Datasets:

ccdv
/

arxiv-summarization

Dataset card Data Studio Files Files and versions Community

Subset (2)

section · 216k rows

Split (3)

train · 203k rows

article stringlengths 0 745k	abstract stringlengths 0 92.7k
let @xmath0 , where @xmath1 denotes the @xmath2th smallest prime . one variant of the prime number theorem states that @xmath8 that is , @xmath9 is approximately @xmath10 on average over @xmath11 . as to the finer questions pertaining to the distribution of primes , we have little more than conjecture in the way of answers . heuristics based on cramr s model suggest that for any given real numbers @xmath12 , @xmath13 however , we do not even know of any specific limit point of the sequence @xmath14 , except for @xmath15 and @xmath16 , the former having been known for just a decade , thanks to the groundbreaking work of goldston pintz yildirim @xcite . ( the latter follows from a 1931 result of westzynthius @xcite . ) this limit point lacuna notwithstanding , hildebrand and maier @xcite showed in 1988 that a positive ( but unspecified ) proportion of nonnegative real numbers are limit points of @xmath14 . more recently , the second author , banks and maynard @xcite have shown that in fact at least @xmath17 of nonnegative real numbers are limit points of @xmath14 . the proof strategy in @xcite incorporates an `` erds rankin '' type construction for producing long gaps between consecutive primes into the celebrated maynard tao sieve , which was originally developed to produce short gaps between primes . more recently still , ford , green , konyagin , and tao @xcite , and ( independently ) maynard @xcite , have settled the notorious `` erds rankin problem '' by showing that @xmath16 is a limit point of @xmath4 , where , @xmath18 and so on . ] @xmath19 . we are therefore motivated to study limit points of @xmath4 . using basically the same strategy as in @xcite , and the work of ford , green , konyagin , and tao @xcite , pintz @xcite has shown that at least @xmath5 of nonnegative real numbers are limit points of @xmath4 . in fact , pintz s result is that the same statement holds if the normalizing function @xmath6 is replaced by any function subject to certain technical conditions that tends to infinity no faster than @xmath6 , for example @xmath20 . ford , green , konyagin , maynard and tao @xcite have actually shown that , for infinitely many @xmath2 , @xmath21 . the purpose of this paper is to fully integrate the work of the five - author paper @xcite into the study of limit points of normalized prime gaps initiated in @xcite . in so doing , we extend the aforementioned result of pintz in three ways . first , we show that the normalizing function @xmath6 may be replaced by any `` reasonable '' function that tends to infinity more slowly than @xmath7 , for example @xmath22 . second , we show that the @xmath5 may conditionally be improved to @xmath23 or even @xmath24 on a certain conjecture concerning the level of distribution of the primes . third , we also consider `` chains '' of normalized , consecutive gaps between primes ( cf . theorem [ thm : chains ] ) . precisely what we mean by a `` reasonable '' function is best explained in context , so we defer the statement of our main result to [ sec : bfm ] ( cf . theorem [ thm : general ] ) . examples of `` reasonable '' functions are @xmath25 , @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath6 , @xmath30 and @xmath31 . any one of these could replace @xmath30 in the following special case of theorem [ thm : general ] , which will serve as a placeholder . [ thm : main ] let @xmath0 , where @xmath1 denotes the @xmath2th smallest prime , and let @xmath32 $ ] denote the set of limit points in @xmath33 $ ] of the sequence @xmath34 , where @xmath35 and @xmath36 is large enough so that @xmath37 . given any five nonnegative real numbers @xmath38 with @xmath39 , we have @xmath40 \ne \emptyset . $ ] as in ( * ? ? ? * corollary 1.2 ) , one may deduce from theorem 1.1 that , with @xmath41 denoting the lebesgue measure on @xmath42 , @xmath43 \cap \lp[r_1 ] ) \ge x/(4(1 + 1/2 + 1/3 + 1/4 ) ) \quad ( x \ge 0),\ ] ] and we mean a positive quantity that tends to zero as @xmath44 tends to infinity . ] ( with an ineffective @xmath45 ) , @xmath46 \cap \lp[r_1 ] ) \ge ( 1 - o(1 ) ) x/4 \quad ( x \to \infty ) . % \]\ ] ] as we will see , assuming a certain variant of the elliott halberstam conjecture ( cf . hypothesis [ hyp : eh ] below ) , one has @xmath47 \ne \emptyset $ ] for any _ three _ nonnegative real numbers @xmath48 , with corresponding improvements to and ( viz . @xmath49 replaces @xmath50 ) . the second author gratefully acknowledges the hospitality of brigham young university , where the work on this paper commenced . we rely heavily on the paper @xcite of ford , green , konyagin , maynard and tao , and we follow their notation and conventions . we explain these conventions here , among others , for completeness sake . 1 . the set of all primes is denoted by @xmath51 ; @xmath52 stand for primes ; @xmath1 denotes the @xmath2th smallest prime . 2 . for @xmath53 , we define @xmath54 . thus , @xmath55 if and only if @xmath56 . a finite set @xmath57 of integers is _ admissible _ if and only if @xmath57 is not a complete set of residues modulo @xmath58 , for any prime @xmath58 . we say an integer is _ @xmath59-smooth _ ( @xmath60 ) if and only if its prime divisors are all less than or equal to @xmath59 . 5 . for @xmath61 and @xmath62 , we define @xmath63 . 6 . for statements @xmath64 , @xmath65 if @xmath64 is true and @xmath66 if @xmath64 is false . the cardinality of a set @xmath67 is denoted by @xmath68 or @xmath69 . the indicator function for @xmath70 ( with @xmath71 clear in context ) is denoted @xmath72 . that is , for @xmath73 , @xmath74 . we write @xmath75 for probability and @xmath76 for expectation . boldface symbols such as @xmath77 or @xmath78 denote random variables , while non - boldface symbols such as @xmath44 or @xmath79 denote their deterministic counterparts . vector - valued random variables are indicated in arrowed boldface , for instance @xmath80 denotes a random tuple of random variables indexed by the set @xmath67 . if @xmath77 takes at most countably many values , we define the _ essential range _ of @xmath77 to be the set of all @xmath44 such that @xmath81 . 11 . if @xmath82 is an event of nonzero probability , @xmath83 for any event @xmath84 , and @xmath85 for any absolutely integrable real - valued random variable @xmath77 . if @xmath86 is another random variable taking at most countably many values , we define the conditional probability @xmath87 to be the random variable that equals @xmath88 on the event @xmath89 for each @xmath90 in the essential range of @xmath86 , and similarly define the conditional expectation @xmath91 to be the random variable that equals @xmath92 on the event @xmath89 . 12 . throughout , @xmath59 denotes a parameter to be thought of as tending to infinity . thus , @xmath45 signifies a quantity that tends to zero as @xmath93 and @xmath94 denotes that @xmath95 . expressions of the form @xmath96 , @xmath97 and @xmath98 all denote that @xmath99 throughout the domain of @xmath44 , for some constant @xmath100 . the constant @xmath101 is to be taken as independent of any parameter unless indicated otherwise , as in @xmath102 for instance , in which @xmath101 depends on @xmath103 and @xmath104 . we write @xmath105 to denote that one can take @xmath106 . we write @xmath107 to denote that @xmath108 . let @xmath59 be a large number and set @xmath109 , where @xmath100 is a certain small constant . ford , green , konyagin , maynard and tao @xcite show that if @xmath110 is large enough , then there exists a vector @xmath111 of residue classes for which @xmath112 \cap \zz \big ) \setminus \ , { \textstyle \bigcup_{p \le cx } c_p \pod{p } } = \emptyset.\ ] ] thus , if @xmath113 is the residue class modulo @xmath114 for which @xmath115 for each @xmath116 , then for @xmath117 with @xmath118 , @xmath119 = \emptyset.\ ] ] we generalize this slightly by proving that if @xmath57 is any set of @xmath120 primes in @xmath121 $ ] with @xmath122 ( say ) , the residue classes may be chosen so that @xmath112 \cap \zz \big ) \setminus \ , { \textstyle \bigcup_{p \le cx } c_p \pod{p } } = \ch\ ] ] and hence @xmath119 = \bp \cap n + \ch.\ ] ] note that @xmath57 , being a set of @xmath122 primes greater than @xmath123 , is admissible . now let @xmath124 be an integer with @xmath125 and let @xmath126 be a partition of @xmath57 into @xmath127 subsets of equal size . as was shown in @xcite , with @xmath128 , a smaller choice of @xmath129 and a minor technical condition on @xmath57 , the maynard tao sieve method establishes that for large @xmath130 , there exists @xmath131 \cap b \pod{w}$ ] and a pair @xmath132 for which @xmath133 provided @xmath120 is sufficiently large and @xmath134 for some small @xmath135 . choosing @xmath136 to be minimal , we obtain a pair of consecutive primes in @xmath137 . we may carefully choose our primes in @xmath57 so that the spacings between them grow faster than @xmath138 but slower than @xmath129 . actually , for reasons related to level of distribution and `` siegel '' zeros , we require that @xmath139 not be a multiple of a certain putative `` exceptional '' modulus less than @xmath140 . the largest prime divisor @xmath141 of this exceptional modulus , if it exists , satisfies @xmath142 . for this reason , we introduce a set @xmath143 of `` unusable '' primes , which has the properties of @xmath144 . their effect is negligible . pintz @xcite has very recently given an elegant simplification of part of this argument , which we take advantage of in this paper , and which shows that one can take @xmath145 . given a large number @xmath59 we define @xmath146 where @xmath101 is a certain ( small ) fixed positive constant , and @xmath147 we then define @xmath148 for vectors of residue classes @xmath149 and @xmath150 , we define sifted sets @xmath151 we note that in view of the prime number theorem ( with suitably strong error term ) and , @xmath152 finally , let @xmath120 be any natural number satisfying @xmath153 by , we may suppose @xmath59 is large enough so that @xmath154 contains at least @xmath120 primes . since @xmath155 , we may also suppose that @xmath156 . we fix any @xmath157 note that @xmath57 , being a set of @xmath120 primes larger than @xmath158 , is an admissible set . [ thm : fgkmt2 ] for all sufficiently large @xmath59 , there exist vectors of residue classes @xmath159 and @xmath160 such that @xmath161 and @xmath162 where the implied constant is absolute . the only difference between theorem [ thm : fgkmt2 ] and ( * ? ? ? * theorem 2 ) is our additional requirement that @xmath161 . unsurprisingly , the proof of theorem [ thm : fgkmt2 ] follows that of ( * ? ? ? * theorem 2 ) very closely , even verbatim in many parts . nevertheless , the details must be checked , and by including them here we are also able to point out the minor differences between the two proofs . we now introduce a set @xmath143 of `` unusable '' primes with the property that for any @xmath163 , @xmath164[p \in \cz ] \frac{1}{p } \ll \frac{1}{p ' } \ll \frac{1}{\log x}.\ ] ] [ cor : thm2 ] let @xmath110 be a sufficiently large but fixed positive constant . for all sufficiently large @xmath59 , there exists a vector of residue classes @xmath165 such that @xmath166\big ) \setminus \ , { \textstyle \bigcup_{p \le cx , \ , p \ , \not\in \ , \cz } } \ , c_p \pod{p}. $ ] we deduce corollary [ cor : thm2 ] from theorem [ thm : fgkmt2 ] with the aid of lemma 5.1 of @xcite , which is as follows . [ lem : bfm5.1 ] let @xmath167 be sets of integers , @xmath168 a set of primes , such that for some @xmath169 , @xmath170 $ ] and @xmath171 . if @xmath172 is admissible then there exists a vector of residue classes @xmath173 such that @xmath174 we choose @xmath59 , @xmath175 and @xmath176 so that the conclusions of theorem [ thm : fgkmt2 ] hold , and work with the enlarged sifted sets @xmath177 note that if @xmath178 , then either @xmath179 , @xmath180 for some @xmath181 or @xmath182 for some @xmath183 . now , @xmath184[n \equiv a_{s } \pod{s } ] 1 \ , + \sum_{p \ , \in \ , \cp \cap \cz } \sums[n \le y][n \equiv b_{p } \pod{p } ] 1 \le y \sums[p \ , \in \ , \cz][p > ( \log x)^{20 } ] \frac{1}{p ' } \ll \frac{y}{(\log x)^{20}}\ ] ] by , and . thus , the elements of @xmath185 that are not in @xmath186 number at most @xmath187 by . we conclude from that @xmath188 let @xmath189 \cup ( z , x/2 ] \}\ ] ] so that @xmath190 is a partition of the primes less than or equal to @xmath59 . we define a vector of residue classes @xmath191 by setting @xmath192 recalling that theorem [ thm : fgkmt2 ] gives @xmath161 , and noting that @xmath57 consists of primes larger than @xmath59 by definition , we see that @xmath193\big ) \setminus \ , { \textstyle \bigcup_{p \le x , \ , p \not\in \cz } } \ , c_p \pod{p}.\ ] ] now , if @xmath194 then @xmath195 and either 1 . @xmath2 is divisible by a prime @xmath196 , 2 . @xmath2 is divisible by a prime @xmath197 $ ] , or 3 . @xmath2 is @xmath198-smooth . in case ( 1 ) , @xmath199 for some @xmath200 by and so , by , @xmath201 is not divisible by any prime in @xmath143 ( provided @xmath59 is sufficiently large , as we assume ) . nor is @xmath201 divisible by any other prime @xmath202 , since such primes are in @xmath203 and @xmath204 for such primes . hence @xmath205 , and @xmath2 must belong to @xmath185 . thus , @xmath206 by . in case ( 2 ) , the prime @xmath141 must belong to @xmath143 , for otherwise @xmath207 . thus , @xmath208 \frac{1}{p } \ll \frac{y}{z } = o\big(\frac{x}{\log x}\big)\ ] ] by and . as shown in ( * ? ? ? * theorem 2 et seq . ) , smooth number estimates give @xmath209 combining , and , we obtain @xmath210 in view of . we may therefore choose our constant @xmath110 to be large enough so that @xmath211 , \ , p \not\in \cz \ } > k + \#\ct.\ ] ] ( the number of primes in @xmath143 that belong to @xmath212 $ ] is negligible , for @xmath213 as can be seen from [ write @xmath214 and sum dyadically ] . ) as @xmath57 is an admissible subset of @xmath215 $ ] , we must conclude , in view of lemma [ lem : bfm5.1 ] , that for sufficiently large @xmath59 there exist residue classes @xmath216 for @xmath217 $ ] , @xmath218 , such that @xmath219 , \ , p \ , \not\in \ , \cz } } \ , c_p \pod{p } = \big(\zz \cap ( x , y]\big ) \setminus \ , { \textstyle \bigcup_{p \le cx , \ , p \ , \not\in \ , \cz } } \ , c_p \pod{p}.\ ] ] in order to prove theorem [ thm : fgkmt2 ] we must first establish the following result , which is analogous to ( * ? ? ? * theorem 4 ) . [ thm : fgkmt4 ] let @xmath59 be sufficiently large . there exists a positive number @xmath110 with @xmath220 the implied constants being independent of @xmath101 , a set of positive integers @xmath221 with @xmath222 , and random vectors @xmath223 and @xmath224 of residue classes @xmath225 and integers @xmath226 respectively , satisfying the following . 1 . for every @xmath227 in the essential range of @xmath228 , we have @xmath229 2 . for every @xmath230 in the essential range of @xmath231 , we have @xmath232 3 . for every @xmath175 in the essential range of @xmath228 , we have @xmath233 where @xmath234 . 4 . with probability @xmath235 , we have @xmath236 5 . call an element @xmath175 in the essential range of @xmath228 `` good '' if , for all but at most @xmath237 elements @xmath238 , one has @xmath239 then @xmath228 is good with probability @xmath235 . let @xmath240 and @xmath57 be as in theorem [ thm : fgkmt4 ] . we set @xmath241 and let @xmath221 be the admissible set with @xmath242 for @xmath243 . our first lemma is a special case of ( * ? ? ? * theorem 5 ) . [ thm : fgkmt5 ] there exist positive quantities @xmath244 and a function @xmath245 supported on @xmath246 $ ] , satisfying the following . 1 . uniformly for @xmath247 , @xmath248 2 . uniformly for @xmath249 and @xmath243 , @xmath250 3 . uniformly for @xmath251 , @xmath252 we choose @xmath253 , @xmath254 and @xmath245 according to lemma [ thm : fgkmt5 ] , and define @xmath255 by setting @xmath256 [ lem : rcb2 ] statements i , ii and iii of lemma [ thm : fgkmt5 ] all hold with @xmath257 in place of @xmath258 , provided the hypothesis @xmath249 in ii is replaced by the hypothesis @xmath259 . we only need to consider ( i ) and ( ii ) . for every @xmath247 we have @xmath260[n \equiv q_j \pod{p } ] w(p , n ) \ll x^{1/3 + o(1)}y / p \ll x^{-2/3 + o(1)}y,\end{aligned}\ ] ] which gives the analog of for @xmath257 in view of . for every @xmath259 and @xmath243 , we see similarly that @xmath261 which gives the analog of . for each @xmath247 , let @xmath262 denote the random integer with probability density @xmath263 for all @xmath61 . using lemma [ lem : rcb2 ] , we verify that @xmath264 @xmath265 and @xmath266 by , the analog of holds with a single prime deleted from @xmath267 . we choose the random vector @xmath268 by selecting each @xmath225 uniformly at random from @xmath269 independently in @xmath270 and independently of the @xmath262 . note , then , that for any random vector @xmath228 , @xmath271 . the sifted set @xmath272 is a random periodic subset of @xmath273 with density @xmath274 let us compare @xmath275 with the quantity @xmath276 defined in @xcite . as @xmath277 ( cf . and ) , one may verify , in a straightforward manner , that @xmath278 consequently , in the estimates that follow , @xmath275 and @xmath279 are interchangeable . as noted in @xcite , by the prime number theorem ( with suitably strong error term ) , and , @xmath280 so by we have @xmath281 also , by we have @xmath282 let @xmath283 and @xmath284 it will transpire that with probability @xmath235 , most primes in @xmath267 lie in @xmath285 . we now define @xmath226 in a slightly complicated way . let @xmath286 suppose we are in the event that @xmath287 . if @xmath288 , we set @xmath289 . otherwise , let @xmath226 be the random integer with @xmath290 with the @xmath226 jointly conditionally independent on the event @xmath287 . ( we easily verify that @xmath291 , so that makes sense . ) [ deduction of theorem [ thm : fgkmt4 ] i iii ] let @xmath247 . we claim that @xmath292 to prove the claim it suffices to show @xmath293 for every @xmath175 . this is easily checked if @xmath294 , since @xmath295 ; otherwise @xmath296 we see that theorem [ thm : fgkmt4 ] ( i ) and ( ii ) hold and we can now prove theorem [ thm : fgkmt4 ] ( iii ) . given @xmath175 in the essential range of @xmath228 , and @xmath238 , we have @xmath297 the right - hand side is @xmath15 if @xmath294 . otherwise , @xmath298 ( here , we have used , and . ) the following lemma is analogous to ( * ? ? ? * lemma 6.1 ) . [ lem : rcb4 ] let @xmath299 be distinct integers of magnitude @xmath300 , @xmath301 , such that @xmath302 . then for all sufficiently large @xmath59 , @xmath303 where the implied constant is absolute . for @xmath304 , let @xmath305 . we have @xmath306 note that for @xmath304 , @xmath307 . let @xmath308 be the set of primes @xmath304 such that either @xmath309 for some @xmath310 or @xmath311 for some @xmath312 . for @xmath310 we have @xmath313 , so @xmath314 has at most @xmath315 prime divisors . similarly , for each @xmath312 and @xmath316 , @xmath317 has at most @xmath315 prime divisors . we see that @xmath318 ( cf . ) , and @xmath319 for @xmath320 we have @xmath321 . thus , @xmath322 [ deduction of theorem [ thm : fgkmt4 ] iv ] let @xmath323 . recalling and , we have @xmath324 let @xmath325 we have @xmath326 from lemma [ lem : rcb4 ] . note that by we have @xmath327 by , the same estimate holds for @xmath328 , and @xmath329 we similarly have @xmath330[q_1 \ne q_2 ] \sigma^2 \bigg(1 + o\bigg(\frac{1}{(\log x)^{16}}\bigg)\bigg ) + \sum_{q \in \cr } \sigma \bigg(1 + o\bigg(\frac{1}{(\log x)^{16}}\bigg)\bigg ) \\ & = \sigma^2(\#\cr)^2 \bigg(1 + o\bigg(\frac{1}{(\log x)^{16}}\bigg)\bigg).\end{aligned}\ ] ] thus , @xmath331 now we use chebyshev s inequality : @xmath332 recalling again we get @xmath333 with probability @xmath334 , and theorem [ thm : fgkmt4 ] ( iv ) follows on recalling that , by , @xmath335 [ lem : rcb5 ] i with probability @xmath336 , @xmath285 contains all but @xmath337 of the primes in @xmath267 . ii we have @xmath338 we have @xmath339 ( for the second step , we supplement lemma [ lem : rcb4 ] with the observation that @xmath340 whenever @xmath341 for some @xmath243 . ) let @xmath342 and @xmath343 be independent random variables having the same probability distribution as @xmath262 . then @xmath344 arguing as above , @xmath345 where @xmath346 is the number of distinct integers @xmath347 ( @xmath348 , @xmath243 ) . now fix @xmath349 . there are less than @xmath350 values of @xmath351 for which @xmath352 . since @xmath353 ( cf . ) , we obtain @xmath354 arguing as in , , @xmath355 thus , with probability @xmath356 , we have @xmath357 that is , @xmath358 . moreover , @xmath359 so with probability @xmath360 , @xmath285 contains all but @xmath337 of the primes @xmath247 . finally , @xmath361 [ lem : rcb6 ] let @xmath249 and let @xmath175 be in the essential range of @xmath228 . then @xmath362 recalling , the left - hand side of is @xmath363 since @xmath364 if @xmath294 , we may rewrite this as @xmath365 and the lemma follows . it is convenient to write @xmath366 [ lem : rcb7 ] we have @xmath367 and @xmath368 we begin with . let @xmath342 and @xmath343 be independent copies of @xmath262 that are also independent of @xmath228 . we observe that for any @xmath369 , and @xmath370 , @xmath371 the left - hand side of is thus @xmath372 ( for @xmath249 , the event @xmath373 is identical to the event @xmath374 , @xmath373 . ) let @xmath375 be the contributions to the right - hand side of from @xmath376 , respectively @xmath377 . fix @xmath378 in @xmath267 with @xmath376 , and fix @xmath379 . the number of distinct integers @xmath380 ( @xmath381 ) is @xmath382 since @xmath383 for @xmath384 . hence @xmath385 ( both sides are @xmath15 if there are @xmath386 with @xmath387 ; otherwise lemma [ lem : rcb4 ] applies . ) we combine with the remark after to obtain @xmath388[][1 ] & = \sums[q \in \cq \setminus \ch][i_1,i_2 \le r ] \sigma^{2r-1 } \bigg(1 + o\bigg(\frac{1}{(\log_2 x)^{10}}\bigg)\bigg ) \bigg(\frac{ux}{2ry}\bigg)^2 \\ & = \bigg(1 + o\bigg(\frac{1}{(\log_2 x)^{10}}\bigg)\bigg ) \frac{\sigma y}{\log x } \bigg(\frac{\sigma^{r-1}ux}{2y}\bigg)^2 . \end{split}\end{aligned}\ ] ] similarly , @xmath389[][2 ] & = \sum_{q \in \cq } \sum_{i_1 = 1}^r \sum_{p_1 \in \cp } \pp(q + ( h_j - h_{i_1})p_1 \in s(\vec{\ba } ) \ , \ , \forall j \le r ) \pp(\tilde{\bn}_p^{(1 ) } = q - h_{i_1}p_1)^2 \\ & \ll x^{-3/5 } \sum_{q \in \cq } \sum_{i_1 = 1}^r \sum_{p_1 \in \cp } \pp(\tilde{\bn}_p^{(1 ) } = q - h_{i_1}p_1 ) \\ & \ll x^{-3/5}(\#\cq),\end{aligned}\ ] ] which together with yields . much the same argument gives for the left - hand side of the expression @xmath390 we now specify that the quantity @xmath110 in theorem [ thm : fgkmt4 ] is @xmath391 so that @xmath392 . [ lem : rcb8 ] with probability @xmath235 , we have @xmath393 for all but at most @xmath394 of the primes @xmath374 . using and lemma [ lem : rcb7 ] , we find that @xmath395 let @xmath396 be the event @xmath397 evidently @xmath398 combining this with , and recalling that @xmath399 ( cf . ) , we obtain @xmath400 [ lem : rcb9 ] we have @xmath401 and @xmath402 the left - hand side of is @xmath403 by lemma [ lem : rcb5 ] . this proves . now , @xmath404 ( here we have used lemma [ lem : rcb4 ] with a familiar argument . ) we obtain on subtracting from . [ deduction of theorem [ thm : fgkmt4 ] v ] in view of lemmas [ lem : rcb6 ] and [ lem : rcb8 ] it suffices to show that with probability @xmath405 , the number of @xmath406 in @xmath407 with @xmath408 is at most @xmath394 . let @xmath139 be the event that holds for more than @xmath394 primes in @xmath407 . then @xmath409 where @xmath410 thus , @xmath411 we require one further lemma for the proof of theorem [ thm : fgkmt2 ] , viz . the following , which is a special case of ( * ? ? ? * corollary 3 ) . [ lem : rcb10 ] let @xmath412 be a set of primes with @xmath413 . for each @xmath247 , let @xmath414 be a random subset of @xmath412 with @xmath415 suppose that for all but at most @xmath416 elements @xmath417 , we have @xmath418 where @xmath110 is independent of @xmath406 and @xmath419 suppose that for any distinct @xmath420 , @xmath421 then for any positive integer @xmath201 with @xmath422 we can find random sets @xmath423 for each @xmath247 such that @xmath424 is either empty or is in the essential range of @xmath414 , and @xmath425 with probability @xmath235 . by , we may choose @xmath101 small enough so that holds . take @xmath426 let @xmath228 and @xmath231 be as in theorem [ thm : fgkmt4 ] . suppose that we are in the probability @xmath235 event that @xmath228 takes a value @xmath175 for which holds . fix some @xmath175 within this event . we apply lemma [ lem : rcb10 ] with @xmath427 , @xmath428 . we need only check the hypothesis . we have @xmath429[p \in \cp ] \pp(q_1 \in e_p(\vec{a } ) ) \le x^{-3/5}\ ] ] ( the sum has at most one term ) . let @xmath430 be the random variables provided by lemma [ lem : rcb10 ] . recalling , @xmath431 with probability @xmath235 . since @xmath432 is either empty or @xmath433 for some random integer @xmath434 , it follows that @xmath435 with probability @xmath235 . the bound follows on setting @xmath436 for a specific @xmath437 for which this bound holds . that @xmath57 is contained in @xmath438 follows from parts ( i ) and ( ii ) of theorem [ thm : fgkmt4 ] . [ def : w ] we consider functions of the form @xmath439 , with @xmath440 . let us say that such a function @xmath441 is `` of the first kind '' if and only if ( i ) it is a strictly increasing bijection , ( ii ) @xmath442 for @xmath443 , ( iii ) @xmath444 as @xmath445 and ( iv ) for @xmath446 , there exists @xmath447 such that @xmath448 as @xmath445 . [ def : sparse ] we consider ( possibly empty ) sets @xmath449 , @xmath450 , of primes less than or equal to @xmath451 . let us that such a set is `` repulsive '' if and only if for any @xmath452 , @xmath453 . given a function @xmath454 with finite support and any arithmetic progression @xmath455 with @xmath456 , we define @xmath457 where @xmath458 is euler s totient function . [ hyp : eh ] fix @xmath459 $ ] and a function @xmath439 of the first kind . for any given @xmath460 and @xmath461 , if @xmath462 is a sufficiently small , fixed number then , for @xmath463 , there is a repulsive subset @xmath464 of the primes less than or equal to @xmath465 such that , with @xmath466 and @xmath467 , we have @xmath468 [ ( r , wz ) = 1 ] [ \text{$r$ squarefree } ] \max_{n \le m \le 2n } \max_{(rw , a ) = 1 } \|\delta(\ind{\bp}\ind{(m , m + n ] } ; a \pod{rw})\| \ll_{\delta , a } \frac{n}{\phi(w)(\log n)^a}.\ ] ] [ thm : bfm4.3 ] fix @xmath459 $ ] and a function @xmath439 of the first kind . suppose that hypothesis [ hyp : eh ] holds . fix a positive integer @xmath79 . in the notation of hypothesis [ hyp : eh ] , if @xmath469 is a sufficiently large integer multiple of @xmath470 , if @xmath471 is sufficiently large and if @xmath472 is sufficiently small , then the following holds for @xmath473 . let @xmath474 $ ] be an admissible set of @xmath120 distinct integers for which @xmath475 is @xmath476-smooth , and let @xmath477 be an integer such that @xmath478 then for any partition @xmath479 of @xmath57 into @xmath470 sets of equal size , there exists some @xmath131 \cap b \pod{w}$ ] , and @xmath480 distinct indices @xmath481 , such that @xmath482 [ thm : bfm4.2 ] hypothesis [ hyp : eh ] , and therefore the statement of theorem [ thm : bfm4.3 ] , holds with @xmath483 and any function @xmath439 of the first kind . [ proof of theorems [ thm : bfm4.2 ] and [ thm : bfm4.3 ] ] that hypothesis [ hyp : eh ] holds with @xmath483 and any function @xmath439 of the first kind is a consequence of lemma 4.1 and theorem 4.2 of @xcite . we prove theorem [ thm : bfm4.3 ] by following pintz s @xcite modification to the proof of theorem 4.3 ( i ) in @xcite . there are many parameters involved and it is important to keep track of their interdependencies . it is also important to note that the implicit constants in all @xmath484-terms are absolute , that is , independent of all parameters . only the unconditional case @xmath483 is considered in @xcite , whereas here we are considering @xmath485 . to do this , we need to note that on hypothesis [ hyp : eh ] , the term @xmath486 may be replaced by @xmath487 on the right - hand side of the inequality in ( * ? ? ? * lemma 4.5 ( iii ) ) . in the proof of this lemma in @xcite , the support of the smooth function @xmath488 , which is @xmath489 $ ] , may be replaced by @xmath490 $ ] , and the rest of the proof may be carried out , mutatis mutandis . as in @xcite , we begin with the following observation . suppose @xmath120 and @xmath127 are positive integers with @xmath125 , and let @xmath126 be a partition of a set @xmath57 of integers into @xmath127 subsets of equal size . suppose also that @xmath491 and @xmath492 are positive real numbers with @xmath493 given an integer @xmath2 , consider the expression @xmath494[h \ne h ' ] \ind{\bp}(n + h)\ind{\bp}(n + h ' ) \big\},\ ] ] where in the double sum each unordered pair @xmath495 with @xmath496 is counted once only . suppose @xmath497 for all but at most @xmath79 of the subsets @xmath498 . then the above expression is at most @xmath499 . consequently , if @xmath500[h \ne h ' ] \ind{\bp}(n + h ) \ind{\bp}(n + h')\ ] ] is positive then @xmath501 for at least @xmath480 of the subsets @xmath498 . note that when @xmath492 is the reciprocal of a positive integer , we have @xmath502 the maximum being attained when @xmath503 and @xmath504 . now , fix @xmath459 $ ] and a function @xmath439 of the first kind . suppose that hypothesis [ hyp : eh ] holds . fix any positive integer @xmath79 and let @xmath505 be the integer satisfying @xmath506 let @xmath507 be a small , fixed quantity to be specified . set @xmath508 let @xmath469 be the integer satisfying @xmath509 finally , let @xmath510 now let @xmath511 , @xmath512 , @xmath473 , @xmath513 and @xmath113 be as in the statement of the theorem . let @xmath126 be any partition of @xmath57 into @xmath127 subsets of equal size . consider the expression @xmath514[n \equiv b \pod{w } ] \hspace{-3pt } \big\ { \sum_{h \in \ch } \ind{\bp}(n + h ) - \frac{1 + m}{2}a - \frac{1}{m } \sum_{j=1}^m \sums[h , h ' \in \ch_j][h \ne h ' ] \hspace{-5pt } \ind{\bp}(n + h)\ind{\bp}(n + h ' ) \big\ } \nu_{\ch}(n),\ ] ] where @xmath515 is the nonnegative weight given by @xmath516[d_i \mid n + h_i \,\ , \forall i \le k ] \lambda_{d_1,\ldots , d_k } \big)^2,\ ] ] and where @xmath517 is the maynard tao sieve as used in @xcite . the aim is to show that @xmath518 , for in that case , by the observation made at the beginning of the proof , there must exist some @xmath131 \cap b \pod{w}$ ] and @xmath480 subsets @xmath498 for which @xmath519 . at this point we invoke lemmas 4.5 and 4.6 in @xcite ( with @xmath486 in the latter replaced by @xmath520 ) . to ease notation define @xmath521 by the relation @xmath522 with @xmath523 and @xmath524 as defined in @xcite . also let @xmath525 . as in @xcite and ( * ( 3.13 ) ) , we find that the relevant estimates yield @xmath526[h \ne h ' ] \frac{m-1}{a}\cdot \frac{(am^2/(m-1))^2}{k^2 } ( 1 + o(\delta + \xi ) ) \\ & = \frac{am^2}{m - 1}(1 + o(\xi ) ) - \frac{1 + m}{2}a - \binom{k / m}{2 } \frac{am^4(1 + o(\delta + \xi))}{k^2(m-1 ) } .\end{aligned}\ ] ] recalling that @xmath527 , we see that @xmath528 and hence @xmath529 we also have @xmath530 , so @xmath531 . we therefore have @xmath532 taking @xmath533 sufficiently small gives @xmath534 , and hence @xmath518 , as desired . recall definition [ def : w ] , in which functions `` of the first kind '' are introduced . we now define a second kind of function . [ def:2ndkind ] we consider functions @xmath535 , with @xmath536 . let us say that such a function @xmath537 is `` of the second kind '' if and only if ( i ) it is a strictly increasing bijection , ( ii ) @xmath538 as @xmath93 and ( iii ) for any @xmath539 , @xmath540 as @xmath93 . [ thm : general ] fix @xmath459 $ ] and a function @xmath439 of the first kind , and suppose hypothesis [ hyp : eh ] holds . fix a function @xmath541 of the second kind and let @xmath542 . let @xmath0 , where @xmath1 denotes the @xmath2th smallest prime , and let @xmath543 $ ] denote the set of limit points in @xmath33 $ ] of the sequence @xmath544 . then given any @xmath545 nonnegative real numbers @xmath546 with @xmath547 we have @xmath548 \ne \emptyset.\ ] ] consequently , letting @xmath41 denote the lebesgue measure on @xmath42 , we have @xmath549 \cap \lp[f ] ) \ge c_1(\theta)x \quad ( x \ge 0)\ ] ] and @xmath550 \cap \lp[f ] ) \ge ( 1 - o(1 ) ) c_2(\theta)x \quad ( x \to \infty),\ ] ] where @xmath551 in view of theorem [ thm : bfm4.2 ] , we may unconditionally apply theorem [ thm : general ] with @xmath483 and any function @xmath552 of the first kind . let @xmath553 be given by @xmath554 , and let @xmath555 be given by @xmath556 . then @xmath441 is of the first kind , @xmath537 is of the second kind and @xmath557 for @xmath558 . [ lem : goodktuple ] let @xmath120 be a natural number and let @xmath559 be a partition of @xmath120 . let @xmath59 and @xmath129 be real numbers such that @xmath560 . if @xmath59 is sufficiently large , then for any @xmath127 possibly overlapping subintervals @xmath561 \subseteq ( x , y]$ ] , @xmath562 , there exist @xmath127 pairwise disjoint sets of primes @xmath563 $ ] with @xmath564 , such that if @xmath565 , then @xmath566 is @xmath59-smooth . for any @xmath127 sets @xmath567 with @xmath568 , @xmath562 , there exist @xmath127 pairwise disjoint sets @xmath569 such that @xmath570 and @xmath564 , @xmath562 . for the sets @xmath567 , let @xmath571 be the integer satisfying @xmath572 . as @xmath573 and @xmath574 , a suitably strong version of the prime number theorem for arithmetic progressions ( cf . * ( 4 ) ) ) yields , for @xmath561 \subseteq ( x , y]$ ] , @xmath575[p \equiv 1 \pod{d } ] 1 = \frac{x}{\phi(d)(\log x)^2 } + o\bigg(\frac{y}{(\log y)^5}\bigg ) \ge \frac{x}{(\log x)^3 } + o\bigg(\frac{x}{(\log x)^4}\bigg).\ ] ] as @xmath122 , we see that if @xmath59 is sufficiently large , then there are at least @xmath120 primes @xmath576 in @xmath561 $ ] . if @xmath577 are any two such primes , as @xmath578 and @xmath579 , any prime divisor @xmath58 of @xmath580 must either divide @xmath573 or be less than or equal to @xmath581 . hence @xmath580 is @xmath59-smooth . fix @xmath459 $ ] and a function @xmath439 of the first kind , and suppose hypothesis [ hyp : eh ] holds . in accordance with theorem [ thm : bfm4.3 ] ( in which we take @xmath582 ) , let @xmath583 be a sufficiently large integer multiple of @xmath545 , let @xmath471 be sufficiently large and @xmath584 and @xmath585 be sufficiently small . in accordance with corollary [ cor : thm2 ] , let @xmath110 be a sufficiently large but fixed positive constant and let @xmath59 be a sufficiently large number . suppose , as we may , that @xmath586 , and ( cf . definition [ def : w ] ( i ) ) set @xmath587 thus , @xmath588 and @xmath130 tends to infinity with @xmath59 . suppose @xmath59 is large enough so that in accordance with theorem [ thm : bfm4.3 ] , @xmath473 . fix a function @xmath535 of the second kind and consider the intervals @xmath593 , \quad i \le \lceil ( 2/\theta ) \rceil + 1.\ ] ] recall that @xmath594 , where @xmath100 is a certain constant ( cf . ) . by definition [ def:2ndkind ] ( ii ) we have @xmath595 and @xmath596 . suppose , then , that @xmath59 is large enough ( in terms of @xmath597 ) so that the intervals in are all contained in @xmath121 $ ] . in accordance with lemma [ lem : goodktuple ] , choose @xmath545 pairwise disjoint sets of primes @xmath569 of equal size , with @xmath598 , \quad i \le \lceil ( 2/\theta ) \rceil + 1.\ ] ] thus , letting @xmath599 we have that @xmath566 is @xmath59-smooth , and hence @xmath476-smooth ( we may suppose that @xmath600 ) . as @xmath120 is fixed we may of course suppose that @xmath122 and @xmath156 , so that is satisfied and @xmath57 , being a set of @xmath120 primes larger than @xmath158 , is admissible . we may of course also suppose that @xmath601 , so that @xmath602 $ ] . thus , @xmath57 satisfies each of the hypotheses of theorem [ thm : bfm4.3 ] . let @xmath603 be as in hypothesis [ hyp : eh ] , so that @xmath603 is repulsive ( cf . definition [ def : sparse ] ) , and note that since @xmath604 ( cf . definition [ def : w ] ( ii ) ) , @xmath605[p \ge p ' ] \frac{1}{p } \ll \frac{1}{p ' } \ll \frac{1}{\log_2 n^{\eta } } \le \frac{1}{\log f_1(n^{\eta } ) } \le \frac{1}{\log x}.\ ] ] thus , is satisfied with @xmath606 . therefore , by corollary [ cor : thm2 ] there exists a vector of residue classes @xmath607 such that @xmath608\big ) \setminus \ , { \textstyle \bigcup_{p \le f_1(n^{\eta } ) , \ , p \ , \not\in \ , \cz(n^{\eta } ) } } \ , c_p \pod{p}.\ ] ] each hypothesis of theorem [ thm : bfm4.3 ] now accounted for , we conclude that there is some @xmath131 \cap b \pod{w}$ ] , and a pair of indices @xmath614 , @xmath615 , such that @xmath616 if there are more than two such indices , we take @xmath617 to be minimal . thus , if @xmath58 is the largest prime in @xmath618 and @xmath141 is the smallest prime in @xmath619 , then @xmath58 and @xmath141 are _ consecutive _ , that is , @xmath620 and @xmath621 for some @xmath622 . indeed , by and the definition of @xmath113 , for any @xmath117 with @xmath623 we have @xmath119 = \bp \cap n + \ch.\ ] ] by and , and since @xmath595 , we have @xmath624 since there are only @xmath625 distinct pairs of indices from which @xmath626 and @xmath627 may be chosen , we deduce that there exists a single pair @xmath615 such that , for arbitrarily large @xmath130 , we have @xmath628 for some pair of consecutive primes @xmath629 \subseteq ( n,3n]$ ] . finally , using definition [ def : w ] ( i ) , ( iii ) and ( iv ) and definition [ def:2ndkind ] ( i ) and ( iii ) , we find that @xmath630 we conclude that @xmath631 we deduce and by using the argument of ( * ? ? ? * corollary 1.2 ) . as in ( * theorem 1.3 ) , we may also consider `` chains '' of normalized , consecutive gaps between primes . using essentially the same argument as above , but using ( the unconditional ) theorem 4.3 ( ii ) of @xcite in place of theorem [ thm : bfm4.3 ] , one may verify the following result . [ thm : chains ] fix any integer @xmath79 with @xmath632 . fix functions @xmath439 and @xmath541 of the first and second kinds respectively , and let @xmath542 . let @xmath0 , where @xmath1 denotes the @xmath2th smallest prime , and let @xmath633 $ ] denote the set of limit points in @xmath33^a$ ] of the sequence of `` chains '' @xmath634 for @xmath635 . given @xmath636 , let @xmath637 be the set @xmath638 for any @xmath639 nonnegative real numbers @xmath640 , we have @xmath641^a \ne \emptyset.\ ] ] let us call a function `` reasonable '' if it is of the form @xmath642 , where @xmath441 is a function of the first kind and @xmath537 is a function of the second kind . theorem [ thm : chains ] shows that for any @xmath79 there are infinitely many chains of consecutive prime gaps with @xmath643 for any reasonable function @xmath644 . there are reasonable functions @xmath644 for which @xmath645 tends to @xmath15 arbitrarily slowly ( recall that @xmath646 is the erds rankin function ) . we believe that in a forthcoming paper @xcite , ford , maynard and tao show that for any @xmath79 there are infinitely many chains of consecutive prime gaps with @xmath647 . `` the distribution of prime numbers . '' pp.5983 in : granville , a. and z. rudnick ( eds . ) . _ equidistribution in number theory , an introduction . _ nato sci . ii math . springer , dordrech , 2007 .	let @xmath0 , where @xmath1 denotes the @xmath2th smallest prime , and let @xmath3 ( the `` erds rankin '' function ) . we consider the sequence @xmath4 of normalized prime gaps , and show that its limit point set contains at least @xmath5 of nonnegative real numbers . we also show that the same result holds if @xmath6 is replaced by any `` reasonable '' function that tends to infinity more slowly than @xmath7 . we also consider `` chains '' of normalized prime gaps . our proof combines breakthrough work of maynard and tao on bounded gaps between primes with subsequent developments of ford , green , konyagin , maynard and tao on long gaps between consecutive primes .
ever since the reports of the detection of two cosmic ray showers of energy well above @xmath3 ev @xcite the origin and the nature of such events have been subjects of strong interest and intense discussion . it is not only very difficult @xcite to extend our understanding of particle acceleration to such extraordinarily high energies but the propagation of these particles in the microwave background and possibly other universal radiation fields restricts the distance to their potential sources to several tens of mpc . conservatively minded astrophysicists are looking for astrophysical sources which may contain the environment necessary for stochastic particle acceleration to energies in excess of @xmath3 ev . powerful ( frii ) radio galaxies @xcite have been suggested as possible sources . if this suggestion were true , the highest energy cosmic rays ( hecr ) would be most likely protons , reflecting the composition of the matter that is available for injection in the termination shocks of frii jets . others @xcite search for powerful astrophysical sources in the cosmologically nearby universe . hecr then could also be heavier nuclei , for which the acceleration is less demanding . the propagation of heavy nuclei on short distances ( o(10 ) mpc ) without huge energy loss is possible . some cosmologists relate the origin of hecr to topological defects @xcite . topological defects ( td ) scenarios avoid the problems of particle acceleration since they are based on ` top down ' evolution . very massive ( @xmath4 ev ) x particles are emitted by the topological defects that later decay into baryons and mesons of lower energy . most of the energy is eventually carried by @xmath0rays and neutrinos , that are products of meson decay . detected hecr would then most likely be @xmath0rays . most radically , the origin of hecr has been related to those of gamma ray bursts @xcite , replacing two extremely luminous mysteries with a single one . in such scenarios hecr are most likely to be again protons . we may not be able to observe the sources of hecr since every source might only emit a single observed ultrahigh energy particle . the nature , the type of the particle that interacted in the atmosphere to generate these giant air showers , could be the key to the understanding the origin of the highest energy cosmic rays . the current experimental evidence on the nature of hecr is not conclusive . the fly s eye experiment , for example , has reported correlated changes in the spectra and the composition of the ultra high energy cosmic rays @xcite . the analysis of the fly s eye experimental statistics suggests that a change of the chemical composition of the cosmic rays from heavy nuclei to protons at @xmath5 ev is accompanied by a change of the spectral index of the cosmic ray energy spectrum . one may then conclude that the hecr are protons . the other currently running air shower experiment , agasa , does not observe @xcite such a correlation . a re analysis of the archival data from the sugar experiment @xcite makes the opposite conclusion a large fraction of the highest energy showers seem to be generated by heavy nuclei .. a correlation between the arrival directions of hecr with energy @xmath6 ev with the supergalactic plane , that contains most of the galaxies of redshift @xmath7 0.03 , has been reported @xcite . the agasa experiment has also observed similar correlation in their data set @xcite , although not fully consistent with the conclusions of ref . @xcite . on the other hand the fly s eye experiment does not see such a correlation ( p. sommers for the fly s eye group , _ private communication _ ) . it also has not been observed in the sugar data @xcite . even if confirmed in the future , a correlation with the structure of the local universe would not answer the question of the nature of hecr . if topological defects are seeds for galaxy formation most powerful galaxies and td would have similar distribution and td and astrophysical scenarios of the origin of hecr are indistinguishable . the profile of the 3@xmath8 ev shower detected by the fly s eye develops higher in the atmosphere than expected for either proton or @xmath0ray showers of that energy @xcite . the highest energy shower seen by the agasa experiment ( @xmath9 ev ) exhibits , apart from its energy , features that are typical for most of the high energy showers . the currently existing air shower arrays can not drastically increase the experimental statistics and the hope for answering the important questions for the nature and origin of hecr is in the construction of much bigger shower arrays , such as the auger project @xcite . even with auger , however , the nature of hecr will be difficult to study . shower parameters are subject of strong intrinsic fluctuations and the cross sections that govern inelastic interactions at @xmath10 = 100 tev are not well enough known . at lower energy ( @xmath11 ev ) showers generated by heavy nuclei , protons and @xmath0rays could be at least statistically distinguished by their muon content . @xmath0ray showers have on the average @xmath12% of the muon content of proton showers of the same energy @xcite . at ultrahigh energies such approach may not be possible calculations of the muon content of the @xmath0ray induced showers predict that the fraction of gev muons could be even higher than in proton generated showers @xcite . we suggest a different approach to the study of the nature of the cosmic rays with energy above @xmath13 ev to prove ( or disprove ) that hecr are @xmath0rays by observing their interactions with the geomagnetic field . while protons and heavier nuclei are not affected by the geomagnetic field , ultra high energy @xmath0rays interact on it to produce @xmath14 pairs . the electrons themselves quickly lose their energy through magnetic bremsstrahlung ( synchrotron radiation ) before they enter the atmosphere of the earth . air showers are thus replaced by ` magnetic + atmospheric ' showers that start far away from the surface of the earth and are absorbed faster compared to usual air showers . with high experimental statistics one can observe the interactions of ultra high energy @xmath0rays with the geomagnetic field by a study of the shower arrival direction in geographical coordinates . if the detected showers do not show signs of interactions with the geomagnetic field , the suggestions for @xmath0ray nature of hecr could be proven wrong . this article is organized in the following way . section 2 . gives a brief discussion of the photon and electron interactions on magnetic fields and of the structure of the geomagnetic field . section 3 . describes a calculation of the ` geomagnetic + atmospheric ' cascades and gives some general results of that calculation . section 4 . calculates shower parameters that could be used to confirm the @xmath0ray origin of hecr and section 5 . contains the conclusions from this research interactions of photons , and especially of electrons , on magnetic fields have been exhaustively studied because of all the problems they create in particle accelerators . the theoretical and some experimental knowledge is reviewed by t. erber in ref . @xcite . magnetic pair production is guided by the parameter @xmath15 [ h\nu / mc^2 ] [ b_\perp / b_{cr}]$ ] , where @xmath16 gauss and @xmath17 is the component of the magnetic field that is normal to the @xmath0ray trajectory . the @xmath0ray attenuation coefficient , i.e. the fraction of photons that undergo pair production in magnetic field of strength @xmath17 per unit distance is given by @xmath18 the maximum attenuation is reached at @xmath0ray energy of @xmath19 while the cross section of the process is linearly proportional to the magnetic field strength @xmath17 . similarly the magnetic bremsstrahlung ( synchrotron radiation ) is guided by @xmath20 [ b_\perp / b_{cr}]$ ] . the radiation emitted by an electron of energy @xmath21 in magnetic field @xmath17 per unit distance is distributed as @xmath22 where @xmath23[1 + h\nu / e_e]/3\upsilon_e$ ] . to demonstrate the strength of the @xmath0ray interactions in the geomagnetic field we show in fig . [ fig1 ] the distributions of the distances from the surface of the earth at which @xmath0rays of different energy pair produce . the @xmath0ray trajectory is taken to be normal to the field lines of a magnetic dipole centered at the center of the earth with magnetic moment of @xmath24 gauss / m . one could see that the @xmath0rays of the energies of interest interact in a relatively narrow range of distances not further than 3@xmath25 . the narrow peak plotted at altitude of 20 km represents @xmath0rays that survive , i.e. interact in the atmosphere before they interact in the geomagnetic field . 12% of the @xmath0rays with energy @xmath3 ev ( and none at higher energy ) survive . the spectra of the @xmath0rays emitted in magnetic bremsstrahlung depends quite strongly on the magnetic field strength . for strong fields the energy distribution of the secondary photons is quite flat . [ fig2 ] shows the energy loss of @xmath3 electrons in magnetic fields of strength @xmath26 = 0.5 , 1 , 1.5 , etc . gauss as a function of the secondary photon energy . in the dipole field model described above a field of 0.1 gauss corresponds to a distance of 0.468 @xmath25 above the surface of the earth , and 0.032 gauss to 1.15 @xmath25 . these distances cover much of the primary @xmath0rays interaction range shown in fig . since lower energy @xmath0rays pair produce close to the earth , the magnetic bremsstrahlung of their secondary electrons is harder . the energy spectra of the @xmath0rays in the bunch that enters the atmosphere after pair production and magnetic bremsstrahlung tend to be almost independent of the primary @xmath0ray energy @xmath27 . gamma rays arriving at any experimental location under different zenith ( @xmath28 ) and azimuthal ( @xmath29 ) angle will see a different geomagnetic field . they will thus cascade differently before reaching the atmosphere . at small @xmath28 , close to the vertical direction , the variation with @xmath29 is insignificant . at relatively large @xmath28 , more than 30@xmath30 , the field strength for most locations changes by factors of 3 or more for different values of @xmath29 . a more quantitative calculation of the strength of the field encountered by the incoming @xmath0ray is trivial for any model of the geomagnetic field but has to be performed for each location , @xmath28 and @xmath29 separately . we have attempted to obtain a slightly more general result for several experimental locations . fig . [ fig3 ] shows the transverse component ( @xmath17 ) of the geomagnetic field as a function of the azimuthal angle at which it arrives to the detector . since the @xmath29 variations for different zenith angles @xmath28 have the same aspect , we have integrated over @xmath28 from 0 to 60@xmath30 , weighting the field values with the solid angle . the 1991 international geomagnetic reference field model ( igrf ) is used for this calculation . four of the locations for which @xmath17 is shown are in the northern hemisphere and only one ( sydney , shown with dash dash ) is south of the equator . since the smallest @xmath17 is seen in the direction of the magnetic pole that is closer to each location , northern and southern locations have opposite field strength dependence on @xmath29 . at the moderate latitudes of these detector locations , the detailed differences between the northern hemisphere detectors are minor . ( a detector located at the geomagnetic equator would have a symmetric response to geomagnetic north and south directions ) . the difference between the maximum and minimum field strengths is almost a factor of 5 . for each one of these detectors , as well as for any other detector location , one could determine a region in azimuth , where the field strength is the lowest and the incoming @xmath0rays would be affected minimally by the geomagnetic field and a region where the effect of the geomagnetic field is at maximum . for the location of sydney , e.g. @xmath0rays arriving with @xmath31 would see @xmath320.02 gauss at a distance of 1 @xmath25 and @xmath0rays with @xmath33 would see more than 0.04 gauss at the same distance . the idea is that @xmath0ray fluxes arriving from these two regions may have observable characteristics that are different enough to be distinguished experimentally . we continue to study the cascading of ultra high energy gamma rays in geomagnetic fields with different strength , corresponding to these two regions . we simulate the electromagnetic cascading in the geomagnetic field by injecting @xmath0rays of energy @xmath27 at a distance of 5 @xmath25 from the surface of the earth on a trajectory with angle @xmath28 relative to the vertical direction at the intersection with the surface . the @xmath0ray is propagated with a stepsize @xmath34 ( from 1 to 10 km ) until the @xmath0ray pair produces or reaches the atmosphere . the atmosphere is defined to be at altitude of 20 km above the earth s surface . gamma rays that reach the atmosphere ` survive ' and interact in the atmosphere to produce air showers with their original injection energy . if the @xmath0ray produces an electron positron pair , the pair electrons are followed in similar way , by calculating their radiation spectrum on every step of propagation . the synchrotron @xmath0rays are tabulated in energy , starting at @xmath35 ev . the assumption here is that secondary @xmath0rays of energy less than @xmath35 ev do not contribute significantly to the cascades that are observed deep in the atmosphere . this lower energy end of the magnetic bremsstrahlung spectrum , as well as the electrons of energy below @xmath35 ev that enter the atmosphere , contain always less than 2% of the primary @xmath0ray . each particle produced in the geomagnetic field , as well as the ` surviving ' primary @xmath0rays then generate atmospheric cascades . the profiles of these cascades are added up to calculate the composite shower profile , generated in the atmosphere by the injected primary @xmath0ray or the products of its interaction in the geomagnetic field . the actual calculation is performed using the dipole magnetic field model with magnetic moment of @xmath24 gauss / m with two scale factors of 0.25 ( low field ) and 1.25 ( high field ) , which generate magnetic field strengths approximately equal to the maximum and minimum values shown in fig . [ fig3 ] . to study the ` survival ' probability in these two field models we made calculations for two extreme zenith angles : @xmath28 = 0@xmath30 and 60@xmath30 , which is the maximum zenith angle at which air showers can be reliably detected and analyzed . [ fig4 ] shows the survival probabilities at high ( lefthand strip ) and low magnetic field strengths . the lefthand boundary of each strip corresponds to propagation at @xmath28 = 60@xmath30 and the righthand boundary is for @xmath28 = 0@xmath30 . @xmath0rays approaching the earth at higher zenith angles spend significantly more time in higher geomagnetic field strengths and have a higher interaction probability . the lefthand edge of the high field strip and the righthand edge of the low field strip practically bracket the survival probability space for @xmath0rays approaching any location at the earth surface with zenith angles smaller than 60@xmath30 . @xmath0rays arriving at higher angles may be absorbed faster . several calculations of the @xmath0ray cascading in the geomagnetic field have been previously performed @xcite . our results are in a good agreement with the main results of all of them . our calculation is generally a refinement of previous ones , which nevertheless reveals some practically important features in the cascading process . previous calculations conclude that there will be a ` cutoff ' in the energy spectrum of the @xmath0rays that reach the atmosphere , because of the very soft spectrum of the secondary photons , generated by magnetic bremsstrahlung . this conclusion is partially due to the relatively rough treatment and low statistics in the previous work . [ fig5 ] shows the number of secondary @xmath0rays and the energy that they carry . @xmath27 = @xmath36 ev in this example . although the number of secondary @xmath0rays of energy above @xmath13 ev is on the average only 6.3 , they carry 48% of the primary energy . this is also important for the development of the subsequent air showers , because at energies above @xmath13 ev the lpm effect @xcite , which suppresses the electromagnetic cross sections at high energy and slows the development of the air showers , becomes important in air . gamma rays of energy above @xmath13 ev , if they do exist , would only be detectable by giant air shower arrays located on the surface of the earth . air shower arrays consist of a large number of counters that trigger in coincidence when the shower from arrives . the shower direction is determined by the arrival time of the shower front at the different counters . a fit of the density in the separate counters reconstructs the total number of shower particles , the showers size @xmath37 , which is then used to determine the primary energy . the output of our monte carlo simulation includes the shower sizes calculated for several atmospheric depths from the cascading of all secondary ( and primary , if the injected gamma rays did not pair produce ) @xmath0rays in the atmosphere . the profiles from individual secondary @xmath0rays of energy above @xmath38 ev are calculated with an account for the lpm effect , although the effect is not significant below @xmath13 ev . the depths are arbitrarily chosen to include a realistic range for typical large air shower experiment and correspond to an array at vertical depth of 860 g / cm@xmath2 and zenith angles with @xmath40 = 0.9 , 0.8 , 0.7 , 0.6 and 0.5 . because the development of purely electromagnetic showers in the atmosphere ( apart from their muon content ) does not depend significantly on the atmospheric density profile , the examples given below could be scaled also to different altitudes and zenith angles . [ fig6 ] shows a general and important shower parameter the average size ( @xmath41 ) generated by @xmath0rays of different energy . the solid lines are for low field ( scale factor of 0.25 ) and the dashed lines are for high field ( scale factor of 1.25 ) . from top to bottom the lines show @xmath42 at the 5 different depths of 956 , 1075 , 1229 , 1433 and 1720 g / cm@xmath2 . except for the deepest observation level , @xmath37 is multiplied by the factor shown by each curve to make the figure readable . in the absence of interactions in the geomagnetic field , and for lower @xmath27 , the shower size have a power low dependence on @xmath43 , @xmath44 with @xmath451 . the power law index @xmath46 depends on the column density between the depth of the shower maximum @xmath47 and the detector . the size at maximum @xmath48 is exactly proportional to @xmath27 and @xmath46 is bigger than unity because the depth of shower maximum grows with energy as @xmath49 radiation lengths ( 1 r.l . = 37.1 g / cm@xmath2 in air ) . the dependence shown in fig . [ fig6 ] is more complicated because in this energy range showers are already at or before their maximum development at some of the shallower observation levels . the role of the magnetic field strength on the @xmath37 dependence on @xmath27 is easier to understand for the deepest levels of observation . compare , for example , the two curves for depth of 1720 g / cm@xmath2 with the @xmath0ray survival probability of fig . [ fig4 ] . at low energy , where there are no interactions on the geomagnetic field , the two curves are the same . the solid curve ( low field ) starts bending at @xmath0ray energy @xmath50 ev where the primary @xmath0rays start interacting in the geomagnetic field . because of these interactions the primary @xmath0ray is replaced by a bunch of @xmath0rays of lower energy . the composite shower reaches maximum at shallower atmospheric depths and is significantly absorbed at the deep observation level . the same happens at energy lower by about one order of magnitude in the high field case . although it is outside of the energy range of fig . [ fig6 ] , at some higher energy , where all @xmath0rays interact on the geomagnetic field , the two curves will join again . to explain the behaviour at the shallow observation levels one has to take into account some of the details of the cascading in the geomagnetic field , namely the shape of the energy spectra of the secondary photons as a function of field strength , which is shown in fig . although the primary @xmath0rays interact in the same way , in the high field case the energy spectra of the secondary @xmath0rays are harder , hard enough to generate showers that are not absorbed at the level of 956 g / cm@xmath2 . one could hardly see a tiny deviation of the strong field ( dashed ) curve in the region of @xmath51 ev . at higher energies the secondaries are energetic enough to produce @xmath37 dependence very close to a power law . when the primary @xmath0rays start interacting in the low field , however , the picture is slightly different . the secondary @xmath0ray spectra are softer , the composite showers reach maxima at shallower depths and are correspondingly absorbed when they reach the observation level . the two curves will join asymptotically . all other levels show intermediate behaviour where the relation between the depth of observation and @xmath47 also contributes to the exact shape of the curve . [ fig6 ] shows the strong differences in the observable parameter @xmath37 which is introduced by the strength of the geomagnetic field . it can not be used , however , for analysis of experimental data because @xmath27 is not a directly measurable parameter . what experiments can do , and usually do , is to produce a spectrum of the measured shower sizes @xmath37 . such spectra for the three deeper observation levels are shown in fig . the solid histogram corresponds to the low field and the dashed one to the high field case . the histograms are result of a simulation , where @xmath27 is sampled from a @xmath52 differential primary spectrum between @xmath13 and @xmath53 ev . at low @xmath37 the spectra are always higher for the low field case , including the two observation levels that are not shown in fig . [ fig7 ] . at the high @xmath37 side and for shallow observation levels the high field case shows higher spectrum , as could be expected by the results shown in fig . [ fig6 ] and as seen for the shallowest level plotted in fig . the biggest difference is at the deepest observation level , where the spectra are different as much as a factor of 10 . the differences between the size spectra decreases for shallower observation levels , and is probably not detectable for the two shallowest levels , which are not shown . the size spectra of fig . [ fig7 ] show that it is possible to detect the difference between a flux of @xmath0rays that reach the earth after cascading in geomagnetic field of different effective strength . in practical terms this means that any experiment that is able to collect large enough experimental statistics should see different @xmath37 spectra in different azimuthal directions if hecr are indeed @xmath0rays . we have not attempted to look for this effect in the existing experimental statistics , because it is not large enough to reveal such effects . the auger project @xcite is an entirely different story . it proposes the construction of two air shower arrays , at least 3 000 km@xmath2 each , in the northern and in the southern hemisphere . for comparison , the area of the largest current detector ( agasa ) is 100 km@xmath2 . each one of these detectors will have the collecting power of more than 5 000 showers above @xmath13 ev per year . if the hecr primaries are @xmath0rays the norther hemisphere detector should see spectra similar to the low field case of fig . [ fig7 ] in northern direction ( the exact direction and value of the minimum field depends on the location of the array ) and the strong field case in showers coming from south . the southern detector will have the opposite effect . if this were the case , the difference of the size spectra would prove that the hecr are gamma rays . the actual effects could even be stronger than shown in fig . [ fig7 ] , because the simulation on which it is based propagated all primary @xmath0rays along the vertical ( @xmath28 = 0@xmath30 ) direction . in fact , as shown in fig . [ fig4 ] , the increase of the interaction probability in the geomagnetic field increases by a non negligible factor when the exact particle trajectory is accounted for . our air shower simulation also does not account for the magnetic bremsstrahlung of the shower electrons which at high @xmath21 and low atmospheric density @xmath54 g/@xmath55 could be important and could accelerate the shower development . in principle the interplanetary magnetic field has to be added to the ` target ' magnetic field . gamma ray arriving from a cone centered on the sun would be absorbed far away from the earth and possibly not detectable . the sun could thus be visibly in ultra high energy @xmath0-rays . the exact dimensions of the region where @xmath0rays are absorbed in pair production on the solar magnetic field , carries valuable information on the magnetic field in the vicinity of the sun . this is an interesting although purely academic problem , because the statistics of such events is always going to be negligible . although we have not done it for this paper , there will be effects , similar to the @xmath37 ones , on the muon content of the @xmath0ray initiated air showers . it is well known that at these extremely high energies the number of soft muons ( 0.3 to 2 gev ) in @xmath0 initiated showers is comparable to this of hadronic showers @xcite . the number of soft muons has an @xmath27 dependence very similar to @xmath37 , because the low energy muons decay readily when @xmath47 is distant from the observation level . the decay length of 1 gev muons is @xmath566 km . a picture similar to the @xmath37 spectra in fig . [ fig7 ] will develop as a result of the cascading in geomagnetic fields of different effective strength . the major difference between the behaviour of the electron size and the muon size is that @xmath37 attenuates as a function of the column density , while @xmath57 attenuates as a function of the distance , i.e. @xmath57 will depend strongly on the shower zenith angle @xmath28 . the study of the two major components of the giant air showers can reveal the nature of the highest energy cosmic rays . if the specific dependence on the shower arrival direction is observed , then the highest energy cosmic rays are @xmath0rays . a non observation of this effect would leave us with the choice between protons and heavy nuclei . * acknowledgements . * ts is grateful to a.a . watson for inspiring discussions on the subject of giant air shower . the work of ts is supported in part by the u.s . department of energy under contract de - fg02 - 91er40626 . hpv is thankful to us nsf for partial support of his visit to u.s . where this work was conceived , and to the bartol research institute for its hospitality . et al . , phys . * 71 * , 3401 ( 1993 ) . n. hayashida _ lett . _ * 73 * , 3491 ( 1994 ) . a.m. hillas , _ ann . revs . . astrophys . _ , * 22 * , 425 ( 1984 ) . rachen and p.l . biermann , _ astron . astrophys . _ * 272 * , 161 ( 1993 ) ; j.p . rachen , t. stanev and p.l . biermann , _ astron . astroph . _ * 273 * , 377 ( 1993 ) . al dargazelli _ et al . g : nucl . part . phys _ , submitted ; j. szabelski , j. wdowczyk & a.w . wolfendale , _ j. phys.g : nucl.phys._ * 12 * , 1443 ( 1986 ) . p. bhattacharjee , c.t . hill and d.n . schramm , _ phys . _ , * 69 * , 567 ( 1992 ) ; g. sigl , d.n . schramm and p. bhattacharjee , _ astropart . * 2 * , 401 ( 1994 ) . e. waxman , _ phys . rev . letters _ , * 75 * , 386 ( 1995 ) . m. vietri , _ ap . j. _ * 453 * , 883 ( 1995 ) . m. milgrom and v. usov , _ ap . j. _ * 449 * , l37 ( 1995 ) . d.j . bird _ 71 , 3401 ( 1993 ) . n. hayashida _ et al . j. phys . _ , * 21 * 1101 ( 1995 ) . x. chi , j. wdowczyk & a.w . wolfendale , _ j. phys . g : nucl . part . _ * 18 * ( 1992 ) and references therein . t. stanev _ et al . _ , * 75 * , 3056 ( 1995 ) . n. hayashida _ et al . _ ( the agasa collaboration ) _ phys . lett _ , submitted . kewley , r.w . clay & b.r . dawson , _ astropart . _ , in print . f. halzen _ * 3 * , 151 ( 1995 ) . see `` cosmic rays above 10@xmath58 ev 1992 '' , eds . m. boratav _ et al . * 28b * ( 1992 ) ; `` the pierre auger project '' , design report , the auger collaboration ( 1995 ) . gaisser _ et al . , _ , * d43 * , 314 ( 1991 ) . mccomb , r.j . protheroe & k.e . phys _ , * 5 * , 1613 ( 979 ) . aharonian , b.l . kanewski & v.v . vardanian , _ astrophys . _ , * 167 * , 111 ( 1990 ) . t. erber , _ revs . _ , * 38 * , 626 ( 1966 ) . b. mcbreen & c.j . lambert , _ proc . cosmic ray conf . _ ( paris ) * 6 * , 70 ( 1981 ) . h.p . vankov and p.v . stavrev , _ phys . b _ , * 266 * , 178 ( 1991 ) . see t. stanev _ rev . _ * d25 * , 1291 ( 1982 ) for the influence of the lpm effect on the shower development .	ultra high energy gamma rays produce electron positron pairs in interactions on the geomagnetic field . the pair electrons suffer magnetic bremsstrahlung and the energy of the primary gamma ray is shared by a bunch of lower energy secondaries . these processes reflect the structure of the geomagnetic field and cause experimentally observable effects . the study of these effects with future giant air shower arrays can identify the nature of the highest energy cosmic rays as either @xmath0rays or nuclei . the nature of the highest energy cosmic rays @xmath1 bartol research institute , university of delaware , newark , de 19716 + @xmath2 institute for nuclear research and nuclear energy , sofia # # , bulgaria +
any network found in the literature is inevitably just a sampled representative of its real - world analogue under study . for instance , many networks change quickly over time and in most cases merely incomplete data is available on the underlying system . additionally , network sampling techniques are lately often applied to large networks to allow for their faster and more efficient analysis . since the findings of the analyses and simulations on such sampled networks are implied for the original ones , it is of key importance to understand the structural differences between the original networks and their sampled variants . a large number of studies on network sampling focused on the changes in network properties introduced by sampling . lee et al . @xcite showed that random node and link selection overestimate the scale - free exponent @xcite of the degree and betweenness centrality @xcite distributions , while they preserve the degree mixing @xcite . on the other hand , random node selection preserves the degree distribution of different random graphs @xcite and performs better for larger sampled networks @xcite . furthermore , leskovec et al . @xcite showed that the exploration sampling using random walks or forest - fire strategy @xcite outperforms the random selection techniques in preserving the clustering coefficient @xcite , different spectral properties @xcite , and the in - degree and out - degree distributions . more recently , ahmed et al . @xcite proposed random link selection with additional induction step , which notably improves on the current state - of - the - art . their results confirm that the proposed technique well captures the degree distributions , shortest paths @xcite and also the clustering coefficient of the original networks . lately , different studies also focus on finding and correcting biases in sampling process , for example observing the changes of user attributes under the sampling of social networks @xcite , analyzing the bias of traceroute sampling @xcite and understanding the changes of degree distribution and hubs inclusion under various sampling techniques @xcite . however , despite all those efforts , the changes in network structure introduced by sampling and the effects of network structure on the performance of sampling are still far from understood . real - world networks commonly reveal communities ( also link - density community @xcite ) , described as densely connected clusters of nodes that are loosely connected between @xcite . communities possibly play important roles in different real - world systems , for example in social networks communities represent friendship circles or people with similar interest @xcite , while in citation networks communities can help us to reveal relationships between scientific disciplines @xcite . furthermore , community structure has a strong impact on dynamic processes taking place on networks @xcite and thus provides an important insight into structural organization and functional behavior of real - world systems . consequently , a number of community detection algorithms have been proposed over the last years @xcite ( for a review see @xcite ) . most of these studies focus on classical communities characterized by higher density of edges @xcite . however , some recent works demonstrate that real - world networks reveal also other characteristic groups of nodes @xcite like groups of structurally equivalent nodes denoted modules @xcite ( also link - pattern community @xcite and other @xcite ) , or different mixtures of communities and modules @xcite . despite community structure appears to be an intrinsic property of many real - world networks , only a few studies considered the effects between the community structure and network sampling . salehi et al . @xcite proposed page - rank sampling , which improves the performance of sampling of networks with strong community structure . furthermore , expansion sampling @xcite directly constructs a sample representative of the community structure , while it can also be used to infer communities of the unsampled nodes . other studies , for example analyzed the evolution of community structure in collaboration networks and showed that the number of communities and their size increase over time @xcite , while the network sampling has a potential application in testing for signs of preferential attachment in the growth of networks @xcite . however , to the best of our knowledge , the question whether sampling destroys the structure of communities and other groups of nodes or are sampled nodes organized in a similar way than nodes in original network remains unanswered . in this paper , we study the presence of characteristic groups of nodes in different social and information networks and analyze the changes in network group structure introduced by sampling . we consider six sampling techniques including random node and link selection , network exploration and expansion sampling . the results first reveal that nodes in social networks form densely linked community - like groups , while the structure of information networks is better described by modules . however , regardless of the type of the network and consistently across different sampling techniques , the structure of sampled networks exhibits much stronger characterization by community - like groups than the original networks . we therefore conclude that the rich community structure is not necessary a result of for example homophily in social networks . the rest of the paper is structured as follows . in section [ sec : sampl ] , we introduce different sampling techniques considered in the study , while the adopted node group extraction framework is presented in section [ sec : nodegroups ] . the results of the empirical analysis are reported and formally discussed in section [ sec : analys ] , while section [ sec : conclusion ] summarizes the paper and gives some prominent directions for future research . network sampling techniques can be roughly divided into two categories : random selection and network exploration techniques . in the first category , nodes or links are included in the sample uniformly at random or proportional to some particular characteristic like the degree of a node or its pagerank score @xcite . in the second category , the sample is constructed by retrieving a neighborhood of a randomly selected seed node using random walks , breadth - first search or another strategy . for the purpose of this study , we consider three techniques from each of the categories . from the random selection category , we first adopt random node selection by degree @xcite ( rnd ) . here , the nodes are selected randomly with probability proportional to their degrees , while all their mutual links are included in the sample ( fig . [ subfig : rnd ] ) . note that rnd improves the performance of the basic random node selection @xcite , where the nodes are selected to the sample uniformly at random . rnd fits better spectral network properties @xcite and produces the sample with larger weakly connected component @xcite . moreover , it shows good performance in preserving the clustering coefficient and betweenness centrality distribution of the original networks @xcite . nevertheless , it can still construct a disconnected sample network , despite a fully connected original network . next , we adopt random link selection @xcite ( rls ) , where the sample consists of links selected uniformly at random ( fig . [ subfig : rls ] ) . rls overestimates degree and betweenness centrality exponent , underestimate the clustering coefficient and accurately matches the assortativity of the original network @xcite . the samples created with rls are sparse and the connectivity of the original network is not preserved , still rls is likely to capture the path length of the original network @xcite . last , we adopt random link selection with induction @xcite ( rli ) , which improves the performance of rls . in rli , the sample consists of randomly selected links as before , while also all additional links between their endpoints ( fig . [ subfig : rli ] ) . rli outperforms several other methods in capturing the degree , path length and clustering coefficient distribution . it selects nodes with higher degree than rls , thus the connectivity of the sample is increased @xcite . techniques from random selection category imitate classical statistical sampling approaches , where each individual is selected from population independently from others until desired size of the sample is reached . from the network exploration category , we first adopt breadth - first sampling @xcite ( bfs ) . here , a seed node is selected uniformly at random , while its broad neighborhood retrieved from the basic breadth - first search is included in the sample ( fig . [ subfig : bfs ] ) . the sample network is thus a connected subgraph of the original network . bfs is biased towards selecting high - degree nodes in the sample @xcite . it captures well the degree distribution of the networks , while it performs worst in inclusion of hubs in the sample quickly in the sampling process @xcite . bfs imitates the snowball sampling approach for collecting social data used especially when the data is difficult to reach @xcite . selected seed participant is asked to report his friends , which are than invited to report their friends . the procedure is repeated until the desired number of people is sampled . next , we adopt a modification of bfs denoted forest - fire sampling @xcite ( ffs ) . in ffs , the broad neighborhood of a randomly selected seed node is retrieved from partial breadth - first search , where only some neighbors are included in the sample on each step ( fig . [ subfig : ffs ] ) . the number of neighbors is sampled from a geometric distribution with mean @xmath0 , where @xmath1 is set to @xmath2 @xcite . ffs matches well spectral properties @xcite , while it underestimates the degree distribution and fails to match the path length and clustering coefficient of the original networks @xcite . however , ffs corresponds to a model by which one author collects the papers to cite and include them in the bibliography @xcite . the author starts with one paper , explores its bibliography and selects the papers to cite . the procedure is recursively repeated in selected papers until desired collection of citations is reached . last , we adopt expansion sampling @xcite ( exs ) , where the seed node is again selected uniformly at random , while the neighbors of the sampled nodes are included in the sample with probability proportional to @xmath3 where @xmath4 is the concerned node , @xmath5 the current sample and @xmath6 the neighborhood of nodes in @xmath5 ( fig . [ subfig : exs ] ) . expression @xmath7 denotes the expansion factor of node @xmath4 for sample @xmath5 and means the number of new neighbors contributed by @xmath4 . the parameter @xmath8 is set to @xmath9 @xcite . note that exs ensures that the sample consists of nodes from most communities in the original network and that the nodes that are grouped together in the original network , are also grouped together in the sample @xcite . exs imitates the modification of snowball sampling approach mentioned above , where for example we want to gather the data about individuals from different countries . thus , on each step we include in the sample the individuals , which knows larger number of others from various countries . the node group structure of different networks is explored by a group extraction framework @xcite with a brief overview below . let the network be represented by an undirected graph @xmath10 , where @xmath11 is the set of nodes and @xmath12 the set of links . next , let @xmath5 be a group of nodes and @xmath13 a subset of nodes representing its corresponding linking pattern ( i.e. , the pattern of connections of nodes from @xmath5 to other nodes @xcite ) , @xmath14 . denote @xmath15 and @xmath16 . the linking pattern @xmath13 is selected to maximize the number of links between @xmath5 and @xmath13 , and minimize the number of links between @xmath5 and @xmath17 , while disregarding the links with both endpoints in @xmath18 . for details on the group objective function see @xcite . the above formalism comprises different types of groups commonly analyzed in the literature ( fig . [ fig : groups ] ) . it consider communities @xcite ( i.e. , link - density community @xcite ) , defined as a ( connected ) group of nodes with more links toward the nodes in the group than to the rest of the network @xcite . communities are characterized by @xmath19 . furthermore , the formalism consider possibly disconnected groups of structurally equivalent nodes denoted modules @xcite ( i.e. , link - pattern community @xcite ) , defined as a ( possibly ) disconnected group of nodes with more links towards common neighbors than to the rest of the network @xcite . modules have @xmath20 . communities and modules represent two extreme cases with all other groups being the mixtures of the two @xcite , @xmath21 and/or @xmath22 . the reader may also find it interesting that the core - periphery structure is a mixture with @xmath23 , while the hub & spokes structure is a module with @xmath24 . the type of group @xmath5 can in fact be determined by the jaccard index @xcite of @xmath5 and its corresponding linking pattern @xmath13 . the group parameter @xmath25 @xcite , @xmath26 $ ] , is defined as @xmath27 communities have @xmath28 , while modules are indicated by @xmath29 . mixtures correspond to groups with @xmath30 . for the rest of the paper , we refer to groups with @xmath31 as community - like and groups with @xmath32 as module - like . groups in networks are revealed by a sequential extraction procedure proposed in @xcite . one first finds the group @xmath5 and its linking pattern @xmath13 with random - restart hill climbing @xcite that maximizes the objective function . next , the revealed group @xmath5 is extracted from the network by removing the links between groups @xmath5 and @xmath13 , and any node that becomes isolated . the procedure is then repeated on the remaining network until the objective function is larger than the @xmath33th percentile of the values obtained under the same framework in a corresponding erds - rnyi random graph @xcite . all groups reported in the paper are thus statistically significant at @xmath34 level . note that the above procedure allows for overlapping @xcite , hierarchical @xcite , nested and other classes of groups . section [ subsec : nets ] introduces real - world networks considered in the study . section [ subsec : orig ] reports the node group structure of the original networks extracted with the framework described in section [ sec : nodegroups ] . the groups extracted from the sampled networks are analyzed in section [ subsec : sampled ] . for a complete analysis , we also observe the node group structure of a large network with more than a million links in section [ subsec : large ] . clrr & & & + _ collab _ & high energy physics collaborations @xcite & @xmath35 & @xmath36 + _ pgp _ & pretty good privacy web - of - trust @xcite & @xmath37 & @xmath38 + _ p2p _ & gnutella peer - to - peer file sharing @xcite & @xmath39 & @xmath40 + _ citation _ & high energy physics citations @xcite & @xmath41 & @xmath42 + the empirical analysis in the following sections was performed on four real - world social and information networks . their main characteristics are shown in table [ tbl : nets ] . the _ collab _ @xcite is a social network of scientific collaborations among researchers , who submitted their papers to high energy physics theory category on the arxiv in the period from january @xmath43 to april @xmath44 . the nodes represent the authors , while undirected links denote that two authors co - authored at least one paper together . the _ pgp _ @xcite is a social network , which corresponds to the interaction network of users of the pretty good privacy algorithm collected in july @xmath45 . the nodes represent users , while undirected links indicate relationships between those , who sign each other s public key . the _ p2p _ @xcite is an information network , which contains a sequence of snapshots of the gnutella peer - to - peer file sharing network collected in august @xmath46 . the nodes represent hosts in the gnutella network , which are linked by undirected links if there exist connections between them . the _ citation _ @xcite is an information network , again gathered from the high energy physics theory category from the arxiv in the period from january @xmath43 to april @xmath44 and includes the citations among all papers in the dataset . the network consists of nodes , which represent papers , while links denote that one paper cite another . we first analyze the properties of groups extracted from the original networks summarized in table [ tbl : orig ] . the number of groups differs among networks , still the mean group size @xmath47 ( denoted @xmath48 ) is comparable across network types . groups @xmath5 in social networks consist of around @xmath49 nodes , while @xmath48 in information networks exceeds @xmath50 nodes . the mean linking pattern size @xmath51 ( denoted @xmath52 ) of social networks is comparable to @xmath48 . the latter relation @xmath53 is expected due to the pronounced community structure commonly found in social networks @xcite . on the other hand , @xmath54 is expected for information networks , due to the abundance of module - like groups . the characteristic group structure of networks is reflected in the group parameter @xmath25 . for social networks , its values are around @xmath55 , which indicates the presence of communities , modules and mixtures of these . in contrast to social networks , the information networks have @xmath25 closer to @xmath56 and consist mostly of module - like groups . to summarize , social networks represent people and interactions between them , like a few authors writing a paper together , therefore we can expect a larger number of community - like groups in these networks . on the other hand , in information networks the homophily is less typical and thus the structure of these networks seem better described by module - like groups . crrrrrrrr & & & & + & & & & & & + _ collab _ & @xmath57 & @xmath58 & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 + _ pgp _ & @xmath65 & @xmath66 & @xmath67 & @xmath60 & @xmath68 & @xmath69 & @xmath70 & @xmath71 + _ p2p _ & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & @xmath79 + _ citation _ & @xmath80 & @xmath81 & @xmath82 & @xmath83 & @xmath84 & @xmath77 & @xmath85 & @xmath86 + sampling techniques outlined in section [ sec : sampl ] enable setting the size of the sampled networks in advance . we consider sample sizes of @xmath87 of nodes from the original networks , that provides for an accurate fit of several network properties @xcite . table [ tbl : samplsoc ] and [ tbl : samplinf ] present the properties of the node group structure of sampled social and information networks , respectively . notice that rls and ffs show different performance than other techniques . the samples obtained with rls and ffs contain less groups , which consist of no more than @xmath88 nodes . additionally , almost all groups in these samples are modules , which reflects in the mean group parameter @xmath25 ( denoted @xmath89 ) approaching @xmath56 for all networks . to verify the above findings , we compute externally studentized residuals of the sampled networks that measure the consistency of each sampling technique with the rest . the residuals are calculated for each technique as the difference between the observed value of considered property and its mean divided by the standard deviation . the mean value and standard deviation are computed for all sampling techniques , excluding the observed one ( for details see @xcite ) . statistically significant inconsistencies between techniques are revealed by two - tailed student @xmath90test @xcite at @xmath91value of @xmath92 , rejecting the null hypothesis that the values of the considered property are consistent across the sampling techniques . statistical comparison of sampling techniques for the number of groups and the mean group parameter @xmath25 is shown on fig . [ fig : stau ] . we confirm that the samples obtained with rls and ffs reveal significantly less groups with significantly smaller @xmath89 than other sampling techniques . moreover , if we compare the number of links in the sampled networks , rls and ffs create samples that contain on average @xmath93 of links from the original networks . in contrast , the samples obtained with rnd , rli , bfs and exs consist of around @xmath94 of links from the original networks . as mentioned before , the sizes of all samples are @xmath87 of the original networks , thus the sampled networks obtained with rls and ffs are much sparser than others . in addition , the performance of rls and ffs can also be explained by their definition . since in rls we include only randomly selected links in the sample , the variance is very high , while it commonly contains a large number of sparsely linked components , whose structure is best described as module - like . on the other hand , the samples obtained with ffs consist of one connected component with a low average degree of @xmath95 . thus , the sparsely connected nodes also form groups , which are more similar to modules . due to the above reasons , we exclude rls and ffs from further analysis . we focus on rnd , rli , bfs , and exs , whose performance is clearly more comparable . the selected sampling techniques perform similarly across all networks as shown in table [ tbl : samplsoc ] for social and table [ tbl : samplinf ] for information networks . the samples consist of various number of groups , still in most cases less than the original networks . the mean sizes @xmath47 and @xmath51 are around @xmath96 , in contrast to groups with @xmath97 nodes on average in the original networks . still , @xmath98 irrespective of network type and the sampling technique , which implies stronger characterization by community - like groups , as already argued in the case of social networks in section [ subsec : orig ] . ccrrrrrrrrrr & & & & & + & & & & & & & + & / & & @xmath58 & @xmath59 & & @xmath61 & @xmath62 & @xmath85 & @xmath62 + & rnd & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 + & rls & @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath77 & @xmath112 & @xmath113 + & rli & @xmath114 & @xmath101 & @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 + & bfs & @xmath121 & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 & @xmath86 + & ffs & @xmath128 & @xmath129 & @xmath130 & @xmath131 & @xmath76 & @xmath77 & @xmath77 & @xmath132 + & exs & @xmath133 & @xmath134 & @xmath135 & @xmath136 & @xmath137 & @xmath138 & @xmath139 & @xmath140 + & / & & @xmath66 & @xmath67 & & @xmath68 & @xmath69 & @xmath141 & @xmath142 + & rnd & @xmath143 & @xmath144 & @xmath145 & @xmath146 & @xmath147 & @xmath148 & @xmath149 & @xmath150 + & rls & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 & @xmath138 & @xmath156 & @xmath157 + & rli & @xmath158 & @xmath144 & @xmath159 & @xmath160 & @xmath147 & @xmath161 & @xmath162 & @xmath163 + & bfs & @xmath164 & @xmath165 & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171 + & ffs & @xmath172 & @xmath100 & @xmath173 & @xmath131 & @xmath76 & @xmath77 & @xmath77 & @xmath132 + & exs & @xmath174 & @xmath175 & @xmath144 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 + ccrrrrrrrrr & & & & & + & & & & & & & + & / & & @xmath73 & @xmath74 & & @xmath76 & @xmath77 & @xmath78 & @xmath79 + & rnd & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath157 + & rls & @xmath188 & @xmath189 & @xmath190 & @xmath191 & @xmath76 & @xmath77 & @xmath77 & @xmath132 + & rli & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath163 & @xmath197 & @xmath198 + & bfs & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath185 & @xmath140 & @xmath203 & @xmath204 + & ffs & @xmath205 & @xmath206 & @xmath207 & @xmath131 & @xmath76 & @xmath77 & @xmath77 & @xmath132 + & exs & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath196 & @xmath212 & @xmath213 & @xmath214 + & / & & @xmath81 & @xmath82 & & @xmath84 & @xmath77 & @xmath85 & @xmath86 + & rnd & @xmath215 & @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 + & rls & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath77 & @xmath77 & @xmath132 + & rli & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath220 & @xmath233 & @xmath234 + & bfs & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & @xmath240 & @xmath241 & @xmath242 + & ffs & @xmath243 & @xmath244 & @xmath245 & @xmath131 & @xmath76 & @xmath77 & @xmath77 & @xmath132 + & exs & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 & @xmath220 & @xmath251 & @xmath252 + indeed , the majority of groups found in sampled social networks are community - like , which reflects in the parameter @xmath253 . in sampled information networks the number of mixtures decreases and communities appear , thus @xmath25 is larger than in the original networks . [ fig : tau - hists ] shows a clear difference in the distribution of @xmath25 between the original and sampled networks . furthermore , to confirm that differences exist between the structure of the original and sampled networks , we compute externally studentized residuals , where we include the value of considered property of the original network in computing the mean over different sampling techniques . we compare the number of groups and the parameter @xmath89 for the original networks and their samples ( fig . [ fig : stauo ] ) . the results prove that the original networks contain a significantly larger number of groups with significantly smaller @xmath89 than the sampled networks . yet , larger parameter @xmath25 and consequently more community - like groups in sampled social networks and less module - like groups in sampled information networks indicate clear changes in the network structure introduced by sampling . we conclude that these changes occur regardless of the network type or the adopted sampling technique . + + notice that the largest @xmath25 and thus the strongest characterization by community - like groups is revealed in the sampled networks obtained with both random selection techniques , rnd and rli . in rnd nodes with higher degrees are more likely to be selected to the sample by definition , while rli is biased in a similar way @xcite . thus , densely connected groups of nodes have a higher chance of being included in the sampled network , while sparse parts of the networks remain unsampled . on the other hand , bfs and exs sample the broad neighborhood of a randomly selected seed node and thus the sampled network represents a connected component . in the case of bfs , all nodes and links of some particular part of the original network are sampled . the latter is believed to be representative of the entire network @xcite , yet bfs is biased towards sampling nodes with higher degree @xcite and overestimates the clustering coefficient , especially in information networks @xcite . on the other hand , exs ensures the smallest partition distance among several other sampling techniques , which means that nodes grouped together in communities of sampled network are also in the same community in the original network @xcite . therefore , the stronger characterization by community - like groups in sampled networks can also be explained by the definition and behavior of the sampling techniques . + due to the relatively high time complexity of the node group extraction framework , we consider only networks with a few thousand nodes . however , our previous study @xcite proved that the size of the original network does not affect the accuracy of the sampling . still , for a complete analysis , we also inspect the changes in node group structure introduced by sampling of a large _ notredame _ network with more than a million links . due to the simplicity and execution time , we present the analysis for two sampling techniques , rnd from random selection and bfs from network exploration category . we also limit the number of groups extracted from the networks to @xmath254 ( i.e. , we consider top @xmath254 most significant groups with respect to the objective function ) . the _ notredame _ data are collected from the web pages of the university of notre dame _ nd.edu _ domain in @xmath255 . the network contains @xmath256,@xmath257 nodes representing individual web pages , while @xmath258,@xmath259,@xmath260 links denote hyperlinks among them . table [ tbl : wnd ] shows the properties of groups , found in the original and sampled networks . the samples consist of smaller groups , still the mean size @xmath47 remains larger than the mean size @xmath51 . the majority of groups extracted from the original network are module - like , which reflects in the parameter @xmath25 slightly larger than @xmath56 . on the other hand , the changes introduced by sampling are clear , since the samples contain less modules , which is revealed by a larger parameter @xmath25 . these findings are consistent with the results on smaller networks from previous sections . the _ notredame _ as an information network expectedly consists of densely linked groups similar to modules , while the structure of sampled networks exhibits stronger characterization by community - like groups . that is again irrespective of the adopted sampling technique . crrrrrrrr & & & & + & & & & & & + / & @xmath254 & @xmath261 & @xmath262 & & @xmath263 & @xmath77 & @xmath264 & @xmath265 + rnd & @xmath254 & @xmath266 & @xmath81 & @xmath267 & @xmath268 & @xmath77 & @xmath269 & @xmath270 + bfs & @xmath254 & @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath77 & @xmath275 & @xmath276 + in this paper , we study the presence of characteristic groups of nodes like communities and modules in different social and information networks . we observe the groups of the original networks and analyze the changes in the group structure introduced by the network sampling . the results first reveal noticeable differences in the group structure of original social and information networks . nodes in social networks form smaller community - like groups , while information networks are better characterized by larger modules . after applying network sampling techniques , sampled networks expectedly contain fewer and smaller groups . however , the sampled networks exhibit stronger characterization by community - like groups than the original networks . we have shown that the changes in the node group structure introduced by sampling occur regardless of the network type and consistently across different sampling techniques . since networks commonly considered in the literature are inevitably just a sampled representative of its real - world analogue , some results , such as rich community structure found in these networks , may be influenced by or are merely an artifact of sampling . our future work will mainly focus on larger real - world networks , including other types of networks like biological and technological . moreover , we will further analyze the changes in the node group structure introduced by sampling and explore techniques that could overcome observed deficiencies . this work has been supported in part by the slovenian research agency _ arrs _ within the research program no . p2 - 0359 , by the slovenian ministry of education , science and sport grant no . 430 - 168/2013/91 , and by the european union , european social fund . j. leskovec , j. kleinberg , c. faloutsos , graphs over time : densification laws , shrinking diameters and possible explanations , in : proceedings of the 11th acm sigkdd international conference on knowledge discovery and data mining , acm , 2005 , pp . 177187 . h. park , s. moon , sampling bias in user attribute estimation of osns , in : proceedings of the 22nd international conference on world wide web companion , international world wide web conferences steering committee , 2013 , pp . 183184 . a. lakhina , j. w. byers , m. crovella , p. xie , sampling biases in ip topology measurements , in : proceedings of the 22nd annual joint conference of the ieee computer and communications , vol . 1 , ieee , 2003 , pp . 332341 . a. s. maiya , t. y. berger - wolf , benefits of bias : towards better characterization of network sampling , in : proceedings of the 17th acm sigkdd international conference on knowledge discovery and data mining , acm , 2011 , pp . 105113 . l. ubelj , n. blagus , m. bajec , group extraction for real - world networks : the case of communities , modules , and hubs and spokes , in : proceedings of the international conference on network science , 2013 , pp .	any network studied in the literature is inevitably just a sampled representative of its real - world analogue . additionally , network sampling is lately often applied to large networks to allow for their faster and more efficient analysis . nevertheless , the changes in network structure introduced by sampling are still far from understood . in this paper , we study the presence of characteristic groups of nodes in sampled social and information networks . we consider different network sampling techniques including random node and link selection , network exploration and expansion . we first observe that the structure of social networks reveals densely linked groups like communities , while the structure of information networks is better described by modules of structurally equivalent nodes . however , despite these notable differences , the structure of sampled networks exhibits stronger characterization by community - like groups than the original networks , irrespective of their type and consistently across various sampling techniques . hence , rich community structure commonly observed in social and information networks is to some extent merely an artifact of sampling . complex networks , network sampling , node group structure , communities , modules + _ pacs : _ 64.60.aq , 89.75.fb , 89.90.+n
cavity quantum electrodynamics ( cavity qed ) enables the study of the interaction between light and matter at the most elementary level , thanks to the achievement of a strong coupling between a single atom and a single photon trapped in a microwave or optical cavity@xcite . this paradigm has been recently brought into superconducting circuits : artificial atoms consisting of two level superconducting circuits have been coupled to superconducting cavities@xcite , in the context of circuit qed . these experiments provide an ideal playground to test the basic laws of quantum mechanics because they can be described in terms of simple models like the jaynes - cummings hamiltonian . however , such models conceal essential physical differences between cavity and circuit qed . on the one hand , the coupling between isolated atoms and cavity photons mainly occurs due to the sensitivity of the atom electric dipole to the cavity electric field . this coupling depends on microscopic details since the atomic dipole is set by the structure of the atom electronic orbitals . furthermore , one can generally perform the `` electric - dipole approximation '' which assumes that the cavity field varies little on the scale of the atomic system@xcite . on the other hand , the behavior of submicronic superconducting circuits is essentially insensitive to microscopic details due to the rigidity of the superconducting phase@xcite . for instance , the behavior of a superconducting charge qubit can be described with one macroscopic variable , i.e. the total charge of a superconducting island@xcite . this charge can vary due to the josephson coupling between the island and an external superconducting reservoir . the coupling between the superconducting charge qubit and the cavity is usually described in terms of a capacitive coupling between the superconducting island and the cavity central conductor . as a result , the chemical potential of the superconducting island is shifted proportionally to the cavity electric potential@xcite . this picture implies strong inhomogeneities of the photonic electric field on the scale of the superconducting qubit , in contrast to what is generally considered in atomic cavity qed for a single atom . recent technological progress is enabling the development of a new type of experiments where nanocircuits based on carbon nanotubes , semiconducting nanowires , two - dimensional electron gases or graphene , are coupled to coplanar microwave cavities@xcite . this paves the way for the development of `` mesoscopic qed '' , a denomination introduced in a pioneering theory work@xcite . mesoscopic qed opens many possibilities because nanoconductors can be tunnel - coupled to various types of fermionic reservoirs such as normal metals , ferromagnets@xcite or superconductors@xcite , in a large variety of geometries . so far , theoretical studies on mesoscopic qed have mainly focused on quantum dot circuits@xcite . several configurations have been suggested to reach the strong coupling regime between an electronic spin and cavity photons@xcite , or more generally , to develop quantum computing schemes@xcite . mesoscopic qed also tackles problems which go beyond the mechanics of closed two level systems coupled to cavities , usually considered in cavity or circuit qed . the interaction between electronic transport and the light - matter interaction leads to a rich phenomenology@xcite . besides , coplanar cavities could be used as a powerful probe to reveal some exotic properties of hybrid nanocircuits , like for instance the existence of topological superconducting phases@xcite , majorana quasiparticle modes@xcite , or spin - entanglement in a cooper pair beam splitter@xcite . on the experimental side , pioneering works have focused on mesoscopic rings@xcite and metallic tunnel junctions@xcite . more recently , experiments have been performed with single quantum dots@xcite and double quantum dots ( dqds ) @xcite with normal metal reservoirs . reaching the strong coupling regime between the charge states of a dqd and a cavity is still a challenge@xcite . nevertheless , interesting resonance phenomena have already been observed@xcite . several experiments have also provided evidence for a modification of the cavity behavior by finite bias transport through a dqd@xcite , including a maser effect@xcite . these recent developments call for a full description of the coupling between a hybrid nanocircuit and cavity photons . one question naturally arises : is mesoscopic qed closer to atomic cavity qed or superconducting circuit qed ? what are the specificities of the coupling between a nanocircuit and a cavity ? so far , most theory works have considered a capacitive coupling between the nanocircuit and the cavity central conductor , by analogy with circuit qed@xcite . this approach implies a coarse grained electric description of the nanocircuit , and a concentration of the non - homogeneous photonic electric field inside some capacitive elements . a few works have considered a direct coupling between the motion of electrons trapped in the nanoconductors and the bare cavity electric field , which is assumed to be constant on the scale of the nanocircuit . this is reminiscent of the descriptions used in cavity qed@xcite . in this article , we introduce a description of mesoscopic qed which bridges between these two approaches . we use a model which focuses on conduction electrons tunneling between the different elements of a nanocircuit . the tunneling electrons occupy quasi - localized orbitals in each nanocircuit element , which recalls the atomic orbital degree of freedom of cavity qed . however , there also exists collective plasmonic modes in the nanocircuit , which can screen at least partially the cavity fields . we use a gauge - invariant mesoscopic qed hamiltonian which accounts for the non - uniform screening of the cavity fields inside the nanocircuit , and for the electromagnetic boundary conditions provided by cavity conductors and voltage - biased nanocircuit dc gates . in the limit where photon - induced magnetic effects are negligible , we can reexpress the mesoscopic qed hamiltonian in terms of a scalar photonic pseudo - potential . this picture unifies the different approaches used so far to describe mesoscopic qed devices , since the photonic pseudo - potential can vary linearly with space in the case of a locally uniform photonic electric field ( dipolar coupling limit ) , as well as remain constant inside a given circuit element in the limit of a coarse grained circuit model . in the framework of a tunneling model , the photonic pseudo - potential leads to various types of linear electron / photon coupling terms : cavity photons shift the orbital energies of the different nanocircuit elements , but also induce simultaneously tunneling and local orbital transitions . this general description can be used to study the behavior of many different types of mesoscopic qed devices . for instance , it can be applied to quantum dot circuits , molecular circuits , quantum point contacts , metallic tunnel junctions , and superconducting nanostructures enclosing andreev bound states or majorana bound states . to illustrate the richness of our approach , we study in details the elementary example of a cavity coupled to a `` quantum rc circuit '' , i.e. a single quantum dot coupled to a single normal metal reservoir . the photon - induced tunneling terms between the quantum dot and the reservoir induce a non - universal relation between the cavity frequency pull and the cavity damping pull , contrarily to what is expected with purely capacitive coupling schemes at low temperatures@xcite . this paper is organized as follows . in section ii a , we discuss the gauge - invariant mesoscopic qed hamiltonian , which involves a photonic vector potential . in section ii b , we perform a unitary transformation to obtain a new hamiltonian where the electron / photon coupling is due to the scalar photonic pseudo - potential . in section ii c , we reexpress the photonic pseudo - potential scheme in the framework of a tunnelling model . in section ii d , we discuss the application of our formalism to the case of nanocircuits with superconducting elements . in particular , we give an explicit hamiltonian for a nanostructure with majorana bound states coupled to a cavity . in section iii , we work out in details the case of the quantum rc circuit coupled to a cavity . section iv concludes . appendix a gives a detailed mathematical justification for the form of the hamiltonian of section ii a , on the basis of an effective model which separates physically the tunneling electrons occupying individual orbital states in the different elements of a nanocircuit , from the plasmonic collective modes of this nanocircuit . appendix b discusses the advantages of the photonic pseudo - potential scheme . a nanocircuit encloses a large variety of degrees of freedom . our main interest in this paper is the interaction of the cavity with the ensemble @xmath0 of the `` tunneling '' conduction electrons , which can occupy orbital states in the different elements of a nanocircuit and tunnel between them . however , there also exists plasmonic electronic modes which are responsive for the screening of external fields from a massive metallic electrode . plasmons can also convey screening currents in the reservoir electrical lines , which reequilibrate locally charge distributions on tunnel junctions after a tunneling event . the question of plasmonic modes in a nanocircuit is complicated since nanoconductors have generally a low electronic density , which allows only a partial screening of the cavity field@xcite . in principle , even the metallic contacts of the nanocircuit are not able to screen totally the cavity field due to their thinness . in the limit where the nanocircuit contacts are connected to outside electrical lines with a low impedance , it is reasonable to assume that the dynamics of plasmonic modes in the nanocircuit is very fast compared to the dynamics of the cavity modes and of the tunneling electrons@xcite . in this case , one can assume that the plasmonic modes in the nanocircuit generate a distribution of screening charges which is enslaved to the position of the tunneling charges and to the value of the cavity field . this produces a renormalization of the cavity field properties , and in particular its spatial dependence nearby and inside the nanocircuit . it also produces a renormalization of the coulomb interactions between the tunneling charges . in general , the coulomb interactions between the nanocircuit charges is also renormalized by the screening charges on the cavity conductors ( see appendix a ) . the positive ions and valence electrons constituting the crystalline structure of the nanocircuit can be treated in a mean field approach , leading to an effective potential @xmath1 which confines tunneling electrons inside the nanocircuit . this confinement potential can delimit quantum dots , or induce disorder effects in nanoconductors , for instance . this approach is sufficient provided the background charges producing @xmath1 and the microwave cavity are far off resonant . in this framework , gauge invariance imposes the hamiltonian @xmath2 with@xmath3@xmath4@xmath5 and @xmath6 the elementary charge . the field operator @xmath7 includes all the tunneling charges of the nanocircuit . above , @xmath8 is a single electron hamiltonian , @xmath9 describes the coulomb interaction between the tunneling electrons ( see appendix a for details ) , and @xmath10 is the hamiltonian of a ( renormalized ) cavity electromagnetic mode . the potential @xmath11 is due to dc voltages applied on the nanocircuit electrostatic gates and to constant charges on floating conductors like the cavity central conductor in a coplanar geometry . the coupling between the cavity and the tunneling electrons arises from the photonic vector potential term of eq.([6 ] ) . for simplicity , we have considered a single cavity mode with creation operator @xmath12 and vector potential profile @xmath13 . this description can be generalized straightforwardly to the multimode case by introducing a sum on a cavity mode index in eqs . ( [ 1 ] ) and ( [ 6 ] ) . note that @xmath13 can strongly vary on the scale of the nanocircuit . the formal description of the electromagnetic field and the light / matter interaction in mesoscopic qed requires some care . in atomic cavity qed , the cavity field is generally assumed to vary slowly on the scale of an atom@xcite . in contrast , we have to take into account strong spatial variations of the photonic field on the scale of a nanocircuit . in particular , a protrusion of a cavity conductor can be used to increase the photonic field locally , close to a given quantum dot @xcite ( see fig . the dc voltage - biased electrostatic gates , used to control the nanocircuit spectrum , provide supplementary boundary conditions on the electromagnetic field . besides , we have already mentioned that the plasmonic screening charges on the nanocircuit conductors can modify the photonic vector potential profile , especially inside and around fermionic reservoirs . in order to describe this complex reality and justify mathematically the shape of eqs.([1]-[6 ] ) , we introduce in appendix a an effective model which separates physically the tunneling electrons occupying individual orbital states in the nanocircuit from the collective plasmonic modes of the nanocircuit . in this framework , one can introduce a hodge decomposition of the electromagnetic field@xcite , which can be obtained thanks to the use of an auxiliary electrostatic green s function . this leads to the terms in @xmath14 , @xmath9 and @xmath11 in eqs . ( [ 1]-[6 ] ) . in this section , we disregard the spin of charge carriers because we want to focus on regimes where charge effects prevail . the introduction of the spin degree of freedom in hamiltonian ( [ 1 ] ) would be straightforward by invoking gauge invariance ( see for instance section [ ad ] ) . the direct coupling between the particles spin and the photonic magnetic field can usually be disregarded because this coupling is expected to be very weak unless collective excitations are considered in a large ensemble of spins@xcite . from recent predictions , real@xcite or effective@xcite spin - orbit coupling provide an alternative way to obtain a spin - photon coupling . in eq.([202 ] ) , the coupling of tunneling electrons to cavity photons occurs through two terms , one in @xmath15 , corresponding to single photon transitions , and the other one in @xmath16 , corresponding to two photon transitions . the second has no reason to be disregarded in the general case , even if it has no structure in the electronic sector . for instance , this term is crucial for determining the existence of a dicke quantum phase transition when many two - level systems are coupled transversely to a cavity@xcite . the @xmath16 term can also bring a non - negligible contribution to the cavity frequency pull caused by one nanocircuit , which is a central quantity in mesoscopic qed experiments ( see appendix b ) . hence , it is convenient to introduce a new representation in which the @xmath16 term is eliminated . this task is completed in the next section . in atomic cavity qed , the power - zienau - wooley ( pzw ) transformation formally enables the elimination of a space - dependent @xmath17 term from the system hamiltonian@xcite . this generalizes the widely known `` electric - dipole '' transformation used in the case where @xmath18 can be considered as constant on the scale of an atom . however , the natural variables of the pzw hamiltonian are polarization and magnetization densities associated to the charge distribution of an atomic system , which are not directly useful in our case . to reexpress the pzw hamiltonian in terms of particles coordinates and momenta , it is necessary to perform a multipolar development of the light / matter interaction@xcite . in practice , this development is performed to a limited order , which means that only a moderate space dependence of the photonic electric and magnetic fields is taken into account . in this section , we perform a unitary transformation of hamiltonian ( [ 1 ] ) in the same spirit as the pzw transformation , but with specificities required for mesoscopic qed . upon disregarding magnetic effects , we can take into account a strong space dependence of the photonic electric field , thanks to the use of a scalar photonic pseudo - potential . this procedure avoids the development of the light / matter interaction in terms of multipoles . this is necessary to describe e.g. multi - quantum dot circuits with strongly asymmetric capacitive couplings to the cavity . furthermore , we use a second - quantized description of charges , which is more convenient to describe central phenomena of mesoscopic physics like tunneling but also superconductivity ( see next section ) . to account for the possibly strong spatial variations of @xmath13 on the scale of a nanocircuit , we define inside the volume of the nanocircuit a reference point @xmath19 and a continuous functional path @xmath20 relating any points @xmath21 and @xmath22 . we also define a photonic pseudo - potential @xmath23 with@xmath24 importantly , one can use @xmath25 inside the nanocircuit volume , provided the flux of @xmath26 through the nanocircuit can be disregarded . this criterion is generally satisfied , considering the typical amplitude of the cavity magnetic field and the size of nanoconductors . in this case , the arbitrariness on the choice of the contours @xmath20 leads only to marginal effects . one exception is nanocircuits with large loops leading to magnetic aharonov - bohm effects . this case is beyond the scope of the rest of this article , which focuses on photon - induced electric effects . we perform the transformation @xmath27 with @xmath28 this gives @xmath29 with@xmath30 and @xmath31 . in this new representation , the single electron potential and the coulomb interactions are renormalized by the term in @xmath32 , but the electron / photon coupling takes a simpler linear form . the motion of tunneling electrons is modified by the pseudo - potential @xmath33 . note that the arbitrariness on the choice of @xmath19 is not an issue , since a change in @xmath19 shifts @xmath34 by a global constant and leads to a unitary transformation of @xmath35 . in the limit where the spatial dependence of the photonic vector potential can be disregarded i.e. @xmath36 , one finds @xmath37 with @xmath38 . this corresponds to the electric - dipole approximation frequently used in atomic cavity qed . in the opposite limit of a small enough conductor or a perfect screening of @xmath39 inside this conductor , one can use a @xmath40 constant inside this conductor . this is reminiscent of the capacitive network models frequently used to describe superconducting charge qubits in circuit qed . note that we do not assume that the ensemble @xmath0 of the tunneling charges is neutral . in this section , we consider a nanocircuit gathering quantum dots and reservoirs connected by tunnel junctions . the effect of time - dependent classical fields on tunneling through quantum dots has been studied since the 1990s@xcite . mesoscopic circuit qed offers the opportunity to study the interaction between microwave photons and quantum dot circuits from a different perspective . quantum dot circuits are often described in terms of a tunneling model in which the dots and reservoirs contain quasi - localized orbitals coupled by tunnel elements@xcite . this decomposition is instrumental in order to particularize the treatment of the reservoirs and account for the irreversibility of transport processes or other reservoir - induced damping effects . importantly , the tunnel couplings arise from an overlap between neighboring orbitals . this means that these orbitals can not be considered as orthogonal , except in the limit of weak tunnel couplings ( see for instance @xcite ) . however , in practice , it may be necessary to take into account a strong inter - dot tunneling . for example , in the case of a dqd , there can exist a strong splitting between bonding and antibonding states , which can be resonant with the cavity@xcite . depending on the absolute energy scale of the dqd confinement potential , it might be necessary to go to the limit of a non - perturbative inter - dot tunnel coupling to obtain such a splitting . in order to circumvent the difficulty related to the non - orthogonality of the nanocircuit orbitals , we divide the nanocircuit into the ensemble @xmath41 of tunnel - coupled quantum dots from one side , and the different reservoirs with an index @xmath42 with @xmath43 , from the other side . we use an index @xmath44 ... @xmath45 to denote these different elements , and an index @xmath46 to denote the different orbitals in a given element @xmath47 . we assume that the low energy spectrum for the whole ensemble @xmath41 of the quantum dots can be determined in the absence of the reservoirs as the discrete spectrum of a potential profile with one or several potential wells . this procedure can be performed for an arbitrary inter - dot tunneling strength and leads to exactly orthogonal orbitals , so that @xmath48 . one can also define exactly orthogonal orbitals in the isolated reservoir @xmath42 , with @xmath43 , so that @xmath49 . in a second step , one can evaluate the tunnel coupling @xmath50 between the orbitals @xmath46 and @xmath51 of @xmath41 and @xmath42 from the overlap between their wavefunctions . this gives @xmath52 only at lowest order in tunneling@xcite . however , using a weak tunneling between @xmath41 and @xmath42 is not a severe restriction since the large density of states in the reservoirs can compensate for the smallness of the tunnel coupling elements , and lead to a large tunneling rate compatible with the kondo effect , for instance . to summarize , in the absence of the cavity , the tunnel hamiltonian writes@xmath53 where the creation operator @xmath54 corresponds to an orbital @xmath46 with energy @xmath55 and wavefunction @xmath56 mainly localized inside element @xmath44 ... @xmath45 . one can use @xmath57 at lowest order in the dot / reservoir tunnel couplings . we now reexpress the photonic pseudo - potential scheme of the previous section in the framework of the tunneling model . for this purpose , one needs to decompose the field operator @xmath58 associated to tunneling charges on the nanocircuit orbital states . at lowest order in the dot / reservoir tunneling , one can use@xcite @xmath59 hence , hamiltonian ( [ htottild ] ) directly gives@xmath60 with@xmath61 note that in principle , the term in @xmath32 from eq.([htottild ] ) renormalizes @xmath55 , @xmath62 and the coulomb interaction term @xmath63 , but this is not essential for the physics we discuss below . since the photonic pseudo - potential modifies the potential landscape seen by the tunneling charges , it can naturally affect all the parameters in the tunneling model . first , cavity photons shift the orbital energy @xmath64 , with a coupling coefficient@xmath65 in general , @xmath66 strongly depends on the indexes @xmath47 and @xmath46 due to the space dependence of @xmath67 . this makes the cavity - induced orbital energy shifts particularly relevant experimentally . for a standard metallic reservoir , it is reasonable to disregard the dependence of @xmath66 on the orbital index @xmath46 , i.e. @xmath68 , because the properties of the electronic wavefunctions in the reservoir can be considered as constant near the fermi energy@xcite . in this framework , a behavior similar to the capacitive coupling of the cavity central conductor to the reservoirs is recovered . this type of reservoir / cavity coupling enables one to interpret data obtained for a quantum dot coupled to two normal metal leads@xcite . in contrast , @xmath66 may strongly depend on @xmath46 for @xmath69 , as illustrated for instance by the example of a dqd with an asymmetric @xmath67@xcite . the hamiltonian @xmath70 also contains coupling terms in @xmath71 with @xmath72 . these terms include photon - induced tunneling terms in @xmath73 , with @xmath74 , and photon - induced transitions internal to @xmath47 , in @xmath75 . in principle , all the above cavity / nanocircuit coupling terms can coexist . in practice , we expect the terms in @xmath66 to be dominant over the tunnel terms in @xmath73 with @xmath74 , which involve a weak overlap between wavefunctions . nevertheless , as shown in the next section , the signatures of the photon - induced tunneling terms can be boosted by the large number of states they affect in the reservoirs . if @xmath67 varies slowly inside @xmath41 , the photon - induced transitions internal to @xmath41 are negligible , i.e. @xmath76 , since @xmath77 and @xmath78 are orthogonal by definition . however , these transitions become possible if the cavity field is weakly screened ( @xmath79 ) , or if a protrusion of a cavity conductor is placed close to one of the quantum dots in order to reinforce the spatial variations of @xmath80@xcite . transitions in @xmath81 can be particularly interesting in case of real or artificial spin orbit coupling which mixes spin and orbital states . in this case , the @xmath82 terms can correspond to spin - transitions inside a single quantum dot@xcite or a dqd@xcite . in the reservoirs , the @xmath83 terms are probably always non - resonant due to strong energy relaxation . hence , we will assume that these terms do not need to be treated explicitly because they can be included in a renormalization of the states @xmath84 . in practice , the potential @xmath67 can be evaluated numerically by removing all nanoconductors from the device and simulating the microwave electromagnetic field in the cavity in the presence of the metallic gates , sources and drains of the nanocircuit . this can be done using standard microwave simulation tools , which can take into account imperfect metals . it enables a realistic evaluation of the elements @xmath66 and @xmath73 . even if @xmath67 is not calculated numerically but replaced by a phenomenological expression , the expressions ( [ jj1 ] ) and ( [ jj2 ] ) remain interesting because they set constraints between the different @xmath66 and @xmath85 , which all depend on the same @xmath67 profile . we do not give details on the coulomb interaction term @xmath63 which stems directly from eq.([htottild ] ) . if coulomb blockade in the nanocircuit is already strong in the absence of the cavity , the effect of the cavity on interactions may be disregarded or treated perturbatively . in the next section , we discuss another situation where we assume that the effects of @xmath63 are negligible due to the large tunnel rate between a quantum dot and a normal metal reservoir . in the general case , coulomb interactions between the tunneling charges can lead to a large variety of effects , which we will not discuss in this work . to conclude , we obtain a decomposition of the system hamiltonian in terms of nanocircuit elements connected by tunnel couplings . while this is reminiscent of the coarse graining description of superconducting microcircuits frequently used in circuit qed , a full analogy is not possible due to the presence of the orbital degree of freedom in the mesoscopic qed case . due to this orbital degree of freedom , we obtain a large variety of electron / photon coupling terms . in this section , we have discussed the case of quantum dot circuits which raises most experimental efforts so far . however , the above approach can be used for any other type of system in which a decomposition in terms of a tunneling model is relevant . this includes quantum point contacts , molecular circuits , metallic tunnel junctions , and hybrid superconducting systems enclosing andreev bound states or majorana bound states , for instance . this last case is discussed in more details in section [ supra ] . many theoretical works on mesoscopic qed use as a starting basis a tunnel hamiltonian where all parameters ( orbital energies , tunnel rates , ... ) are perturbed by corrections proportional to the cavity electric field . these empirical approaches lack of a formal justification and can lead to unphysical predictions . our work provides a rigorous framework for the use of tunnelling models with a linear light / matter interaction ( the superconducting case will be discussed explicitly in section [ supra ] ) . as already mentioned above , our approach proceeds along a spirit similar to the pzw transformation , but with modifications necessary to take into account the specificities of mesoscopic qed , among which a strong space dependence of the cavity electric field on the scale of a nanocircuit , and boundary conditions provided by cavity conductors and voltage - biased dc electrostatic gates . interestingly , the effect of boundary conditions provided by grounded conductors on the pzw transformation has been discussed recently@xcite . however , this ref . considers only neutral atomic systems affected by locally constant cavity fields ( dipolar coupling limit ) . in the above sections , superconductivity was not explicitly taken into account . it is important to discuss the generalization of our approach to the case of superconducting hybrid nanocircuits , considering the present interest for majorana fermions@xcite or andreev bound states for instance@xcite . for simplicity , we consider below standard bcs superconducting correlations characterized in the mean field approach by a superconducting gap @xmath86 coupling electrons and holes with opposite spins . the gauge invariant minimal coupling hamiltonian of eq.([1 ] ) can be generalized as@xmath87 with @xmath88 the field operator for tunneling electrons with spin @xmath89 and@xmath90 the term @xmath91 is the straightforward generalization of eq.([1 ] ) to the spin dependent case , i.e. the single electron term @xmath92 can have a structure in spin - space , due to magnetic fields and/or spin - orbit coupling , for instance . hence , @xmath7 must be replaced by @xmath93 $ ] in the first term of eq.([1 ] ) and by @xmath94 in eq.([coul ] ) . due to gauge invariance , @xmath92 can involve the electron momentum through @xmath95 only . importantly , we define @xmath86 as a gauge invariant quantity , so that the phase parameter @xmath96 ensures the gauge invariance the pairing term of @xmath97 . the form used above for @xmath96 is valid provided photon - induced magnetic effects can be disregarded , in agreement with the approach of section [ pseudop ] . using eqs.([unit ] ) and ( [ unit2 ] ) with @xmath98 replaced by @xmath99 , hamiltonian ( [ hsupr ] ) is transformed into@xmath100 where @xmath101 is again the generalization of eq.([htottild ] ) to the spin dependent case . importantly , eq.([bim ] ) shows that in the photonic pseudo - potential scheme , the phase of the gap term is not affected by photons anymore . this represents one more advantage of this scheme . this result remains valid in the case of d - wave or p - wave superconducting correlations@xcite . in the case where electron - electron interactions can be disregarded , bogoliubov - de gennes equations are a widely used approach@xcite , which enables a diagonalization of the circuit hamiltonian in terms of quasiparticles with creation operator @xmath102 . one can define quasiparticles modes in the absence of the cavity , by [email protected]_{n}(\overrightarrow{r})\ ] ] with a spinorial wavefunction @xmath104\ ] ] fulfilling @xmath105 with @xmath106 $ ] and @xmath107{cc}\widetilde{h}_{\mathcal{t}}(\vec{r } ) & \delta(\vec{r})\sigma_{0}\\ \delta^{\ast}(\vec{r})\sigma_{0 } & -\sigma_{y}\widetilde{h}_{\mathcal{t}}^{\ast}(\vec{r})\sigma_{y}\end{array } \right ] ~\text{.}\ ] ] above , @xmath108 and @xmath109 are the identity and the second pauli matrices in spin space . from the definition of @xmath102 , one gets @xmath110 with@xmath111@xmath112@xmath113 and @xmath114 , @xmath115 , @xmath116 pauli matrices in nambu space . in this section , we discuss the case of a nanocircuit enclosing majorana bound states , coupled to a cavity*. * in principle , these self - adjoint bound states can appear at the interface between the topological and non - topological superconducting phases of a nanostructure@xcite . although an indirect majorana bound state / cavity coupling mediated by a superconducting quantum bit or a josephson junction has been considered in many works@xcite , a direct coupling can also exist@xcite . using our formalism , we can provide a general hamiltonian describing such a situation . the direct majorana bound state / cavity coupling occurs due an overlap between neighboring majorana bound states , caused by the finite size of the nanocircuit . this overlap is naturally dependent on the photonic pseudo - potential . proposals to obtain majorana bound states in condensed matter systems rely on the use of superconducting elements , which calls for the use of the results of section [ supra ] . for simplicity , we disregard coulomb interactions and use the framework of the bogoliubov - de gennes equations of section [ bdgsec ] . to remain general , we do not specify the details of the system . we consider an ensemble of majorana bound states with creation operators @xmath117 and wavefunction @xmath118 , localized inside a nanoconductor . in general , the nanoconductor also encloses a continuum of ordinary fermionic states @xmath119 with wavefunction @xmath120 located above an energy gap . we assume that the nanoconductor is tunnel coupled to a fermionic reservoir with orbital states @xmath121 with wavefunctions @xmath122 . such a reservoir can be used to measure the conductance through the majorana system@xcite . in the limit of weak tunneling , one can reexpress eq.([bdg ] ) by using the ensemble of the operators @xmath123 with @xmath124 , following similar considerations as for the tunneling model of section [ tunnel ] . this gives:@xmath125 above , the coefficients @xmath126 have expressions similar to the @xmath127 coefficients of eqs . ( [ htot]-[ho ] ) with @xmath128 replaced by @xmath129 . we use energies @xmath130 for the orbitals in the reservoir ( @xmath131 ) and the nanoconductor continuum of states ( @xmath132 ) , and a tunnel coupling parameter @xmath133 between overlapping majorana bound states . the majorana bound states and the reservoir are connected by tunnel coupling terms in @xmath134 . for simplicity , we have omitted the dc tunnel coupling between the reservoir and the nanoconductor continuum of states . photons can shift all the above mentioned parameters , due to the two last lines of eq.([tt ] ) . in particular , photons can modify the tunnel coupling between neighboring majorana bound states , due to the terms in @xmath135 , as discussed in ref.@xcite . note that in hamiltonian ( [ tt ] ) , there is no dc coupling between the majorana bound states and the nanoconductor continuum of states , which are assumed to be orthogonal by construction of @xmath136 and @xmath137 . however , there can be photon - induced transitions between the majorana bound states and the nanocircuit continuum of states due to terms in @xmath138 and @xmath139 . effects related to these last transitions have been discussed in @xcite . our approach provides a general justification for such terms . we now give an example of a specific prediction given by hamiltonian ( [ ht ] ) . we consider a cavity coupled to a `` quantum rc circuit '' made out of a single quantum dot tunnel contacted to a single normal metal reservoir . we assume that the dot has a large intrinsic level spacing so that a single dot orbital with energy @xmath140 needs to be considered . the dot is capacitively coupled to a dc gate electrode , which enables one to tune @xmath140 . we study the effects of the dot on the photonic mode with frequency @xmath141 . following the previous section , the total hamiltonian of the system is:@xmath142 the hamiltonian @xmath143 of the dot and its reservoir involves the lead density of states @xmath144 . for later use we have introduced a cutoff @xmath145 . the terms in @xmath146 describes tunneling between the dot and the reservoir , whose orbitals correspond to fermionic creation operators @xmath147 and @xmath148 , respectively , with @xmath149 and @xmath150 . one can use extra bosonic modes @xmath151 with @xmath152=\delta(\epsilon-\epsilon^{\prime})/n_{0}$ ] to describe the intrinsic damping of the cavity , i.e.@xmath153 the coupling between the quantum dot circuit and the cavity is described by the term @xmath154 with@xmath155 it is not necessary to write explicitly the coupling between the chemical potential of the reservoir and the cavity because since there is a single reservoir , it is only the difference between the potentials of the dot and the reservoir which matters . note that the terminology `` quantum rc circuit '' mainly refers to the fact that no dc current can flow through the device and dissipation is due to transport through a single tunnel barrier . however , in the general case , the concept of a local capacitor is not sufficient to describe the coupling of the electromagnetic field to the device , since the photonic pseudo potential modifies @xmath140 and @xmath146 simultaneously . in practice , one can measure the response of the cavity to a microwave drive with frequency @xmath156 , which can be described by@xmath157 above , we have disregarded coulomb interactions between the tunneling charges ( including the contribution from the @xmath158 term arising from the unitary transformation of section ii.b ) , because we assume that these interactions are smaller than the tunnel rate @xmath159 between the dot and the reservoir . such a simplification is frequently performed for interpreting standard current measurements in open quantum dot circuits@xcite . in practice , one can check the relevance of this hypothesis by measuring for instance the ac current through the dot versus @xmath140 . . the wavy lines correspond to photonic propagators , the simple dashed lines to bare electronic propagators in the normal metal reservoir , and the double full lines to electronic propagators in the quantum dot dressed by the dot / reservoir tunneling processes.,title="fig:"][diagrams ] our purpose is to identify signatures of the terms in @xmath160 , which could become visible in the cavity response , considering the recent progress on the mesoscopic qed technology@xcite . in the framework of the linear response theory , the cavity electromagnetic response to @xmath161 is determined by the green s function@xcite @xmath162 with @xmath163\right\rangle $ ] and @xmath164 . one can use the exact relation @xmath165 with @xmath166 the free cavity green s function . in the simplest model where @xmath167 and @xmath168 are independent of energy , the constant @xmath169 corresponds to pure cavity damping . in the following , we consider the limit @xmath170 @xmath171 which corresponds to standard cavities . in the limit @xmath172 , one can use a random phase approximation which leads to @xmath173 with @xmath174 the electronic response function calculated in the absence of the cavity . in the framework of a non - interacting diagrammatic calculation , @xmath175 can be decomposed into the eleven contributions represented in fig . our calculation is not perturbative in @xmath146 since we use an exact expression for the dressed propagator of electrons on the dot , corresponding to the double full lines in fig.2 . there are nine types of diagrams corresponding to the possible ordered pairs formed with the interaction constants @xmath176 , @xmath160 , and @xmath177 . the diagrams of fig.2a correspond to pure contributions from photon - induced orbital shifts in @xmath178 . the diagrams of figs . 2b and c correspond to contributions from photon - induced tunneling in @xmath179 , @xmath180 or @xmath181 . note that the contributions in @xmath179 have been separated into the diagrams of fig . 2b which diverge logarithmically for @xmath182 large and the bottom diagrams in fig.2c which are regular . there also exist interferences between photon - induced orbital shifts and photon - induced tunneling , as shown by the diagrams of fig . 2d which depend on @xmath183 or @xmath184 . the constants @xmath185 and @xmath182 set the scale of variations of @xmath186 . therefore , in the limit @xmath187 , one can replace @xmath188 by @xmath189 in eq.([rep ] ) . this gives a cavity frequency pull @xmath190 and a cavity damping pull @xmath191 . in the limit @xmath192 and at zero temperature , one obtains @xmath193 where @xmath194^{2}\gamma + 4\pi\alpha\operatorname{re}[t\gamma^{\ast}]n_{0}\varepsilon_{d}-\alpha ^{2}\gamma\right ) \nonumber\end{aligned}\ ] ] stems from the diagrams of fig.2,a , c and d , and@xmath195+\left\vert \varepsilon_{d}\right\vert ( 2\arctan[\frac{\gamma}{\left\vert \varepsilon _ { d}\right\vert } ] -\pi)\right ) \nonumber\end{aligned}\ ] ] stems from the diagrams of fig.2,b . note that the cutoff @xmath196 on the reservoir density of states is necessary to avoid a divergence of @xmath197 . the cavity damping pull is given by@xmath198n_{0}\varepsilon_{d}\right ) ^{2}~\text{. } \label{pos}\ ] ] interestingly , @xmath199 and @xmath200 vanish for @xmath201 , whereas @xmath197 vanishes only for @xmath202 . above , we have implicitly included the spin degree of freedom in the model . in the spinless case , @xmath203 and @xmath200 should be divided by a factor 2 . from eqs.([a1]-[pos ] ) , at low temperatures , when the cavity has a frequency smaller than the dot / reservoir tunnel rate @xmath185 and the reservoir bandwidth @xmath182 , and when @xmath204 , the cavity frequency pull and the damping pull fulfill the relation @xmath205 with@xmath206 this property is a manifestation of the korringa - shiba relation@xcite , whose universality is still actively discussed for quantum rc circuits @xcite . remarkably , we find that this universality is broken for @xmath160 finite . hence , one can identify the existence of the @xmath160 coupling by comparing the @xmath207 and @xmath208 curves ( fig . 3 ) or studying @xmath209 ( fig . 4 ) . here , we present results for @xmath210 because from eqs . ( [ jj1]-[jj2 ] ) , this is the most probable regime of parameters . when @xmath204 , @xmath203 and @xmath211 show identical variations with @xmath140 , and are both even functions of @xmath140 , while @xmath205 . when @xmath212 , @xmath211 presents a resonance with @xmath140 which is wider than that of @xmath203 ( fig.3 , middle panel ) . a good resolution on the tails of the @xmath203 and @xmath211 resonances could also reveal that @xmath213 and @xmath200 are not even with @xmath140 in the general case , except if @xmath214=\pi/2 $ ] ( fig.3 , bottom panel ) . in principle , even a relatively small @xmath160 can be detected . indeed , for @xmath215 , @xmath209 already significantly deviates from @xmath216 at @xmath217 where the signals @xmath203 and @xmath211 are maximal ( fig.4 , pink curves ) . this is because the effect of the photon - induced tunneling term is boosted by the large number of states involved in the reservoirs . to conclude , in order to reveal the photon - induced tunneling terms , one promising possibility is to compare the cavity frequency pull and damping pull caused by a quantum dot with a single normal metal reservoir . even a relatively weak photon - induced tunneling can affect these quantities because the large number of reservoir states reinforces the tunnel effect . and @xmath211 versus @xmath140 for different values of @xmath160 . we have used @xmath218 , @xmath219 , and pulsation scales @xmath220 and @xmath221 . ] versus @xmath140 for different values of @xmath160 . we have used the same values of @xmath146 , @xmath182 and @xmath222 as in fig.3 ] in principle , circuit - qed should enable an accurate determination of @xmath209 thanks to the use of resonant techniques . the fabrication of a quantum dot with a single normal metal reservoir inside a coplanar cavity is fully accessible with present techniques . however , so far , the closest case studied experimentally is a quantum dot with two normal metal reservoirs , which leads to more complicated physics due to an asymmetric coupling of the two reservoirs chemical potentials to the cavity@xcite . different types of quantum rc circuits based on 2-degs but also carbon nanotubes or semiconducting nanowires deserve an investigation . the ratio @xmath223 should be strongly dependent on the type of nanoconductor and the contacts configuration used . the physics discussed in this section is related to the problem of a quantum rc circuit which is not coupled to a cavity but directly excited by a classical microwave field with amplitude @xmath224 and frequency @xmath225 , which corresponds to a term in @xmath226 . at low frequencies and @xmath227 , it has been predicted@xcite that a spinless quantum rc circuit excited classically should have the same admittance as an rc circuit , with a universal resistance @xmath228 . however , this result was obtained in a purely capacitive coupling scheme i.e. @xmath204 . assuming that the admittance of the circuit is determined from the current @xmath229 with @xmath230 , this value of @xmath231 can be recovered by calculating the charge susceptibility @xmath232 , i.e. the diagram of fig . 2.a , which leads to @xmath233 because @xmath204 . however , in the general case , one should calculate the current response of the quantum rc circuit from @xmath234 because the oscillating drive can also modify the tunnel barrier transparency . only the diagram of fig . 2.a and half of the diagrams of fig . 2d should contribute to @xmath231 , whereas the diagrams of fig.2.b leading to the logarithmic term of eq.([a2 ] ) should not contribute . therefore , @xmath231 should show a behavior qualitatively different from @xmath209 . we will not discuss the detailed behavior of @xmath231 here . we will simply point out that for @xmath212 , @xmath231 is not universal , and furthermore , @xmath231 and @xmath209 are not trivially related . interestingly , the universal value @xmath228 expected for a spinless system with @xmath204 has been checked experimentally in the limit @xmath235 for a 2-deg based circuit@xcite . however , this measurement is not incompatible with a finite @xmath160 , due to its uncertainty of the order of @xmath236@xcite . note that since @xmath231 and @xmath209 are qualitatively different signals , it can be interesting to measure both in the same experiment in order to obtain more information on the system . more generally , the study of tunneling effects can benefit from a simultaneous measurement of the cavity response and the current through the nanocircuit . such a joint measurement has already been realized in the case of a dqd . the cavity dispersive shift is directly sensitive to the population imbalance between the dqd bounding and antibounding states whereas the current though the dqd corresponds to a more complex combination of state populations and tunnel rates . the simultaneous measurement of the two signals gives stronger constraints to determine the system parameters@xcite . another possibility to gain more information on mesoscopic qed systems might be to measure cross - correlations between the electronic current and the cavity output field . importantly , the example of the quantum rc circuit shows that dissipation of the cavity photons has to be considered carefully , since an open quantum dot circuit with a single contact can already induce photon dissipation . on a more general level , photonic dissipation in mesoscopic qed is a very rich problem . another interesting possibility would be to prepare non - trivial quantum photonic states by combining the effects of photonic and electronic dissipation with the non - linearity of quantum dot circuits , by analogy with the methods developped for two level systems@xcite . this so - called `` reservoir engineering '' could benefit from the specificities of nanocircuits like for instance the use of reservoirs with specific electronic orders or finite bias electronic transport . mesoscopic qed has common ingredients with atomic cavity qed , like the relevance of an orbital degree of freedom , which does not exist for superconducting qubits . it has also common ingredients with standard circuit qed , like the tunneling physics and the strong inhomogeneities of the photonic modes . it is therefore necessary to develop a specific description which combines all these ingredients . in the case where photon induced magnetic effects can be disregarded , one can express the coupling between tunneling quasiparticles and photons in terms of a scalar photonic pseudo - potential . in the framework of a tunneling model , this leads to photon - induced orbital energy shifts , which coexist with photon - induced tunneling terms and photon - induced local orbital transitions . * * to illustrate the richness of our approach , we have discussed the example of a cavity coupled to a quantum rc circuit , i.e. a single quantum dot coupled to a single normal metal reservoir . the photon - induced tunneling terms between the dot and the reservoir induce a non - universal relation between the cavity frequency pull and the cavity damping pull , contrarily to what is expected with purely capacitive coupling schemes at low temperatures@xcite . this case represents only one example of use for our formalism . we have given explicit hamiltonians for the cases of a multi - quantum dot circuit in a cavity , and a superconducting nanostructure enclosing majorana bound states coupled to a cavity . one could also consider cavities coupled to quantum point contacts , molecular circuits , and metallic tunnel junctions , for instance . _ we acknowledge useful discussions with m. baillergeau , l. e. bruhat , c. cohen - tannoudji , m. c. dartiailh , m. r. delbecq , m. p. desjardins , r. b. saptsov , m. trif , and j. j. viennot . this work was financed by the erc starting grant cirqys , the eu fp7 project se2nd[271554 ] and the anr - nanoquartet [ anr12bs1000701 ] ( france ) . _ _ _ of the perfect conductors considered for the generation of the photonic modes . the photonic field in yellow is inhomogeneous , as represented by darker yellow areas near the cavity protusions , and by the white ( screening ) areas between the small purple conductors ( c ) effective orbital nanocircuit @xmath237 . ] in the main text , eqs ( [ 1])-([6 ] ) have been introduced on the basis of a physical discussion and gauge invariance considerations . it is also instructive to use an effective model which separates spatially the tunneling and plasmonic parts of the nanocircuit . note that this effective model does not aim at calculating quantitatively the nanocircuit / cavity couplings . it rather aims at providing a full mathematical justification for the form of eqs . ( [ 1])-([6 ] ) and a deeper insight on the underlying physics . for simplicity , we assume that the nanoconductors host only individual electronic orbital modes since their electronic density is generally very low . we assume that the plasmonic modes in the fermionic reservoirs ( blue elements in fig . 1 ) are ultra fast due to the absence of dynamical coulomb blockade@xcite . in this case , one can decompose each fermionic reservoir into a purely orbital conductor ( dark grey elements in fig.5.a ) which contains no plasmons and is tunnel coupled to the nanocircuit , and perfect conductors ( small purple elements in fig.5.a ) which can host plasmonic modes , but are only capacitively coupled to the nanoconductors . these perfect conductors , or `` effective plasmonic reservoirs '' , can be placed such that the partial or full decrease of the photonic fields around the nanocircuit orbital states is reproduced at least qualitatively . we denote with @xmath238 the ensemble of the perfect conductors , which includes the cavity conductors , gate electrodes , and the effective plasmonic reservoirs ( see fig . some of these conductors are voltage biased and some others are left floating , like for instance the central conductor in a coplanar cavity . the rest of the circuit represents an effective orbital nanocircuit @xmath237 where tunneling physics prevails ( see fig . note that in principle , to grant current conservation , the `` orbital '' reservoirs have to be connected to voltage sources through wirings which host plasmonic modes , but one can assume that they are too far from the nanoconductors to have any significant influence on the nanocircuit / cavity coupling . for later use , we attribute to each conductor of the ensemble @xmath238 an index @xmath239 and we note @xmath240 and @xmath241 the corresponding surface and volume . in the following treatment we make a distinction between the conductors @xmath242 which are left floating with a constant total charge @xmath243 , and the conductors @xmath244 which are voltage - biased with a constant voltage @xmath245 or grounded ( @xmath246 ) . the generators imposing the voltages @xmath245 can be omitted from the description and included in electromagnetic boundary conditions since they are far from the cavity and nanocircuits . the volume @xmath247 complementary to @xmath248 hosts a charge distribution @xmath249 of particles @xmath176 with charges @xmath250 at positions @xmath251 , which belong to the nanocircuit @xmath237 . we also define the corresponding current distribution @xmath252 . in cavity qed , the treatment of the boundary conditions provided by the cavity conductors is usually omitted , because the atoms are very far from the cavity mirrors . therefore , a procedure based on a spatial fourier transformation can be used to separate the transverse electric field @xmath253 , which has no divergence , from the longitudinal electric field @xmath254 , which has no rotational@xcite . it is found that the value of @xmath254 is instantaneously imposed by the charge distribution @xmath255 associated to the atomic system . in contrast , the transverse fields @xmath253 and @xmath256 correspond to propagating modes . in mesoscopic qed , it is necessary generalize this procedure to account for the presence of the perfect conductors of the ensemble @xmath238 , including voltage - biased electrostatic gates . as already mentioned in the previous section , we use a charge distribution @xmath255 which includes the charges from the @xmath237 subsystem only . we will take into account the screening charges on the surface of the conductors of @xmath238 through electromagnetic boundary conditions . inside @xmath257 , @xmath255 , @xmath258 , and the total electric and magnetic fields @xmath259 and @xmath260 are related by maxwell s equations . to understand how these relations are affected by the screening charges lying on @xmath238 , we introduce the hodge decomposition of a generic field @xmath261 as @xmath262 , where @xmath263 has a finite divergence but no rotational , @xmath264 has a finite rotational but no divergence , and @xmath265 has none@xcite . a similar decomposition was invoked very recently in the context of cavity qed @xcite . however , in this ref . , the effects of the dc voltage biases @xmath245 or the floating charges @xmath243 are not considered ( i.e. @xmath246 is used for any @xmath239 ) . besides , the volume @xmath257 outside the cavity conductors is assumed to be bounded . this hypothesis is not possible in our case due to openings in the cavity planes , which are necessary to connect the nanocircuits to their outside voltage bias circuitry . hence , we need to perform the hodge decomposition in the fully general case . below , we explain how to perform this task . it is convenient to introduce a potential @xmath11 and a green s function @xmath266 which fulfill the static laplace equations:@xmath267 and @xmath268 for @xmath269 , with boundary conditions@xmath270@xmath271 and @xmath272@xmath273 above , @xmath274 is the outward pointing unit vector perpendicular to @xmath240 . one can check that each of these two sets of eqs . has a unique solution provided @xmath275 is not empty@xcite . * * from eqs . ( [ bbc0]-[bc2 ] ) and the maxwell eqs . , one can check that:@xmath276@xmath277 and@xmath278 while @xmath279 is still valid . we also find @xmath280 with@xmath281 and@xmath282 the harmonic component @xmath283 represents a static contribution to the electric field , set by the charges @xmath243 or potentials @xmath245 imposed on the conductors @xmath239 . thanks to the use of @xmath266 , the definitions ( [ d1]-[d4 ] ) take into account the effect of the screening charges lying on @xmath238 . using the maxwell equations and the expressions ( [ d0]-[d4 ] ) , one can check that the longitudinal components @xmath284 and @xmath285 are set instantaneously by @xmath255 since @xmath286 for @xmath287 . furthermore , @xmath288 , @xmath253 and @xmath289 follow an independent system of propagation equations , i.e. @xmath290@xmath291 for @xmath287 . the above equations have to be supplemented with the boundary conditions@xmath292@xmath293 with @xmath294 any trajectory connecting conductors @xmath239 and @xmath46 . at this stage , it is convenient to introduce scalar and vector potentials @xmath295 and @xmath296 in the coulomb gauge i.e. @xmath297 . we also define the decomposition @xmath298 , such that@xmath299@xmath300 and@xmath301 a comparison between eqs . ( [ bc0]-[bc2 ] ) , and ( [ b1],[bb1],[bb2 ] ) leads to the identification @xmath302 up to a global constant which one can disregard . hence , @xmath284 but also @xmath303 are instantaneously set by the charge distribution @xmath255 . in contrast , the transverse fields @xmath253 and @xmath304 are determined by @xmath296 which follows the propagation equation @xmath305 given by eq.([b4 ] ) , with @xmath306 , and boundary conditions similar to eqs.([bb1])-([bb2 ] ) . in this framework , a proper classical lagrangian is @xmath307 with @xmath308 in the sums . the lagrangian equations deriving from @xmath309 correspond to eq.([field ] ) and to the standard newton equations for the motion of particles @xmath176 . from eq . ( [ ll ] ) , one can check that @xmath310 is the conjugate of @xmath311 and @xmath312 is the conjugate of @xmath313 . therefore one can express the hamiltonian of the system as @xmath314 . this hamiltonian can be quantized straightforwardly by replacing @xmath315 , @xmath311 , @xmath316 and @xmath317 by operators @xmath318 , @xmath319 , @xmath320 and @xmath321 . this gives@xmath322 in principle , the ensemble of the charges @xmath308 includes tunneling electrons ( @xmath323 ) but also ions and valence electrons from an underlying crystalline structure . in the main text , the two latter are assumed to be decoupled from the cavity and are thus treated in a mean field approach . note that the harmonic potential term in the above hamiltonian is omitted by ref . @xcite . in our case , this term is crucial to account for the effect of dc electrostatic gates . assuming that the coupling between @xmath238 and @xmath237 is perturbative , the fields @xmath321 and @xmath324 can be expressed in terms of photonic modes calculated in the absence of @xmath325 , using for instance eq.([field ] ) with @xmath326 we treat explicitly only one of these modes in eq.([6 ] ) . screening charges on the nanocircuits reservoirs can strongly modify @xmath321 . this is why , in the general case , it is not adequate to make a perturbative treatment between the photonic field generated by the empty cavity and the whole nanocircuit . instead , the present work considers a perturbative treatment between the photonic field generated by the plasmonic modes in the whole mesoscopic qed device , and the effective orbital nanocircuit @xmath237 . similarly , in superconducting circuit qed , the cavity modes can be strongly renormalized by the presence of a superconducting quantum bit , and perturbation schemes must therefore be defined carefully@xcite . in this appendix , the renormalization of the bare cavity modes by the nanocircuit plasmons is described formally by supplementary boundary conditions imposed by effective plasmonic reservoirs . this enables a simple mathematical justification for the form of eqs ( [ 1])-([6 ] ) . in order to get realistic estimates of the fields @xmath321 and @xmath320 , one can make a numerical microwave simulation using the real device geometry , with the nanoconductors omitted ( see main text ) . the advantages of a formalism without the @xmath16 terms have already been discussed thoroughly for cavity qed@xcite . however , since the physical ingredients and measured quantities are different in cavity and mesoscopic qed , it is useful to discuss more specifically the advantages of the photonic pseudo - potential scheme of section [ ppp ] . the first quantity which is generally measured in mesoscopic qed experiments is the cavity frequency pull for a given cavity mode , which is modified by the nanocircuits coupled to the cavity . since the minimal coupling scheme and the photonic pseudo - potential scheme are related by a unitary transformation , they must of course predict the same cavity pull . it is nevertheless instructive to check this equivalence for a simple example . here , we use a simpler approach than in section iv , because we do not need to particularize the effect of the nanocircuit reservoirs and we will not discuss the cavity damping pull . we consider a cavity coupled to a nanocircuit with unperturbed many electron states @xmath327 which satisfy the general equation @xmath328 with@xmath329 and @xmath330 . we discuss the limit where the nanocircuit and the cavity are not resonant i.e. @xmath331 for any @xmath46 and @xmath51 . at second order in the cavity / nanocircuit coupling ( i.e. @xmath332 or @xmath333 ) , an elementary perturbation theory gives the cavity pull @xmath334 with@xmath335 note that @xmath336 depends on the state @xmath337 occupied by the nanocircuit . this can be used , in principle , for a noninvasive readout of the nanocircuit state in the non - resonant regime@xcite . equation ( [ pull ] ) can be obtained by using hamiltonian ( [ htottild ] ) or equivalently hamiltonian ( [ 1 ] ) . in order to check the equivalence between the two approaches , it is necessary to invoke the completeness relation @xmath338 for the nanocircuit states . interestingly , one can check that the @xmath16 terms in eq.([1 ] ) give a contribution @xmath339 to @xmath336 , which is not negligible for @xmath340 . we conclude that in the framework of mesoscopic qed where the cavity frequency pull is a central quantity , it is particularly important to eliminate in a rigorous way the @xmath16 terms if one wants to use a linear light / matter coupling . the photonic pseudo - potential scheme completes such a task . the cavity frequency pull is not the only means to characterize the interactions between a nanocircuit and a cavity . a more general workout of mesoscopic qed devices requires the knowledge of the matrix elements between the different nanocircuit eigenstates , generated by the cavity photons . another argument in strong favor of the use of the photonic pseudo - potential scheme is the general relation @xmath341 it indicates that the single - photon coupling elements decay more quickly with @xmath342 in the photonic pseudo - potential scheme . in practice one often uses a truncated space for the nanocircuit electronic states which is difficult to handle or determine globally . this is what we do for instance in section iv , where we consider a single orbital quantum dot . equation ( [ el ] ) shows that in this case it is more accurate to use eq.([htottild ] ) to predict the cavity behavior . finally , the photonic pseudo - potential scheme involves only photon - independent superconducting gap terms ( see section [ supra ] ) . this can be another significant advantage considering the already rich structure of the nanocircuit / light coupling in the absence of superconductivity . a. blais , r .- s . huang , a. wallraff , s. m. girvin , and r. j. schoelkopf , _ cavity quantum electrodynamics for superconducting electrical circuits : an architecture for quantum computation _ , physical review a * 69 * , 062320 ( 2004 ) . j. j. viennot , m. r. delbecq , m. c. dartiailh , a. cottet , and t. kontos , _ out - of - equilibrium charge dynamics in a hybrid circuit quantum electrodynamics architecture _ , phys . b * 89 * , 165404 ( 2014 ) . strictly speaking , this assumption is not valid for non ergodic reservoirs such as integrable billiards or anderson localized systems . our theory can be applied to these cases provided one keeps the @xmath46 dependence of the coupling parameter @xmath66 . the results of section [ supra ] can be generalized straighforwardly to d - wave or p - wave superconductivity by using a non - local gap term in @xmath343 in eq.([hsupr ] ) , with a different structure in spin - space in case of p - wave superconductivity .	understanding the interaction between cavity photons and electronic nanocircuits is crucial for the development of mesoscopic quantum electrodynamics ( qed ) . one has to combine ingredients from atomic cavity qed , like orbital degrees of freedom , with tunneling physics and strong cavity field inhomogeneities , specific to superconducting circuit qed . it is therefore necessary to introduce a formalism which bridges between these two domains . we develop a general method based on a photonic pseudo - potential to describe the electric coupling between electrons in a nanocircuit and cavity photons . in this picture , photons can induce simultaneously orbital energy shifts , tunneling , and local orbital transitions . we study in details the elementary example of a single quantum dot with a single normal metal reservoir , coupled to a cavity . photon - induced tunneling terms lead to a non - universal relation between the cavity frequency pull and the damping pull . our formalism can also be applied to multi quantum dot circuits , molecular circuits , quantum point contacts , metallic tunnel junctions , and superconducting nanostructures enclosing andreev bound states or majorana bound states , for instance .
18.5pt in a recent article @xcite blte and nienhuis performed numerical investigations of what they termed the fully - packed loop ( fpl ) model . this is a statistical model where the ensemble is the set of all combinations of closed paths on the honeycomb lattice that visit every vertex and do not intersect . the boltzmann weight of such a filling set of paths is just the exponential of the number of paths , i.e. the energy of a configuration is the number of closed loops used to cover the lattice . an example of a fully - packed configuration of loops on this lattice is shown in figure [ example ] . the partition function for this model may be represented as @xmath1 where the sum is over all @xmath2 , the coverings of the vertices of the hexagonal lattice by closed nonintersecting paths , @xmath3 is the number of paths in the covering @xmath2 , and @xmath4 is a generalized activity . this model was originally studied for its interest as the low - temperature limit of the @xmath5 vector lattice models @xcite . in this limit , the dimensionality of vectors @xmath4 is just the activity @xmath4 in equation ( [ zfpl ] ) . the partition function ( [ zfpl ] ) is apparently the generating function for the numbers of ways to cover the hexagonal lattice by any number of closed paths . its calculation in the thermodynamic limit is an interesting combinatorial problem . more recently , batchelor , suzuki and yung @xcite pointed out that previous authors @xcite had exploited an identification of the fpl model with the integrable lattice model associated to the quantum group @xmath0 @xcite . this integrable model is a vertex model on the square lattice where each link of the lattice can be in one of three states , and the vertex weights are given by the r - matrix for @xmath0 as in figure [ weights ] . the r - matrix depends on a deformation parameter @xmath6 , as well as a spectral parameter @xmath7 typical of integrable theories . denoting the @xmath0 partition function as @xmath8 , the precise idenfication is @xmath9 where @xmath10 is the volume of the lattice ( the number of hexagonal faces ) . since the model is integrable , much exact information can be derived . in particular , the model s bethe equations have been constructed and solved . ( 400,100 ) ( 50,25)(0,1)50 ( 25,50)(1,0)50 ( 50,25)a ( 50,75)b ( 25,50)i ( 75,50)j ( 100,50 ) @xmath11 \\ \end{array } $ ] one of the more important results that have been derived in this way is the existence of a phase transition in the model ( [ zfpl ] ) at @xmath12 @xcite . at larger @xmath4 , larger numbers of loops are favored and at smaller @xmath4 configurations with fewer loops are favored . it has been conjectured that this transition is between a large-@xmath4 phase where the average loop length is finite and a small-@xmath4 phase where this average is infinite . in section [ flsection ] a simple relation between the free energy of the fpl model and the ensemble average length of loops is derived . from the known solution to the bethe equations of the @xmath0 integrable lattice model , the free energy is identified and used to graph the exact value of the average loop length as a function of @xmath4 . identifying @xmath13 , it becomes apparent that @xmath14 corresponds to the integrable model for @xmath15 real and @xmath16 corresponds to @xmath15 purely imaginary . the former phase is known @xcite to be massive , in the sense that there is a gap in the spectrum of eigenvalues of the transfer matrix between the leading eigenvalue and the next - leading eigenvalue . by standard arguments @xcite , this implies a finite correlation length . the gap tends to zero as @xmath15 goes to zero , showing that @xmath12 is a critical point of the model ( [ zfpl ] ) . in section [ xisection ] this correlation length is studied by considering the spectrum of eigenvalues of this transfer matrix . the spectrum may be deduced directly from the model s bethe equations . in the case that the transfer matrix is symmetric and therefore has real eigenvalues , the correlation length is related to the maximum eigenvalue @xmath17 and next - leading eigenvalue @xmath18 by @xmath19 18.5pt the r - matrix of the @xmath21 quantum group is an @xmath22 matrix that may be interpreted as a matrix of boltzmann weights of a vertex model on the square lattice as shown in figure [ weights ] . since this matrix satisfies the yang - baxter equation , the associated transfer matrix commutes with itself evaluated at differing values of the spectral parameter and the model is exactly solvable by a recursive set of @xmath23 nested bethe anstze @xcite . the formula for eigenvalues of the transfer matrix for the range of parameters , @xmath24 and @xmath15 real and positive , is @xmath25 where the product is over roots , @xmath26 of a set of bethe equations and we have neglected terms that do not contribute in the thermodynamic limit . in this limit the @xmath26 are distributed in the interval @xmath27 $ ] with the density , @xmath28}.\ ] ] for eigenvalues near the maximum eigenvalue , the changes in the distribution are parameterized by the locations of holes @xmath29 , @xmath30 , @xmath31 , according to @xmath32}{\sinh[(d+1)m\gamma ] } \sum_{h=1}^{n_q } e^{-2im\theta^q_h}. \label{density}\ ] ] the numbers of holes are constrained to satisfy the relation , @xmath33 the formulas ( [ eigenvalue ] ) , ( [ density ] ) may be combined in the thermodynamic limit to yield a formula for eigenvalues of the transfer matrix of the integrable model , @xmath34 \,d\lambda . \label{formula}\ ] ] 18.5pt the formulas of the preceding section in the case @xmath35 yield directly the free energy density of the model ( [ zfpl ] ) for the @xmath14 phase as the logarithm of the maximum eigenvalue of the transfer matrix , rescaled by the factor of equation ( [ identification ] ) . this free energy was actually derived in 1970 by baxter @xcite as the solution to a weighted three - coloring problem on the honeycomb lattice . the free energy density of the fpl model in the @xmath14 phase is @xmath36 where @xmath37 , and @xmath38 . this function has an essential singularity at @xmath39 . the free energy for @xmath40 and with periodic boundary conditions is given in integral form in @xcite . it is interesting to note that for both phases , the free energy density gives the ensemble average length of loops . since a configuration @xmath2 on a lattice of @xmath10 faces has @xmath41 occupied links , the total length of loops is always @xmath41 . the average loop length of configuration @xmath2 is therefore @xmath42 . if we define the ensemble average loop length @xmath43 by @xmath44 then from inspection of equation ( [ zfpl ] ) it is clear that @xmath45 = \frac{2n}{n } z_{fpl}(n ) . \label{leqn}\ ] ] the general solution to this equation can be written up to quadrature by direct integration : @xmath46 where the lower limit of integration is an undetermined constant . in terms of the free energy density @xmath47 , this becomes @xmath48 the integral in equation ( [ lintegral ] ) can be evaluated by steepest descent . the result is @xmath49 } { n \frac{df_{fpl}}{dn}(n ' ) } \right]^n_c.\ ] ] the constant of integration may now be determined from the known value of @xmath43 at @xmath50 . as will be shown in section [ perturbsection ] , in this limit @xmath51 , @xmath52 , and @xmath53 . these imply that @xmath54 , so in the thermodynamic limit @xmath55 in this calculation we have neglected corrections of order @xmath56 to @xmath43 . a graph of the ensemble average loop length versus @xmath4 in the large-@xmath4 phase is shown in figure [ lengthgraph ] . this verifies the conjecture of @xcite that loop length diverges at the critical point . 18.5pt to obtain the correlation length , we must compute the expression ( [ formula ] ) for the minimal hole distribution . when @xmath35 , there are two choices for the @xmath57 . either @xmath58 and @xmath59 , or @xmath60 and @xmath61 . in each case , the eigenvalue gap is minimized for holes at @xmath62 where the sum in equation ( [ density ] ) after integration in ( [ formula ] ) is oscillatory . the transfer matrix of the model is symmetric at the point @xmath63 , and its eigenvalues are then real . after setting @xmath7 to this value , the correlation length of the model is given by equation ( [ xidef ] ) . considering the case @xmath58 and @xmath59 , we denote the next - leading eigenvalue for this hole distribution as @xmath64 . the equation ( [ formula ] ) together with the formula for densities ( [ density ] ) gives @xmath65 where @xmath66 are the integrals over roots , @xmath67 \,d\lambda . \label{integral}\ ] ] the integral in equation ( [ integral ] ) can easily be performed by contour integration . after introducing the variables @xmath68 and @xmath69 , the result for @xmath70 is @xmath71 \,d\lambda = \left\ { \begin{array}{ll } \frac{\pi}{m } [ 1-(z^2 q^{-1})^m ] , & m>0 \\ -2\pi\log z , & m=0 \\ \frac{\pi}{m } [ 1-(z^2 q)^m ] , & m<0 \end{array } \right . .\ ] ] substituting this result into equation ( [ formula2 ] ) gives @xmath72 after expanding the demoninator of the summand in a power series in @xmath73 , this may be resummed to the form , @xmath74.\ ] ] this form is now convergent at the symmetric point , @xmath75 or equivalently @xmath76 . we may therefore evaluate it there to obtain the correlation length according to equation ( [ xidef ] ) , @xmath77 this is the desired result , the correlation length of the fpl model where @xmath37 and @xmath78 , or equivalently @xmath79 . the other possible choice of holes , @xmath60 and @xmath61 may be computed in the same way to give @xmath80 this quantity is greater than ( [ xi ] ) for all @xmath78 , so it is not the inverse correlation length . for large @xmath81 , the inequality may be seen by considering the limiting forms of expressions ( [ xi ] ) and ( [ other ] ) . rigorously , the multiplicands in ( [ other ] ) may be seen to be greater than those in ( [ xi ] ) term by term in @xmath82 . 18.5pt the fpl model has a natural large-@xmath4 expansion which allows simple perturbative verifications of results . when @xmath4 is large , the dominant configurations are those with large numbers of loops . the perturbative procedure is to approximate the sum over states by including the configurations with the highest numbers of loops . on a hexagonal lattice with number of faces @xmath10 a multiple of three , there are three configurations with the maximum possible number of loops . in these states , one out of every three faces has a small loop around it and these small loops lie on a triangular lattice . a sample is shown in figure [ nodefects ] . these three configurations differ by translations and each has @xmath83 loops . ( 140,153 ) ( 20,0)(140,153 ) ( 60,0)(-20,0)(3,5)10(-10,17)(1,0)20(20,0)(-3,5)10 ( 120,0)(-20,0)(3,5)10(-10,17)(1,0)20(20,0)(-3,5)10 ( 30,51)(-10,17)(1,0)20(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 90,51)(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10(-10,17)(1,0)20 ( 150,51)(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10(-10,17)(1,0)20 ( 60,102)(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10(-10,17)(1,0)20 ( 120,102)(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10(-10,17)(1,0)20 ( 30,153)(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 90,153)(20,0)(-3,-5)10(10,-17)(-1,0)20(-20,0)(3,-5)10 ( 150,153)(10,-17)(-1,0)20(-20,0)(3,-5)10 the smallest change in the number of loops that can be made is to introduce a defect somewhere in one of the maximal configurations , as shown in figure [ onedefect ] . there are @xmath84 different such defects that can be introduced and each reduces the number of loops by 2 . ( 140,153 ) ( 20,0)(140,153 ) ( 60,0)(-20,0)(3,5)10(-10,17)(1,0)20(20,0)(-3,5)10 ( 120,0)(-20,0)(3,5)10(-10,17)(1,0)20(20,0)(-3,5)10 ( 30,51)(-10,17)(1,0)20(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 90,51)(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10 ( 150,51)(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10(-10,17)(1,0)20 ( 60,102)(20,0)(-3,5)10(10,-17)(-1,0)20(-20,0)(3,-5)10(-20,0)(3,5)10(-10,17)(1,0)20 ( 120,102)(-20,0)(3,5)10(-10,17)(1,0)20(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 30,153)(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 90,153)(20,0)(-3,-5)10(10,-17)(-1,0)20(-20,0)(3,-5)10 ( 150,153)(10,-17)(-1,0)20(-20,0)(3,-5)10 ( 90,85)(20,0)(-3,-5)10(-20,0)(3,-5)10(-10,17)(1,0)20 introducing defects in this way , we can reach all possible configurations . to see that this is so , we can represent a configuration by labelling the links on the lattice that do not contain part of a path . one of every three links is unoccupied , and every vertex touches one unoccupied link . these unoccupied links form a dimer configuration for the vertices of the lattice . if we draw rhombuses around every dimer and interpret the resulting picture as the projection of the edges of a stack of cubes , we see that a fpl configuration is equivalent to a stack of cubes . such an identification is shown in figure [ cubes ] . ( 300,150 ) ( 0,34 ) ( 0,0)(20,0)(-3,-5)10 ( 30,-17)(20,0)(-3,-5)10 ( 60,-34)(20,0)(-3,5)10 ( 0,34)(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 30,17)(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 60,0)(20,0)(-3,5)10(10,-17)(-1,0)20 ( 90,-17)(20,0)(-3,5)10 ( 0,68)(20,0)(-3,5)10(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 30,51)(20,0)(-3,5)10(10,-17)(-1,0)20 ( 60,34)(20,0)(-3,5)10(20,0)(-3,-5)10 ( 90,17)(20,0)(-3,5)10(10,-17)(-1,0)20 ( 30,85)(20,0)(-3,5)10(20,0)(-3,-5)10 ( 60,68)(20,0)(-3,5)10(10,-17)(-1,0)20 ( 90,51)(20,0)(-3,-5)10 ( 60,102)(20,0)(-3,-5)10 ( 90,85)(20,0)(-3,-5)10(10,-17)(-1,0)20 ( 120,68)(10,-17)(-1,0)20 ( 150,34 ) ( 0,0)(0,0)(0,1)34(0,0)(5,-3)30 ( 30,-17)(0,0)(0,1)34(0,0)(5,-3)30 ( 60,-34)(0,0)(0,1)34 ( 0,34)(0,0)(0,1)34(0,0)(5,-3)30 ( 30,17)(0,0)(0,1)34(0,0)(5,-3)30 ( 60,0 ) ( 90,-17)(0,0)(0,1)34(0,0)(-5,-3)30 ( 0,68)(0,0)(5,-3)30 ( 30,51 ) ( 60,34)(0,0)(0,1)34(0,0)(5,-3)30(0,0)(-5,-3)30 ( 90,17)(0,0)(-5,-3)30 ( 120,0)(0,0)(0,1)34(0,0)(-5,-3)30 ( 30,85)(0,0)(5,-3)30(0,0)(-5,-3)30 ( 60,68)(0,0)(-5,-3)30 ( 90,51)(0,0)(0,1)34(0,0)(5,-3)30(0,0)(-5,-3)30 ( 120,34)(0,0)(0,1)34(0,0)(-5,-3)30 ( 60,102)(0,0)(5,-3)30(0,0)(-5,-3)30 ( 90,85)(0,0)(5,-3)30(0,0)(-5,-3)30 ( 120,68 ) in this new representation , the action of inserting a defect is just the action of adding or removing a cube . this identification is exhibited in figure [ cubedefect ] . the result then follows that since every stack of cubes can be made by adding or removing cubes , every fpl configuration can be made from one of the maximal ones by inserting some combination of defects . ( 200,200 ) ( 0,0)(0,0)(0,1)34(0,0)(5,-3)30 ( 30,-17)(0,0)(0,1)34 ( 0,34)(0,0)(5,-3)30 ( 30,17 ) ( 60,0)(0,0)(0,1)34(0,0)(-5,-3)30 ( 30,51)(0,0)(5,-3)30(0,0)(-5,-3)30 ( 60,34)(0,0)(-5,-3)30 ( 120,0 ) ( 0,0)(0,0)(0,1)34(0,0)(5,-3)30 ( 30,-17 ) ( 0,34 ) ( 30,17)(0,0)(0,1)34(0,0)(5,-3)30(0,0)(-5,-3)30 ( 60,0)(0,0)(0,1)34(0,0)(-5,-3)30 ( 30,51)(0,0)(5,-3)30(0,0)(-5,-3)30 ( 60,34 ) ( 0,120 ) ( 0,0)(20,0)(-3,-5)10(-10,17)(1,0)20 ( 60,0)(-20,0)(3,-5)10(-10,17)(1,0)20 ( 0,34)(20,0)(-3,5)10(10,-17)(-1,0)20 ( 60,34)(10,-17)(-1,0)20(-20,0)(3,5)10 ( 30,17)(20,0)(-3,5)10(10,-17)(-1,0)20(-20,0)(3,5)10 ( 120,120 ) ( 0,0)(20,0)(-3,-5)10(-10,17)(1,0)20 ( 60,0)(-20,0)(3,-5)10(-10,17)(1,0)20 ( 0,34)(20,0)(-3,5)10(10,-17)(-1,0)20 ( 60,34)(10,-17)(-1,0)20(-20,0)(3,5)10 ( 30,17)(20,0)(-3,-5)10(-20,0)(3,-5)10(-10,17)(1,0)20 to obtain an approximation for the free energy , consider first the maximal state shown in figure [ nodefects ] . for a lattice of @xmath10 faces , this configuration has @xmath85 loops . there are 3 such configurations corresponding to the three - fold translational degeneracy of the state . to lowest order then @xmath86 $ ] . this result was used in section [ flsection ] to determine the asymptotics of the average loop length . allowing defects , there are @xmath87 locations for a defect and each defect reduces the number of loops by two . defects may be applied in any number and in any combination , giving the usual sum over disconnected diagrams . we can write this as the exponential of the connected diagram ( one defect ) and we will be correct except for the effects of excluded volumes which begin with two - defect connected diagrams and are therefore higher order . to the next order , @xmath88 \exp[o(n^{-4})]$ ] . perturbatively calculating the fpl free energy , we see that @xmath89 in conformity with baxter s result shown in equation ( [ ffpl ] ) . this point of view incidentally leads to a simple expression for the entropy density of the fpl configurations at @xmath90 . at this point , all configurations are weighted equally and @xmath91 is just the number of configurations , or the exponential of the entropy . then calculating the partition function is just the problem of counting the number of coverings of the honeycomb lattice by paths , which is the number of different possible stacks of cubes , which is the old combinatorial problem of counting plane partitions . elser @xcite has calculated the asymptotics of plane partitions for large arrays of numbers . the result applied to this case is entirely dependent on the shape of the boundary , even in the thermodynamic limit . this is to be expected when @xmath90 , because this is in the small-@xmath4 phase where the model is critical . for a lattice of @xmath10 faces and free boundary conditions , the maximum entropy is obtained for a hexagon - shaped boundary and in that case the partition function is asymptotically @xmath92.\ ] ] 18.5pt the ground state of the @xmath93 integrable lattice model is @xmath94-fold degenerate . this implies the existence of a notion of interfacial tension @xmath95 away from the critical point between regions of differing antiferromagnetic polarization . by considering finite - size corrections , de vega @xcite has derived transcendental equations for this interfacial tension and computed the asymptotic behavior of @xmath96 in the limits @xmath97 and @xmath98 . scaling arguments originally due to widom @xcite predict that the scaling relation , @xmath99 should hold near the critical point , @xmath100 or equivalently @xmath101 . it would be interesting to test this relation in this case , but we know of no explicit expression for the interfacial tension . away from the critical point however , a comparison can be made . the asymptotic behavior of the interfacial tension for @xmath98 was extracted by devega , and the result is @xmath102 in the case of the fpl model , @xmath35 , @xmath103 , and @xmath104 this result may be compared with a perturbative calculation . consider the sum over fpl states at large-@xmath4 with the constraint that boundary conditions are fixed to cause frustration in the bulk , as in figure [ interface ] . the configuration in that figure has the maximum number of loops possible and is the analog of the configuration shown in figure [ nodefects ] . denoting the sum over defects in this configuration by @xmath105 , the interfacial tension is defined to be the change in free energy per unit length of the interface : @xmath106 where @xmath107 is the vertical size of the lattice . for a lattice of n faces , the maximum number of loops possible in the presence of the constraint is @xmath108 instead of @xmath85 . the maximal state in the presence of the constraint is now @xmath109-fold degenerate , because there are @xmath110 locations near the interface where defects may be freely introduced without changing the number of loops . to lowest order therefore , @xmath111 reading off the exponents , we have from equation ( [ sdef ] ) the result that @xmath112 equation ( [ s ] ) is apparently consistent with the large-@xmath15 asymptotics derived in @xcite . equation ( [ s ] ) together with equation ( [ xi ] ) show that @xmath113 in the fpl model . more generally , from the correlation length calculation it is clear that for large @xmath15 the leading behavior of the correlation length for any value of @xmath23 will always be @xmath15 , and the leading behavior of @xmath96 is always @xmath114 . h. w. j. blte , b. nienhuis , _ phys . lett _ * 72 * , 1372 ( 1994 ) . b. nienhuis , _ phys . lett . _ * 49 * , 1062 ( 1982 ) . e. domany , d. mukamel , b. nienhuis , a. schwimmer , _ nucl . phys . _ * b190 * , 279 ( 1981 ) . m. t. batchelor , j. suzuki , c. m. yung , preprint cond - mat 9408083 , august 1994 . s. o. warnaar , b. nienhuis , _ j. phys . * 26 * , 2301 ( 1993 ) . n. yu . reshetikhin , _ j. phys . * 24 * , 2387 ( 1991 ) . o. babelon , h. j. de vega , c. m. viallet , _ nucl . phys . _ * b200 * 266 ( 1982 ) ; _ nucl . phys . _ * b220 * 283 ( 1983 ) . r. j. baxter , _ j. math . phys . _ * 11 * , 784 ( 1970 ) . h. j. de vega , _ j. phys . a : math . gen . _ * 20 * , 6023 ( 1987 ) . r. j. baxter , _ exactly solved models in statistical mechanics _ , new york , academic , 1982 . b. widom , _ j. chem . phys . _ * 43 * , 3892 ( 1965 ) . v. elser , ph.d . thesis , university of california , berkeley , 1984 .	the fully - packed loop model of closed paths covering the honeycomb lattice is studied through its identification with the @xmath0 integrable lattice model . some known results from the bethe ansatz solution of this model are reviewed . the free energy , correlation length , and the ensemble average loop length are given explicitly for the many - loop phase . the results are compared with the known result for the model s surface tension . a perturbative formalism is introduced and used to verify results . 8.5 in 6.5 in -.35 in arch - ive/9411132 + .2 in * correlation length and average loop length of the fully - packed loop model * .5 in anton kast + .5 in
an undirected graph is formed by a set of vertices and a set of undirected edges , with each edge connecting between two different vertices . a feedback vertex set ( fvs ) for such a graph is a set of vertices intersecting with every cycle of the graph . in other words , the subgraph induced by the vertices outside the fvs contains no cycle ( it is a forest ) @xcite . the feedback vertex set problem aims at constructing a fvs of small cardinality for a given undirected graph . it is a fundamental nondeterministic polynomial - complete ( np - complete ) combinatorial optimization problem with global cycle constraints @xcite . in terms of complete algorithms , whether a graph @xmath0 has a fvs of cardinality smaller than @xmath1 can be determined in time @xmath2 @xcite . and an fvs of cardinality at most two times the optimal value can be easily constructed by an efficient polynomial algorithm @xcite . an optimal fvs is a feedback vertex set whose cardinality is the global minimum value among all the feedback vertex sets of the graph . for a given graph , an optimal fvs can be constructed in an exact way in time @xmath3 @xcite , where @xmath4 denotes the total number of vertices in the graph . applied mathematicians have obtained rigorous lower and upper bounds for the optimal fvs problem @xcite and have proved its tractability for graphs with specific structures ( see for example , @xcite and references cited therein ) . due to the np - complete nature of the fvs problem , in general it is not feasible to construct optimal feedback vertex sets for large cycle - rich graphs . an important question is then to design efficent heurstic algorithms that are able to obtain near - optimal fvs solutions for given graph instances . such a task is quite nontrivial . a major technical difficulty is that cycles are global objects of a graph and therefore the existence of cycles can not be judged by checking only the neighborhood of a vertext . ( similar difficulties exist in other combinatorial optimization problems with global constraints , such as the steiner tree problem @xcite and the optimal routing problem @xcite . ) in ref . @xcite , one of the authors succeeded in converting the fvs problem to a spin glass problem with local interactions . the fvs problem was then studied from the spin glass perspective , and a message - passing algorithm , belief propagation - guided decimaton ( bpd ) , was impletmented to solve the fvs problem heuristically for single graph instances . this bpd algorithm is quite efficient in terms of computing time and computer memory ( since there is no need of cycle checking ) , and it can obtain fvs solutions that are very close to the optimal ones when applied on large random graph instances and regular lattices @xcite . for the undirected fvs problem it is not yet known whether simple local search algorithms can achieve equally excellent results as the bpd algorithm . motivated by this question , we complement the message - passing approach in this paper by implementing and testing a simulated annealing local searching ( sals ) protocol @xcite for the undirected fvs problem . a similar algorithmic study has already been undertaken in @xcite for directed graphs . here we modify the microscopic search rules of @xcite to make it applicable to undirected graphs . in the sals algorithm , an order is defined for the vertices outside the focal fvs , and this order is constrained by a set of local vertex constraints . our simulation results suggest that this local search algorithm is comparable in performance to the bpd algorithm at least for random graph instances . the feedback vertex set problem has wide practical applications in the field of computer science ( such as integrated circuit design and database management ) . although not yet seriously explored , the fvs problem may have many potential applications in complex systems research as well . for example , if a vertex is contained in a large fraction of the near - optimal feedback vertex sets , we may expect this vertex to play a very significant role for the dynamical processes on the graph . therefore the probability of belonging to a near - optimal fvs can serve as a centrality index of dynamical significance for each vertex of a graph . such a probablity can be computed by sampling many independent near - optimal fvs solutions or be computed directly using the belief propagation iterative equations @xcite . the construction of a near - optimal fvs also facilitates the study of a complex dynamical system as a simpler response problem of a cycle - free subsystem ( which is intrinsically simple ) under the influence of the vertices in the fvs vertices . if the subgraph induced by the vertices in the fvs itself contains many cycles , such a decomposition can be applied on this subgraph again . through this iterated process , an initial cycle - rich complex graph is then organized into a hierarchy of forest ( cycle - free ) subgraphs and the edges between these forests . a simple illustration of this hierarchical organization is shown in fig . [ fig : fvshierarchy ] . we believe such a hierarchical representation of a complex graph will be very helpful in future dynamical applications . ( color online ) . the four filled points form an optimal feedback vertex set for this graph @xmath0 . the subgraph @xmath5 induced by these four vertices and the cycle - free subgraph induced by all the remaining vertices ( shown as open points ) are connected through many edges ( shown in light blue ) . since the subgraph @xmath5 still contains cycles within itself , we decompose it into a tree subgraph of three vertices ( filled magenta points ) and a subgraph formed by a fvs of one vertex ( the central red point ) . by this way , the vertices in the orginal graph @xmath0 are arranged into three different layers . the vertices of each layer form a cycle - free subgraph ( a tree or a forest ) , while different layers are connected by edges . an important property of such an organization is that each cycle must involves vertices from at least two layers . , scaledwidth=35.0% ] the next section describes the sals algorithm in detail , and in sec . 3 we test the performance of this local search algorithm on random graph instances and compare the results with the results obtained by the bpd algorithm and those obtained by the replica - symmetric mean field theory . we conclude this work in sec . the two appendices are the proofs of the theorems of sec . 2 . for a graph @xmath0 of @xmath4 vertices , let us consider an ordered list @xmath6 formed by @xmath7 vertices of this graph @xmath8 following @xcite , we assign to the first vertex @xmath9 of this list an integer rank @xmath10 , to the second vertex an integer rank @xmath11 , ... , and to the last vertex @xmath12 an integer rank @xmath13 . therefore each vertex @xmath14 has an integer rank @xmath15 which marks the position of this vertex in the list @xmath6 . for the purpose of the fvs problem , we introduce for each vertex @xmath14 a ranking condition as @xmath16 where @xmath17 denotes an edge of the graph @xmath0 between vertex @xmath18 and vertex @xmath19 , and @xmath20 if @xmath21 and @xmath22 if @xmath23 . the ranking condition ( [ eq : rc ] ) is satisfied by vertex @xmath18 if and only if among all the nearest neighboring vertices of vertex @xmath18 that are also contained in the list @xmath6 , at most one of them has a lower rank than that of @xmath18 . a list @xmath6 is referred to as a legal list if all its vertices satisfy the ranking condition ( [ eq : rc ] ) . the link between legal lists and feedback vertex sets is setup by the following two theorems : [ th:1 ] if @xmath6 is a legal list , then the subgraph of the graph @xmath0 induced by all the vertices of this list is cycle - free . therefore the set @xmath24 formed by all the remaining vertices of @xmath0 not included in @xmath6 is a fvs . [ th:2 ] if @xmath24 is a fvs for a graph @xmath0 , then it is possible to form a legal list @xmath6 using all the vertices not contained in @xmath24 . these two theoretms are easy to prove , see the appendices for technical details . they suggest that there is a one - to - many correspondance between a fvs @xmath24 and legal lists @xmath6 . therefore the problem of constructing an optimal fvs is converted to a problem of constructing a legal list @xmath6 of maximal cardinality . notice that judging whether a list @xmath6 is legal or not is algorithmically very easy as eq . ( [ eq : rc ] ) involves only the neighborhood a focal vertex @xmath18 but not the connection pattern of the whole graph @xmath0 . a similar conversion from global constraints to local constraints has also been used in the steiner tree problem , which aims at constructing a tree of minimal total edge length connecting a set of specified vertices @xcite . just for simplicity of later discussions , let us define the energy of a legal list @xmath6 as the total number of vertices not contained in it , namely @xmath25 in otherwords , @xmath26 is just the cardinality of the complementary fvs @xmath24 of the list @xmath6 . following ref . @xcite we implement a simulated annealing local search algorithm as follows : 1 . input the graph @xmath0 . initialize the legal set @xmath6 as containing only a single randomly chosen vertex of @xmath0 , and the complementary feedback vertex set @xmath24 then contains all the other vertices of @xmath0 . set the temperature @xmath27 to an initial value @xmath28 . 2 . choose a vertex ( say @xmath18 ) uniformly at random from the feedback vertex set @xmath24 . + \(a ) if the list @xmath6 contains no nearest neighbor of the vertex @xmath18 , delete @xmath18 from @xmath24 and insert it to the head of @xmath6 . then vertex @xmath18 has rank @xmath29 in the updated list @xmath6 and the energy of @xmath6 decreases by @xmath30 . + \(b ) if @xmath6 contains exactly one nearest neighbor ( say @xmath19 with rank @xmath31 ) of the vertex @xmath18 , delete @xmath18 from @xmath24 and insert it to @xmath6 at the position just after vertex @xmath19 . then vertex @xmath18 has rank @xmath32 in the updated list @xmath6 and the energy of @xmath6 decreases by @xmath30 . + \(c ) if @xmath6 contains two or more nearest neighbors ( say @xmath33 with vertex @xmath19 having the lowest rank @xmath31 among these vertices ) of the vertex @xmath18 , we make a proposal of moving @xmath18 from @xmath24 to the ordered list @xmath6 at the position just after vertex @xmath19 and deleting all those nearest neighbors of @xmath18 from @xmath6 if the insertion of @xmath18 causes the violation of the ranking condition ( [ eq : rc ] ) for these vertices . suppose @xmath34 vertices have to be deleted from @xmath6 ( and be added to @xmath24 ) as a result of inserting vertex @xmath18 to @xmath6 , then the energy increase of @xmath6 is @xmath35 . if @xmath36 we accept the proposed action with probability @xmath30 , otherwise we accept it with probability @xmath37 . if this proposal is accepted , then vertex @xmath18 has rank @xmath32 in the updated list @xmath6 . repeat step 1 until the list @xmath6 has been successfully updated for @xmath38 times . during this process , record the best fvs so far reached and the corresponding lowest energy , @xmath39 . 4 . decrease the temperature to @xmath40 , where @xmath41 is a fixed constant , and then repeat steps 1 - 2 . if the energy value @xmath39 does not change in @xmath42 contiguous temperature stages , then we stop the local search process and output the reached best fvs . in our computer experiments we use @xmath43 , @xmath44 and @xmath45 which are identical to the values used in ref . the temperature ratio @xmath46 is set to several different values @xmath47 , @xmath48 and @xmath49 . as shown in fig . 2 , the slower the rate of temperature decrease , the lower is the energy of the constructed best fvs solutions . the bpd algorithm was applied to erds - renyi ( er ) random graphs and regular ( rr ) random graphs in @xcite to test its performance . as we want to compare the performance of the present sals algorithm with the bpd algorithm , we apply the sals algorithm on the same er and rr random graphs used in @xcite . an er random graph of @xmath4 vertices and @xmath50 edges was generated in @xcite by first selecting @xmath51 different vertex pairs uniformly at random from the whole set of @xmath52 candidate vertex pairs and then connecting the two vertices of each selected vertex pair by an edge . a vertex in such a graph has on average @xmath53 nearest neighbors ( i.e. , the mean vertex degree of the graph is @xmath53 ) . when @xmath54 the vertex degree distribution of the graph converges to a poisson distribution of mean value @xmath53 @xcite . a rr graph differs from an er graph with the additional constraint that each vertex has exactly the same number @xmath53 of nearest neighbors ( here @xmath53 must be an integer ) . it was generated by first attaching to each vertex a number of @xmath53 half - edges and then randomly connecting two half - edges into a full edge , but prohibiting self - loops or multiple edges between the same pair of vertices ( see ref . @xcite for more details of the graph generation process ) . the vertex number of all these random graph instances is equal to the same value @xmath55 . there are @xmath56 independently generated er or rr random graph instances at each fixed ( mean ) degree value @xmath53 . we run the sals process once on each of these @xmath56 instances and then compare the mean value of the obtained final @xmath39 values with the mean value of the final energies obtained in @xcite by running the bpd process once on each of these @xmath56 graph instances . the evolution of the minimal energy @xmath39 as a function of the temperature @xmath27 is shown in fig . [ fig : evolution ] for er ( a ) and rr ( b ) random graphs of @xmath57 . when @xmath58 the evolution curves for the three different cooling parameters @xmath49 , @xmath59 and @xmath60 coincide with each other . this indicates that at @xmath58 the typical relaxation time of energy is shorter than the time scale of temperature decreasing . at @xmath61 the curve of @xmath49 starts to separate from the other two curves , and the final value of @xmath39 is considerably higher than those of the other two evolutionary trajectories . this indicates that at cooling parameter @xmath49 the simulated annealing process is eventually trapped in a local region of the configuration space whose energy minimal value is extensively higher than that of the optimal solutions . the two evolutionary curves of @xmath62 and @xmath47 again separate from each other at a lower temperature of @xmath63 , and the final value of @xmath39 reached at cooling parameter @xmath47 is slightly lower than the final value of @xmath39 reached at @xmath62 . if the value of @xmath46 is set to be more closer to @xmath30 the final value of @xmath39 will decrease slightly further ( at the expense of much longer simulation times ) . the observation that lower final energy values can be reached by lowering the cooling rate strongly indicates the system has a very complicated low - energy landscape with many local minima . this is consistent with the prediction that the undirected fvs problem is in a spin glass phase at low enough energy values @xcite . figure [ fig : fvs ] compares the performance of the sals algorithm ( the cooling parameter fixed to @xmath47 ) with that of the bpd algorithm on er ( a ) and rr ( b ) random graphs . we see that bpd slightly outperforms sals for both ensembles of random graphs this is not surprising , since the bpd algorithm takes into account the global structure of the graph through message - passing , while the sals algorithm considers only the local graph structure . it is interesting to see that the results of the two algorithms are actually very close to each other , especially for er random graphs . as also shown in fig . [ fig : fvs ] , at each value of ( mean ) degree @xmath53 , the result obtained by the sals algorithm is very close to the predicted value of the global minimal energy by the replica - symmetric mean field theory @xcite . therefore message - passing algorithms are not the only candidate choices to construct near - optimal feedback vertex sets for undirected random graphs . an advantage of the sals algorithm is that its implementation is very straightforward . it may be the method of choice for many practical applications . , a vertex s probability of being in a feedback vertex set increases with its vertex degree @xmath64 . each star point is the mean result obtained by averaging over the @xmath56 final fvs solutions obtained by the sals algorithm on the @xmath56 random graph instances of @xmath55 vertices . each plus point denotes the result obtained by the rs mean field theory at @xmath65 @xcite . , scaledwidth=35.0% ] we also notice from fig . [ fig : fvs ] that the sals algorithm performs poorer in regular random graphs than in er random graphs . this is probably due to the following fact : each vertex in a regular random graph has the same degree so there is local guide as to whether a vertex should be included into the feedback vertex set or not . on the other hand the degree heterogeneity of an er graph will give some local guide to the sals process to arrive at a near - optimal fvs . based on the final @xmath56 fvs solutions obtained by the sals algorithm on the @xmath56 er random graphs with mean vertex degree @xmath57 , we compute the mean probability @xmath66 that a vertex of degree @xmath64 is contained in the fvs . the results shown in fig . [ fig : vertexfvsprob ] demonstrate that @xmath66 is close to @xmath67 for @xmath68 and close to @xmath30 for @xmath69 , and it is an rapidly increasing function of @xmath64 in the range of @xmath70 . the empirically obtained mean value of @xmath66 is found to be very close to the value of @xmath66 computed using the replica - symmetric mean field theory . notice that for @xmath71 the empirical values of @xmath66 are slightly larger than those predicted by the rs mean field . such small differences cause the energies of the construsted fvs solutions to be higher than the corresponding optimal values . the graphs for real - world complex systems usually are highly heterogenous in terms of vertex degree distributions @xcite . therefore it is very likely that the sals algorithm to have very good performance for such graphs . in this paper , we implemented and tested a local search algorithm for the undirected feedback vertex set problem . similar to the local search algorithm for the directed fvs problem @xcite , our algorithm uses the technique of simulated annealing to explore the highly complex landscape of low - energy congiurations . our simulation results demonstrated that this algorithm is very efficient for large random graph instances . the relative sizes of the constructed fvs solutions by the local search algorithm are very close to the predicted values of the replica - symmetric mean field theory , and they are also very similar to the results obtained by the bpd message - passing algorithm . it should be emphasized that the microscopic dynamical rules of our local search algorithm do not obey the detailed balance condition . these dynamical rules therefore need to be appropriately modified if one is interested in the equilibrial fvs solutions at a given fixed temperature @xmath27 . we are currently using a modified set of microscopic dynamical rules to study the spin glass phase transition of the model proposed in @xcite . such an equilibrium study will offer more physical insights on the cooling - rate dependent behaviors of fig . [ fig : evolution ] . the numerical simulations were performed at the hpc computer cluster of the authors institute . this work was partially supported by the national basic research program of china ( no . 2013cb932804 ) , the knowledge innovation program of chinese academy of sciences ( no . kjcx2-ew - j02 ) , and the national science foundation of china ( grant nos . 11121403 , 11225526 ) . hjz conceived research ; smq performed research ; hjz wrote the paper . consider a generic legal list @xmath6 formed by some vertices of the graph @xmath0 , see eq . ( [ eq : glist ] ) . let us consider the subgraph @xmath72 induced by all the vertices of list @xmath6 and all the edges between these vertices . since @xmath6 is a legal list , every vertex @xmath14 must satisfy the rank condition ( [ eq : rc ] ) . consequently , every vertex in the subgraph @xmath72 has at most one nearest neighbor with rank lower than that of itself . this subgraph @xmath72 must be cycle - free . we prove this statement by contradiction . assume there is a cycle in the subgraph @xmath72 involving @xmath64 vertices : @xmath73 if the rank of vertex @xmath74 is lower than that of @xmath75 , then due to the fact of vertex @xmath75 having at most one nearest neighbor with lower rank than itself , the rank of @xmath76 must be higher than that of @xmath75 . continuing this analysis along the cycle , we obtain that the rank of vertex @xmath77 must be higher than that of @xmath78 but lower than that of @xmath74 . but this is impossible since the rank of @xmath78 must be higher than that of @xmath74 . similarly , if the rank of vertex @xmath74 is higher than that of @xmath75 , we will arrive at the contradicting results that the rank of @xmath75 is higher than that of @xmath79 while the rank of @xmath79 is higher than that of @xmath74 . because of these contradictions , the assumption of @xmath72 containing a cycle must be false . therefore @xmath72 must be a tree or a forest . then the set @xmath24 formed by the vertices not contained in @xmath6 must be a feedback vertex set . suppose the set @xmath24 is a fvs of a graph @xmath0 . then the subgraph @xmath72 induced by all the vertices not included in @xmath24 must be cycle - free . if @xmath72 has only one connected component ( i.e. , being a tree ) , we can pick a vertex ( say @xmath18 ) of @xmath72 uniformly at random and specify this vertex as the root of the tree subgraph @xmath72 . we can then construct an ordered list @xmath6 using all the vertices of @xmath72 in such a way : first the root vertex @xmath18 , then all the vertices of unit path length to @xmath18 ( in random order ) , followed by all the vertices of path length two to @xmath18 ( again in random order ) , ... , followed by the remaining vertices of the longest path length to @xmath18 ( in random order ) . obviously @xmath6 is a legal list with the ranking condition ( [ eq : rc ] ) satisfied for all its vertices . if @xmath72 is a forest with two or more tree components , we can perform the above - mentioned process for each of its tree components and then contatenate the constructed ordered lists in a random order to form a whole ordered list @xmath6 . this list @xmath6 must also be a legal list .	an undirected graph consists of a set of vertices and a set of undirected edges between vertices . such a graph may contain an abundant number of cycles , then a feedback vertex set ( fvs ) is a set of vertices intersecting with each of these cycles . constructing a fvs of cardinality approaching the global minimum value is a optimization problem in the nondeterministic polynomial - complete complexity class , therefore it might be extremely difficult for some large graph instances . in this paper we develop a simulated annealing local search algorithm for the undirected fvs problem . by defining an order for the vertices outside the fvs , we replace the global cycle constraints by a set of local vertex constraints on this order . under these local constraints the cardinality of the focal fvs is then gradually reduced by the simulated annealing dynamical process . we test this heuristic algorithm on large instances of erdos - renyi random graph and regular random graph , and find that this algorithm is comparable in performance to the belief propagation - guided decimation algorithm .
general relativity is commonly regarded as the correct approach to non - quantum gravitation @xcite . einstein s theory views gravity as a manifestation of the curvature of the @xmath0-d space - time @xcite . several authors have proposed to consider this physical curved @xmath0-d space - time as a membrane embedded in a flat space - time of higher dimension called the host space . this point of view is computationally convenient and is also extremely natural in the context of modern string and brane theory @xcite . the aim of the present article is to complement the existing literature on this topic . our main conclusion is that the embedding approach to gr can be successfully implemented in a large variety of contexts and provides some undeniable computational and conceptual advantages . here follows a summary of our principal results . we first introduce two new classes of embeddings ( modeled after nash s classical free embeddings @xcite ) and explain why these two classes are particularly natural from the physical point of view . although they typically require host spaces of higher dimensions than most embeddings proposed by various authors @xcite , these new classes of embeddings present the important physical advantage of being deformable , and therefore physically more realistic . in particular , given an arbitrary space - time , any embedding of this space - time which belongs to one of the two new classes can be deformed to obtain an embedding for gravitational waves propagating in this space - time . we then give explicit examples of embeddings in both classes for the standard minkovski space - time , the schwarzschild black hole and gravitational waves propagating in flat space - time . we then propose new variational principles which give back einstein s general relativity by viewing the 4-d space - time as a membrane moving in a flat host space . some of the variational principles involve new border terms previously not considered by previous authors . actually , the issue of constructing actions which deliver the equations of standard general relativity in terms of embedding functions has been often addressed in the literature . our work is the first to propose a solution to this long standing problem . we finally show that the embedding point of view permits a particularly simple and physically enlightening treatment of the initial value problem in relativistic gravitation . we denote the physical 4-d space - time by @xmath1 and its lorenztian , possibly curved metric by @xmath2 . space - time indices running from @xmath3 ( @xmath4 ) to @xmath5 will be indicated by greek ( latin ) letters and the metric signature will be @xmath6 . the covariant derivative for tensor fields defined on @xmath1 is , as usual , the derivative operator associated with the levi - civita connection of the metric @xmath2 . we also consider a ` host'-space @xmath7 _ i.e. _ an @xmath8-dimensional lorentzian flat space with metric @xmath9 and choose a system of @xmath8 coordinates @xmath10 in the host - space @xmath11 . to view the physical 4-d space - time as embedded in the host - space is tantamount to saying that an arbitrary point @xmath12 in @xmath1 can be considered as a point of @xmath11 as well . we thus define an embedding by a set of @xmath8 functions @xmath13 , @xmath14 , which represent the @xmath15-coordinates of the space - time point @xmath12 . note that these functions are scalars with respect to coordinate changes on the space - time @xmath1 . let us now choose a system of four coordinates @xmath16 on the physical space - time @xmath1 . the squared line element @xmath17 between two infinitesimal points of @xmath1 reads , with obvious notations : @xmath18 but the same squared line element can also be evaluated by viewing both points as belonging to the host - space ; this leads to @xmath19 or @xmath20 where @xmath21 denotes the partial differentiation of @xmath22 with respect to @xmath23 . this partial derivative actually coincides with the covariant derivative @xmath24 of @xmath22 with respect to @xmath23 because , as noted earlier , the function @xmath22 is a scalar respect to coordinate changes on @xmath1 . equating ( [ eq : ds21 ] ) and ( [ eq : ds22 ] ) delivers the important relation : @xmath25 which is manifestly covariant with respect to coordinate changes on @xmath1 . it is a well known result that a given lorentzian ( or riemannian ) metric manifold can be embedded into a flat host space of higher dimension . constructive and existence theorems in the local @xcite as well as in the global sense give conditions on the minimal dimension of the host space , for closed and open manifolds ( see also @xcite , and the references in the review @xcite ) . the minimal dimension of the host - space needed to embed locally a generical 4-dimensional space - time is @xmath26 . usually less dimensions are needed for vacuum space - times @xcite . it has however been argued heuristically by deser _ _ @xcite that embeddings can not _ a priori _ be used with profit by physicists . this conclusion essentially rests on an intuition gained from studying the so - called trivial embedding of @xmath0-d minkovski space - time into itself , which can not be deformed to accomodate standard gravitational waves . the way out of this possible problem is conceptually extremely simple . it consists in working only with particular embeddings which do admit deformations . this is where the notion of freeness enters the picture . put simply , free , @xmath27-free and spatially free embeddings are three particular classes of embeddings which share the common property of being _ by definition _ deformable to accommodate linear variations of the metric tensor . let us now present the technicalities which motivate the three definitions we are about to give . consider a given embedding of the form ( [ embed1 ] ) and let @xmath28 be an arbitrary perturbation or deformation of this embedding . we assume that the vectors @xmath29 of the host space are linearly independent . note that this condition is necessary for the metric to be invertible , since from eq.([embed1 ] ) , the linear dependence of these vectors would imply directly the existence of a nonzero eigenvector of the metric matrix @xmath30 with zero eigenvalue . varying ( [ embed1 ] ) , we obtain at first order in @xmath31 : @xmath32 the embedding variation is made up of two contributions , one tangent to the 4-d space - time and the other one normal to it ; we thus write @xmath33 equation ( [ eq : deltag ] ) then becomes : @xmath34 as ( [ embed1 ] ) , ( [ variationmetricsplit ] ) is manifestly covariant with respect to coordinate changes on @xmath1 . the vectors @xmath35 of the host space define the _ second fundamental form _ and are normal to the 4-d embedded space - time : @xmath36 ( see appendix [ appendix : covariant ] ) . now , the embedding @xmath37 will be useful physically if it can be deformed to accommodate an arbitrary perturbation of the metric @xmath38 . this means that the physically useful embeddings are those for which equation ( [ variationmetricsplit ] ) can be solved in @xmath39 and @xmath40 for any arbitrarily given @xmath41 . let us now introduce the definitions of free , @xmath27-free and spatially @xmath27-free embeddings . * definition 1 : free embeddings . * an embedding ( [ embed1 ] ) is said to be free if , for any metric perturbation @xmath41 and any choice of tangential variation @xmath42 , the 10 equations ( [ variationmetricsplit ] ) can be solved in the normal variations @xmath43 . this definition originated with nash work on embeddings of riemannian manifold and is now standard . nash actually chose to work with vanishing tangent variations and the normal variations can then be obtained by algebraic methods only ( see @xcite for a modern discussion ) . locally , the dimension of the host space of nash free embeddings must be @xmath44 the material presented in the following sections ( in particular , the various action principles discussed in section [ sec : actionprinciples ] ) makes it natural to introduce two other classes of embeddings , the @xmath27-free and spatially free embeddings . these classes have never been considered by earlier authors and we now give their definitions : * definition 2 : @xmath27-free embeddings . * an embedding ( [ embed1 ] ) is @xmath27-free if , for an arbitrary metric perturbation , the 10 equations ( [ variationmetricsplit ] ) are equivalent to : + ( i ) @xmath45 linearly independent linear combinations of ( [ variationmetricsplit ] ) which can be solved in the normal variations , no matter what the tangential variations are , and + ( ii ) @xmath46 remaining equations that can be solved in the tangential variations , independently of the solution obtained in ( i ) for the normal variations . + notice that @xmath47 because there are @xmath0 independent tangential variations and thus the number of equations in ( ii ) must lie between @xmath3 and @xmath0 . note also that , by definition , any free embedding is automatically a @xmath48-free embedding . the host - space dimension for a @xmath27-free embedding must be @xmath49 in order to accommodate the linearly independent @xmath0-dimensional tangent space and the @xmath27-dimensional subspace of the normal space which solve the equations in ( i ) . since @xmath50 , we have @xmath51 . * definition 3 : spatially free embeddings . * an embedding is spatially free if there exist a coordinate system @xmath52 on the 4-d space - time in which the 6 vector fields @xmath53 are linearly independent . spatially free embeddings are particularly important when working in the so - called @xmath54 formalism . they form a subclass of the @xmath27-free embeddings . indeed , let @xmath55 be a spatially free embedding and let @xmath52 be the coordinate system in which the 6 vector fields @xmath53 are linearly independent . in this coordinate system , the @xmath56 components of ( [ variationmetricsplit ] ) can be solved for @xmath57 . the remaining components of ( [ variationmetricsplit ] ) are the @xmath58 components ; taken together , they constitute a system of four inhomogeneous first order differential equations for the fields @xmath59 that can be solved by integration along @xmath60 thus , any spatially free embedding is at least @xmath61-free , hence @xmath27-free . the three above definitions are relevant for physics if at least some general relativistic space - times admit embeddings which are either free , @xmath27-free or spatially free . we prove in the next section that this is so by constructing explicit examples of free , @xmath27-free and spatially free embeddings for several space - times of physical interest , including a schwarzschild black hole . whether all general relativistic space - times admit embeddings in at least one of these three classes remains however an open problem . we now give explicit examples of embeddings of physically relevant space - times belonging to the classes defined in the last section . they represent thus the first examples in the literature of embeddings whose deformations can , by construction , be properly mapped to metric deformations , so that , for example , gravitational waves can be described as embedding waves . let the space - time coordinates be denoted by @xmath62 , where @xmath63 are ` spatial ' coordinates . we have developed a very simple method to construct @xmath27-free embeddings for an interesting class of space - times , including the flat and schwarzschild @xmath0-dimensional space - times . the method consists in splitting the host space as a direct sum of two flat subspaces , ( i ) the ` base ' host space , with flat euclidean metric @xmath64 , and ( ii ) the ` extra ' host space , with flat lorentzian metric @xmath65 ( one and only one time - like direction ) . we will describe the contribution of each subspace to the total embedding in two steps , followed by the explicit examples . on the base host space , we define a ` base embedding ' @xmath66 , depending only on the spatial coordinates , with the following properties : 1 . the metric induced by @xmath67 is flat : @xmath68 2 . the @xmath69 vectors @xmath70 are linearly independent for all @xmath71 , where @xmath72 is a given @xmath5-dimensional subset of @xmath73 . among the several types of base embeddings that can be constructed we have chosen the following @xmath74-dimensional one , for its simplicity : @xmath75 where the functions @xmath76 and the parameters @xmath77 are chosen so that @xmath67 satisfies properties p1 and p2 . property p1 is ensured if there exist real functions @xmath78 such that : @xmath79 where @xmath80 and prime denotes derivative with respect to @xmath81 . property p2 is ensured if the real functions @xmath78 satisfy : @xmath82 where @xmath83 and @xmath84 , for @xmath85 fixed . a particular solution of the previous inequalities is @xmath86 with this solution , @xmath67 is a base embedding on @xmath87 and this embedding is not self intersecting ( _ i.e. _ for any two sets of 3 spatial coordinates @xmath88 and @xmath27 , @xmath89 it is easy to show that , by adding extra embedding functions to this base embedding , the general embedding @xmath90 satisfies automatically the following properties : 1 . the metric induced by @xmath55 decomposes as : @xmath91 2 . the embedding @xmath92 is spatially free for @xmath71 . appropriate choices of these extra embedding functions @xmath93 generate spatially free embeddings of the flat minkovski space - time and of the schwarzschild black hole . here , to accommodate the extra embedding , only one extra dimension ( necessarily time - like ) is needed . the extra embedding function is @xmath94 . the ( @xmath95-dimensional ) host space metric is @xmath96 , and it follows from eq.([eq : splitmetric ] ) that the induced space - time metric is @xmath97 this @xmath98-dimensional embedding is @xmath61-free for @xmath99 in this case , to accommodate the extra embedding we need @xmath5 extra dimensions , one time - like and two space - like . the metric components @xmath100 of this black hole in kerr - schild coordinates do not depend explicitly on @xmath101 , and read @xcite : @xmath102 where @xmath103 and @xmath104 by looking at eq.([eq : splitmetric ] ) , this black - hole metric is obtained if we use the following extra embedding : @xmath105 each spatial coordinate @xmath106 ranges from @xmath107 to @xmath108 the parameter @xmath109 must satisfy @xmath110 finally , the functions @xmath111 and @xmath112 must solve the differential equations : @xmath113 they are well - behaved for all @xmath114 , even near the horizon ( @xmath115 ) . analytical expressions for the functions @xmath111 and @xmath112 can be obtained when @xmath116 figs . [ fig : bh1 ] and [ fig : bh2 ] have been constructed using these expressions . the extra embedding functions @xmath117 account for the causal structure of the black hole . after rescaling the time @xmath101 , the radial coordinate @xmath118 and the embedding functions in units of @xmath119 , we plot them from eqs.([plot ] ) in figs . [ fig : bh1 ] and [ fig : bh2 ] ( interior and exterior region , respectively ) as two - dimensional surfaces parameterized by @xmath120 these are helicoidal surfaces with a pitch @xmath121 . we remark that the interior and exterior regions are smoothly connected . the black hole metric is equal to the metric induced by the plotted surface plus the ( flat ) metric @xmath122 induced by the base embedding ( not plotted ) . the event horizon @xmath123 , denoted in both figures , is an upgoing helix subtending an angle of @xmath124 with respect to the rescaled axis @xmath125 . the interior region @xmath126 is represented in fig . [ fig : bh1 ] . we can see the event horizon @xmath123 as a light - like helix which is close to the vertical axis . the space - like helix @xmath127 is bounding the figure . the actual surface extends to infinity as @xmath128 subtending asymptotically an angle of @xmath124 with respect to the rescaled axis @xmath125 as @xmath129 all physical trajectories get trapped in the interior region , approaching the ` helix at infinity ' @xmath130 . this graphical interpretation is possible since the plotted surface gives the main contribution to the black hole metric in the region @xmath131 . indeed , the ( flat metric ) contribution coming from the base embedding ( not plotted ) is proportional to @xmath132 and is therefore negligible . the exterior region @xmath133 is represented in fig.[fig : bh2 ] by the spiral surface that approaches the vertical axis as @xmath134 , folding and folding indefinitely , never reaching the axis . notice that in this region , the above mentioned base embedding s contribution ( not plotted ) to the black hole metric , being proportional to @xmath132 , is much more important than the _ spatial _ contribution coming from the plotted surface , so that this exterior region plot gives only partial information about the causal structure of the black hole . let us check that the @xmath61-free embedding of flat space - time introduced in section [ firstexample ] allows for the representation of gravitational waves as embedding waves by computing explicitly the perturbation of this embedding that corresponds to the well - known @xcite plane wave : @xmath135 with @xmath136 , @xmath137 and @xmath138 we solve eq.([variationmetricsplit ] ) for the embedding variations by first taking vanishing tangent variations @xmath139 . we then split the normal variations as the embedding itself : @xmath140 , where @xmath141 ; @xmath142 and @xmath143 are explicitly given by : @xmath144 and the other components are obtained by cyclic permutations of @xmath145 . the functions @xmath146 and @xmath147 were defined in section [ sec : firststep ] . the embedding waves are thus simply plane waves , modulated with some smooth functions related to this particular embedding . we have introduced in the previous sections the ( purely kinematical ) concept of free embedding , which is _ by definition _ deformable to accommodate linear variations of the metric tensor , and thus gravitational waves . our next goal , therefore , is to find a * satisfactory action principle * for general relativity in terms of embedding variables , in order to obtain what we will call * ` free embedding field theory for gravity'. this theory that requires host space dimension @xmath148 , turns out to be equivalent to gr , not only at the levels of eom but also for perturbations ( waves around the general solution ) . we then consider the ` @xmath27-free embedding theory for gravity ' * , that allows us to reduce the minimal dimension of the host space from @xmath149 to @xmath48 . finally , we introduce a third , alternative family of action principles that depend on lagrange multipliers . the metric variations and embedding variations are independent and their relation ( eq.([embed1 ] ) ) is given as an eom . these alternative theories are also shown to be equivalent to usual gr , both at the levels of eom and of perturbations ( waves around the general solution ) , provided that the embedding is @xmath27-free . let us denote the @xmath0-d space - time manifold and its @xmath5-d boundary by @xmath150 and @xmath151 respectively and let @xmath8 be the dimension of the host space . consider the usual @xcite einstein action principle but written in terms of embedding variables @xmath152 = \frac{1}{8\pi g}\int_{\mathcal{m } } r[g[\bm{y}]]\,\sqrt{-\|g[\bm{y}]\|}\,d^4 x + \frac{1}{4\pi g}\int_{\partial \mathcal{m } } k[\bm{y}]\ ] ] where @xmath153 $ ] means that @xmath2 is replaced everywhere by its definition , eq.([embed1 ] ) . the scalar curvature @xmath154 $ ] is given explicitly by @xmath155 note that the scalar curvature is quadratic and of second order only in the embedding functions @xmath55 ( see appendix [ appendix : covariant ] ) . finally , @xmath156 $ ] is the trace of the extrinsic curvature of the boundary . the @xmath157-boundary term is here , as usual , @xcite to cancel boundary terms that depend on the normal derivatives of @xmath158 , and allows to impose the boundary conditions @xmath159 . we thus assume from now on the vanishing boundary conditions ( v.b.c . ) @xmath160 on @xmath161 . these boundary conditions explicitly read , using eq.([variationmetricsplit ] ) , @xmath162 let us consider the variation of action ( [ actionrt ] ) , in two steps . first , we obtain in the standard way @xcite @xmath163 where @xmath164 $ ] is the einstein tensor computed in terms of the embedding functions ( see appendix [ appendix : covariant ] ) and @xmath41 is the variation of the metric given in terms of the variation of the embedding functions in eq.([variationmetricsplit ] ) . next we rewrite this variation of the action explicitly in terms of the variations of the embedding functions . using the twice contracted bianchi identity @xmath165 we get a boundary plus a bulk term : @xmath166 where @xmath167 is the normal unitary 3-volume element on the boundary . the above boundary term comes from the differential dependence of the metric in terms of the embedding functions . note that it has nothing to do with the @xmath156$]-boundary term in eq.([actionrt ] ) . we assume in this section the simplest case which is that the embedding is free throughout the manifold and at its boundary . from local existence theorems , this will require the dimension of the host space to be @xmath148 . by definition of free embedding ( see section [ sssec : def ] ) , the @xmath48 @xmath8-dimensional host - space vectors @xmath168 are linearly independent . in this simple case , we can consistently take @xmath139 throughout the manifold . the boundary conditions on the embedding functions ( [ vbc ] ) now imply , because of the freeness of the embedding , @xmath169 on the boundary . using eq.([var - action - y ] ) , under arbitrary normal embedding variations @xmath40 in the bulk we obtain the following euler - lagrange equations : @xmath170 \,\bm{y } _ { ; \mu \nu}=0\,\ ] ] which , taking into account embedding freeness in the bulk , imply @xmath164 = 0,$ ] which are the standard einstein equations but written in terms of the embedding functions . it is worthwhile at this point to warn the reader against a common misinterpretation . in connection with string - like models with arbitrary @xmath8 , the equations for the minimal hypervolume membrane , @xmath171 are correctly understood as constraints on the second fundamental form @xmath172 . but the analogy of these equations with eqs.([eomrt ] ) , with @xmath173 replacing @xmath174 leads to a widespread error : to also understand eqs.([eomrt ] ) as constraints on the second fundamental form . for free embeddings this is clearly a wrong interpretation since , as we have shown , when a free embedding solves eq.([eomrt ] ) , it follows @xmath175 in a perfectly consistent way . on the other hand , for the minimal hypervolume membrane this interpretation is correct , since if a free embedding solved the corresponding equation , then the contradictory result @xmath176 would follow . the earliest work we know on the variational principle for gravitation in terms of embedding variables is the ` gravitation @xmath177 string ' due to regge and teitelboim @xcite ( rt ) . this approach is well - known _ not _ to be equivalent to general relativity . the reason for this is as follows . by requiring isometric embedding only , the authors set @xmath26 as the dimension of the flat host space . they use essentially the action ( [ actionrt ] ) as a starting point , and the equations of motion are again ( [ eomrt ] ) . however , in this case the equations no longer imply the einstein equations since the embedding can not be free , for the dimension of the host space is @xmath178 indeed , there are only @xmath61 independent eom . the free embedding theories introduced in the previous section are equivalent to gr but require a host space of dimension @xmath148 . the following question therefore naturally arises : ` are there embedding theories equivalent to gr , based on a host space of dimension @xmath179 ? ' in this section we consider the most challenging case , @xmath26 . in other words , is it possible to modify @xmath26 rt theory and make it equivalent to gr ? we now show that the only way to make a @xmath26 rt theory equivalent to gr is to require the embedding to be @xmath61-free together with appropriate boundary conditions on the embedding variations . in order to be equivalent to gr , a theory should have the same number ( ten ) of independent eom . however , as pointed out by regge and teitelboim @xcite , deser _ et al . _ , pavi @xcite , and franke and tapia @xcite , it follows from @xmath180 that the second fundamental form @xmath172 has at most @xmath61 independent components ( instead of ten ) . there are therefore at most @xmath61 independent eom ( [ eomrt ] ) . note that they are _ exactly _ @xmath61 independent equations ( by definition 2 , section [ sssec : def ] ) when the embedding is @xmath61-free . what happened to the other 4 ? the answer comes from the fact that the rt boundary conditions do not allow for an arbitrary variation @xmath41 in the bulk , which is not consistent with the standard derivation of the einstein equations from the action principle ( [ actionrt ] ) ( see eq.([eq : delta action 1 ] ) ) . the possibility of having arbitrary @xmath41 in the bulk depends on the boundary conditions imposed on the embedding functions in the theory . but , as we will see below , the naive rt choice of v.b.c . on the _ variations of embedding functions _ and their normal derivatives imposes constraints on the bulk metric variations , independently of the @xmath61-freeness property . however , an arbitrary variation @xmath181 in the bulk is allowed when the boundary conditions : ( a ) @xmath182 arbitrary and ( b ) @xmath40 given by eqs.([vbc ] ) on the boundary , are used instead of the rt v.b.c . to give a specific example , consider variations around a @xmath26 , @xmath61-free and spatially free embedding of an open neighborhood within the flat space - time @xmath183 . here the relevant boundaries for the action principle are the space - like boundaries @xmath184 . we assume for simplicity the following property for the embedding functions : @xmath185 throughout the open neighborhood . now suppose that both @xmath186 and @xmath187 are zero on the boundaries . following the general equations ( [ variationmetricsplit ] ) there are , in the bulk , 4 differential equations we must solve for the tangent embedding variations : @xmath188 and 6 algebraic equations for the normal embedding variations : @xmath189 the last 6 equations imply @xmath190 on the boundaries , with no condition in the bulk . however , the first 4 equations give constraints : the @xmath191 component implies @xmath192 and the @xmath193 components imply @xmath194 but these expressions are zero for the v.b.c . on the embedding variations , so we obtain @xmath195 @xmath196 + conversely ( see eqs . ( [ variationmetricsplit ] ) ) , the assumption of arbitrary metric variations in the bulk imply that , at space - like portions of the boundary , all the @xmath0 embedding tangent variations @xmath197 must be arbitrary . recalling that the action variation ( [ var - action - y ] ) depends on these tangent variations , we are led to @xmath0 extra euler - lagrange equations on the boundary , that will turn out to be the usual constraints of gr , see eq.([rteulerlagrboundary ] ) below . we now turn to the proof that , in the more general case of curved space - times embedded in a spatially free manner , these extra equations will be enough to show the equivalence of the einstein equations with the euler - lagrange equations obtained from the action ( [ actionrt ] ) under the new boundary conditions : @xmath182 arbitrary and @xmath40 given by eqs.([vbc ] ) , which admit solutions because of spatial freeness . for simplicity , we will assume the following properties for the space - time and the embedding : ( i ) the space - time is globally hyperbolic , which allows to define a global time coordinate , and ( ii ) the embedding is @xmath61-free and , with respect to the latter time coordinate , it is spatially free ( see definition 3 in section [ sssec : def ] ) . we define the resulting theory as a @xmath61-free embedding theory of gravity . using the general equation ( [ var - action - y ] ) for the variation of the action , we consider arbitrary variations @xmath197 on the boundary , and arbitrary variations @xmath40 in the bulk . the euler - lagrange equations for this theory are thus : @xmath198 where @xmath199 is the unit normal to the boundary , and @xmath200 using the appropriate time slicing , which exists in globally hyperbolic space - times , the euler - lagrange equations at the space - like boundary are @xmath201 . on the other hand , spatial freeness of the @xmath48-dimensional embedding implies the relations @xmath202 , where @xmath203 are tensor fields defined by the embedding . therefore the bulk eqs.([rteulerlagrbulk ] ) become @xmath204 using this equation we can rewrite @xmath205 as a system of four first order , homogeneous partial differential equations for the unknowns @xmath206 recalling that the initial condition is @xmath207 we obtain @xmath208 in the bulk , and then eqs.([rteulerlagrbulk-1 ] ) imply @xmath209 in the bulk . we have thus shown the equivalence between gr and @xmath61-free , spatially free embedding theory . the case @xmath210 corresponds to the addition of extra embedding functions to the theory , and all derivations above apply with minimal obvious modifications . we now explicitly demonstrate that the above embedding theory linearized about the solution of the eom @xmath175 is equivalent to standard linearized gr , and thus contains the standard gravitational waves . to wit , we compute the variations of eqs.([rteulerlagrboundary][rteulerlagrbulk ] ) , and evaluate them on the general solution , which satisfies the eom @xmath211 . in the rest of this subsection the covariant derivative corresponds to the metric satisfying the eom . we obtain : @xmath212 the tensor @xmath213 is covariantly conserved when the eom hold . this can be deduced from the variation of the identity @xmath214 . we get @xmath215 and , using the eom @xmath216 @xmath217 follows . then , the tensor @xmath218 satisfies the same equations as the tensor @xmath219 . repeating the arguments already used above we conclude that for spatially free @xmath61-free embeddings the solution is @xmath220 , just as expected for the standard perturbation theory of the einstein equations in gr . however , this time one has to compute the variations with respect to the embedding variables . though it is possible to write down the resulting equations , in our opinion it gives no further insight to do it . instead we notice that the perturbations of the embedding functions _ must _ propagate in a proper way , because of the proved correspondence that exists , for spatially free @xmath61-free embeddings , between the variations of the metric and those of the embedding , eqs.([variationmetricsplit ] ) . we refer the reader to the explicit example of waves around flat space - time in section [ sec : example - waves ] . this case is intermediate between the @xmath148 free ( i.e. , @xmath48-free ) case ( section [ sec : freeembtheory ] ) and the @xmath222 @xmath61-free case ( sections [ sssec:6freen=10 ] , [ sssec:6freen>10 ] ) . in this case the eom ( [ eomrt ] ) are exactly @xmath223 equations ( by definition 2 , section [ sssec : def ] ) . therefore there will be only @xmath224 missing equations . these equations will be obtained as boundary equations like eqs.([rteulerlagrboundary ] ) , by assuming arbitrary boundary conditions on @xmath46 out of the four @xmath225 . following the same type of arguments as those between eqs.([rteulerlagrboundary ] ) and eqs.([rteulerlagrbulk-1 ] ) , we can conclude that if the embedding is spatially free and @xmath27-free with @xmath226 , the eom are equivalent to gr . this section deals with a class of action principles where the metric @xmath30 and the embedding functions @xmath227 are considered independent . relation ( [ embed1 ] ) between metric and embedding therefore appears as an eom . consider the following action @xmath228 & = & \frac{1}{8\pi g}\int_{\mathcal{m } } d^4 x\,\sqrt{-\|g\|}r[g ] + \frac{1}{4\pi g}\int_{\partial \mathcal{m } } k[g]\\ \label{actionn}&+ & \frac{1}{8\pi g}\int_{\mathcal{m } } d^4 x\,\sqrt{-\|g\|}\lambda^{\mu_1 \nu_1 \ldots \mu_n \nu_n}\left(g_{\mu_1 \nu_1}-\bm{y}_{,\mu_1}\cdot\bm{y}_{,\nu_1}\right)\ldots\left(g_{\mu_n \nu_n}-\bm{y}_{,\mu_n}\cdot\bm{y}_{,\nu_n}\right),\end{aligned}\ ] ] where @xmath229 , k[g]$ ] are the ricci scalar and the extrinsic curvature of the boundary @xmath230 , both expressed in terms of the metric , and @xmath231 is a lagrange multiplier . this case has already been presented in the literature @xcite , with v.b.c . on the embedding variables ; these b.c . make the theory inequivalent to gr . let us consider now the action ( [ actionn ] ) in the context of free and @xmath27-free embeddings and allow for arbitrary tangent variations @xmath187 of the embedding . the variation of the @xmath233 reads : @xmath234\right ) \right]\sqrt{-\|g\| } d^4 x \\ \label{var - action-1-degree } & -&2\int_{\partial\mathcal{m } } \delta w_\mu \lambda^{\mu \nu } d \sigma_\nu+2\int_{\mathcal{m } } \left[\delta w_\mu \lambda_{\quad ; \nu}^{\mu \nu } + \bm{\delta y}_\perp \cdot \bm{y } _ { ; \mu \nu}\lambda^{\mu \nu}\right]\sqrt{-\|g\| } d^4 x \,,\end{aligned}\ ] ] where the einstein tensor @xmath235 $ ] is viewed as a functionnal of the metric . remember that @xmath236 and @xmath37 are treated as independent at this stage ; their variations are therefore independent too and eq . ( [ var - action-1-degree ] ) thus delivers three euler - lagrange equations in the bulk . the euler - lagrange equation with respect to the lagrange multiplier @xmath237 gives , as a constraint , the relation between the metric and the embedding functions , eq.([embed1 ] ) ; replacing this equation into the euler - lagrange equation with respect to the metric @xmath30 , we obtain the relation @xmath238 ; replacing finally this last relation into the euler - lagrange equation with respect to the normal embedding variations we get the eom @xmath239 in the bulk , which is formally identical to the equation obtained from the action principle presented in the previous section . for free embeddings this eom implies the einstein equations , @xmath240 however , for @xmath27-free embeddings , equivalence with gr can be obtained by supplementing the @xmath46 missing equations by the boundary equations coming from the arbitrary variation of @xmath182 in eq.([var - action-1-degree ] ) ( see previous section ) . variation of eq.([actionn ] ) now reads @xmath242\right ) \bigg]\sqrt{-\|g\| } d^4 x \\ \nonumber & -&4\int_{\partial\mathcal{m } } \delta w_\mu \widetilde{\lambda}^{\mu \nu } d \sigma_\nu+4\int_{\mathcal{m } } \left[\delta w_\mu \widetilde{\lambda}_{\quad ; \nu}^{\mu \nu } + \bm{\delta y}_\perp \cdot \bm{y } _ { ; \mu \nu}\widetilde{\lambda}^{\mu \nu}\right]\sqrt{-\|g\| } d^4 x \,,\end{aligned}\ ] ] where @xmath243 . the euler - lagrange equations stemming from the first two variations in eq.([var - action-2-degree ] ) are equivalent to : @xmath244 & = & 0\ , , \\ \end{array}\ ] ] which imply einstein dynamics for the embedding variables . equations ( [ solbdb ] ) also imply that both remaining surface and bulk variations in eq.([var - action-2-degree ] ) vanish . at this stage , @xmath245 remains completely arbitrary . it thus seems that , at the level of eom , freeness is not needed to recover einstein equations . however this is misleading because the existence of propagating gravitational waves equivalent to those of gr for this theory needs , as we are now going to prove , that the embedding be @xmath27-free . let us perturb the eom of the @xmath246 embedding theory by varying the embedding variables @xmath247 , the metric @xmath30 and the tensor @xmath248 . after the perturbation has been performed to first order , we can use the eom wherever we want . recall that here @xmath173 depends only on @xmath249 first note that the variation @xmath250 vanishes because of the eom . the other variations give , in the bulk : @xmath251 @xmath252_{;\nu}\\ \label{deltaeulerlagrybulk } & + & \bm{y } _ { ; \mu \nu } { \lambda}^{\mu \nu \alpha \beta } \delta(g_{\alpha\beta}-\bm{y}_{,\alpha}\cdot \bm{y}_{,\beta } ) \,.\end{aligned}\ ] ] on the boundary we get : @xmath253 now , using eq.([deltaeulerlagrg ] ) , eqs.([deltaeulerlagrybulk][deltaeulerlagryboundary ] ) become : @xmath254 the first term in eq.([newdeltaeulerlagrybulk ] ) vanishes ( see the discussion after eq.([deltarteulerlagrbulk ] ) ) . equations ( [ newdeltaeulerlagrybulk][newdeltaeulerlagryboundary ] ) are thus equivalent to eqs . ( [ deltarteulerlagrboundary][deltarteulerlagrbulk ] ) and these are in turn equivalent to einstein equations when the embedding is @xmath27-free . note finally that it is necessary to assume that the tensor @xmath245 is invertible in order to deduce from eq . ( [ deltaeulerlagrg ] ) the embedding equation @xmath255 , which permits the identification @xmath256 of the metric in terms of the embedding at the eom and at the perturbative level as well ; this is what makes possible a description of gravitational waves in terms of embedding waves . in this section we discuss the possibility of treating the einstein equations as a dynamical system in terms of the embedding functions . we present some preliminary results regarding the initial value formulation of the vacuum einstein equations in the embedding approach . note that , in the usual formulation , the numerical integration schemes suffer from instabilities ( e.g. , pure gauge modes and violation of constraints @xcite ) , which destroy their performance in finite time . in this context , the direct numerical integration of the equations in the embedding variables is certainly worth developing . another motivation for future numerical work is that it could provide information on some interesting theoretical questions . for example : is @xmath27-freeness ( or spatially free @xmath27-freeness ) kept as long as one integrates in time the evolution equations in embedding variables ? let us assume for simplicity that the dimension of the host space is @xmath26 . let us assume that the host space has only one time - like coordinate : @xmath257 and that the space - time is globally hyperbolic . one can then introduce a coordinate system @xmath258 with @xmath259 a global time - like coordinate and @xmath260 let the initial embedding , defined by the functions @xmath261 , be @xmath61-free and spatially free with respect to this coordinate system . let @xmath262 this vector is timelike and verifes : @xmath263 where @xmath264 let @xmath265 be the components of @xmath266 in the basis @xmath267 i.e. , @xmath268 . one has @xmath269 and @xmath270 defining now the lapse function @xmath271 and the shift vector @xmath272 by @xmath273 one gets the following evolution equation for @xmath274 : @xmath275 in terms of the functions @xmath276 , the metric components read @xmath277 @xmath278 @xmath279 the condition for @xmath280 to be time - like is @xmath281 our definition of lapse and shift agrees with the standard one ( see wald @xcite ) . it is well known that these functions are not dynamical and that fixing them improperly can cause problems in the numerical integration of einstein equations @xcite . because of coordinate transformation invariance , @xmath0 components of @xmath55 can be chosen as arbitrary functions of @xmath282 . the corresponding @xmath0 equations in ( [ yeom ] ) then fix @xmath283 once @xmath266 is known . in this way , the evolution equations ( [ yeom ] ) allow to propagate in time the embedding functions @xmath55 provided the vector @xmath266 is known . as shown in the next section , the choice of @xmath266 as the relevant variables ` conjugate to @xmath55 ' is the natural one to deal with einstein equations . as usual , the vacuum einstein equations @xmath284 and @xmath285 are @xmath0 constraints on the initial data ; in the present case the initial data are the values of @xmath286 at @xmath287 . the first three equations are linear in the derivatives @xmath288 , while the last one is algebraic and quadratic in @xmath289 @xmath290 here , @xmath291 is the @xmath292inverse of @xmath293 and @xmath294 where the spatial projector @xmath295 is defined by @xmath296 the projector and the metric components appearing in eqs.([einsteinconstraints ] ) are clearly given by the initial data . let us define the ` acceleration ' @xmath297 by @xmath298 since @xmath283 have been fixed by a choice of coordinates , the time evolution of @xmath266 will be completely determined if the acceleration @xmath297 is known . its normal components are obtained ( when the embedding is spatially free ) from the vacuum einstein equations @xmath299 : @xmath300 and its tangent components satisfy : @xmath301 these expressions are obtained from straightforward manipulation of the differentiated versions of eqs.([nconstraints ] ) and eqs.([yeom ] ) . equations ( [ neom][nconstraintsderived ] ) can thus be used to propagate in time the vector @xmath266 . we have revisited the embedding approach on general relativity which views the @xmath0-d , possibly curved , physical space - time as a membrane floating in a flat host space - time of higher dimension . we have first introduced two new classes of embeddings , both based on nash notion of freeness . all embeddings in these classes are deformable and , therefore , allow for a description of gravitational waves ; explicit examples of such embeddings have been given for both minkovski space - time and the schwarzschild black hole . we have also presented new variational principles which deliver general relativity as a field theory for embedding variables . einstein s dynamics thus appears as free membrane dynamics in the host space . we have finally considered the general relativistic initial value formulation in terms of embedding variables and argued that this new point of view sheds new light on this particularly difficult issue . this article proposes what is , to the best of our knowledge , the first consistent embedding approach to non quantum gravitation . previous attempts have been marred by essentially two problems . the first one concerns the possibility of deforming the embedding to accommodate for gravitational waves . our approach is the first to use the notion of freeness introduced by nash , and variations thereof , to solve this problem . the second problem is linked with the possibility of constructing an action principle which would deliver einstein s theory in terms of the embedding variables . previous attempts failed in this respect because the action functionals were used with the wrong boundary conditions , and also because the considered embeddings were not free . we have constructed explicitly free embeddings of several general relativistic space - times of astrophysical importance ( for example , a schwarzschild black hole ) . the approach developed in this article is thus relevant to physics . whether all general relativistic space - times can be embedded freely in a flat space - time of higher dimension remains however an open question . we would like to end this article by mentioning a few assets offered by the embedding point of view on gr which have not already been discussed . first , the embedding point of view is remarkably modern , offering obvious links with string theory , m - theory ( or f - theory ) and cosmology ( see @xcite and references therein ) . for example , embedding variables seem ideally suited to the semiclassical study of the spontaneous creation or destruction of universes out of a quantum vacuum @xcite . more generally , the embedding point of view surely appears as _ the _ right tool to study problems involving changes in the topology of space - time . in fact , in our example of flat space - time , the extra parameters and functions ( appearing explicitly in the embedding but not in the metric ) can be chosen in order to change the topology of space - time from @xmath183 to @xmath302 . in this context , the problem of averaging statistically the geometry of ( classical ) space - time has recently been solved for situations in which the topology of space - time is fixed @xcite . does the embedding point of view permit a more general treatment ? finally , it is certainly interesting to investigate how field quantization on the ( flat ) host space translates on the embedded space - time . the free embedding theory of gravity introduced in this paper , precisely because it deals with @xmath27-free embeddings , is suitable for perturbative quantum theory , as opposed to the old approaches ( see , for example , @xcite ) . to cite just a few interesting questions : does the resulting 4-d quantum theory depend , at fixed space - time coordinates @xmath303 , on the choice of the embedding ? we refer the reader to discussions of this topic in the context of rt theory , and in the context of embedding theory of induced gravity in @xcite . what new insight does the embedding point of view bring to the unruh effect @xcite ? and how does black hole thermodynamics appear in the embedding point of view ? ( see @xcite for an account of these two last topics in the context of gr ) . following dirac @xcite we define the christoffel symbols in terms of the embedding functions : @xmath304 the covariant derivative is defined as usual in terms of the christoffel symbols . for example , the second covariant derivative of a scalar function @xmath305 is @xmath306 it is interesting to see that , when @xmath307 is replaced by the embedding functions ( which are scalars by definition ) we get what is known as the second fundamental form : @xmath308 where @xmath309 is the normal projector , whose kernel is by definition the tangent space @xmath310 generated by the tangent vectors @xmath311 then , the second fundamental form is a set of vectors in the normal space ( @xmath312 ) with @xmath313 finally we compute the riemann and einstein tensors in terms of the embedding functions . the riemann curvature tensor depends on second ( partial ) derivatives of the metric tensor , which itself depends on first derivatives of the embedding functions . therefore one could naively expects the curvature tensor to depend on third order derivatives of the embedding functions . however , all terms containing third order derivatives vanish . the riemann tensor thus depends on second derivatives only and reads @xmath314 the einstein tensor is similarly written @xmath315.$ ] geometrically , it is natural that only second covariant derivatives of embedding functions appear in these tensors : the curvature radii of the embedded manifold along the principal axes depend essentially on the _ normal _ components of the matrix of second derivatives of the embedding functions ( the second fundamental form ) .	the formulation of general relativity in which the @xmath0-dimensional space - time is embedded in a flat host space of higher dimension is reconsidered . new classes of embeddings ( modeled after nash s classical free embeddings ) are introduced . they present the important advantage of being deformable and therefore physically realistic . explicit examples of embeddings whose deformations _ do _ describe gravitational waves around their respective backgrounds are given for several space - times , including the schwarzschild black hole . new variational principles which give back einstein s general relativity are proposed . in this framework , the 4-d space - time is a membrane moving in a flat host space . * pacs numbers:04.20.cv , 04.20.fy , 04.30.nk * * keywords : * general relativity ; embeddings ; variational formulations ; waves .
photon - dominated regions ( or photodissociation regions , pdrs ) are regions of the neutral interstellar medium ( ism ) where far - ultraviolet ( fuv ) ( 6 ev @xmath2 ev ) photons control the heating and chemical processes . they are the interface between regions and cold molecular cores . the physical and chemical structure of pdrs depends critically on the fuv intensity and the gas density . in the peripheral regions of the molecular cloud l1204/s140 , the fuv intensity and the gas density are low , allowing tests of the models in an important regime . we are motivated by two primary goals : understanding pdrs in a regime of parameter space that has not been extensively studied ; and understanding the role of the regions with relatively low fuv and density in global star formation , both in our own galaxy and in other galaxies . a great deal of observational and theoretical effort has been devoted to understanding pdrs . comprehensive models of pdrs have been constructed by several groups ( e.g. , black & dalgarno 1977 , van dishoeck & black 1986 , 1988 , tielens & hollenbach 1985a , sternberg & dalgarno 1989 , le bourlot et al . 1993 , kaufman et al . 1999 ) by solving the full chemistry and heating - cooling balance in a self - consistent way . in pdrs with @xmath3 , the most important heating process for gas is photoelectric heating electrons ejected from dust particles by fuv photons heat the gas through collisions . for dust , direct absorption of fuv photons is the primary heating mechanism . far - infrared continuum emission is the major cooling process for dust , and the lines of @xmath4p@xmath5p@xmath6 at 157.7409 ( hereafter ) and @xmath7p@xmath8p@xmath9 at 63.183705 ( hereafter ) are the most important for gas . therefore , the and lines , along with rotational emission , are the most important tracers of pdrs . most previous work has focused on bright pdrs very close to hot ob stars , e.g. , the orion bar ( tielens & hollenbach 1985b ; jansen et al . 1995 ; hogerheijde et al . 1995 ; tauber et al . 1994 ) , the ngc 2023 pdr ( steiman - cameron et al . 1997 , draine & bertoldi 1996 ) , and the s140 pdr ( emery et al . 1996 ; timmermann et al . 1996 ; spaans & van dishoeck 1997 ) . other recent studies include those by liseau et al . ( 1999 ) , and the field has been reviewed by draine & bertoldi ( 1999 ) and by hollenbach & tielens ( 1999 ) . these regions have a intensity of @xmath10 and a density higher than 10@xmath11 , where is the enhancement factor relative to the standard interstellar radiation field as given by habing ( 1968 ) . there has been very little exploration of the physics of pdrs with modest fields and densities , conditions likely to prevail over most of the surface of molecular clouds in our galaxy . federman et al . ( 1995 ) and van dishoeck & black ( 1988 ) have studied pdrs in diffuse clouds ( @xmath12 mag ) and translucent clouds ( @xmath13 , @xmath14 and @xmath15 mag ) . regions with high densities and moderate uv fields ( @xmath16 , @xmath17 ) have also been studied in some detail ( jansen et al . 1995 ) , and kemper et al . ( 1999 ) have used and observations to probe a reflection nebula with @xmath18 and @xmath19 . in this paper , we explore the critical intermediate regime where @xmath20 and @xmath21 . the infrared space observatory ( iso ) provided a unique opportunity to observe low - brightness extended , , and . we used this capability to study the intermediate regime . it is also important to understand the role of regions with modest extinction and column density in star formation . regions like those we are studying include most of the mass in the interstellar medium ( hollenbach & tielens 1995 ) , but their role in star formation is poorly known . based on the jeans criterion , most molecular clouds in the galaxy are not sufficiently supported by thermal energy and therefore should collapse under gravity to form stars . such widespread collapse , however , would lead to a galactic star formation rate hundreds of times higher than what has been observed ( zuckerman & palmer 1974 , evans 1991 ) . the observed low star formation rate seems to indicate that most parts of most molecular clouds are `` sterile '' ( evans 1999 ) . magnetic fields and turbulence are generally considered to play an important role in supporting molecular clouds and preventing or slowing collapse . however , there is no widely accepted theory on the mechanism of magnetic and turbulent support of molecular clouds . recently , elmegreen ( 2000 ) has argued that star formation does in fact happen within a few cloud crossing times , removing the need for cloud support . pringle , allen , & lubow ( 2001 ) have pointed out that such a picture strains methods of cloud formation , and they conclude that visible clouds would have to form out of dark " molecular matter . these scenarios beg the question : what prevents star formation in the great majority ( @xmath22% ) of molecular gas ? mckee ( 1989 ; bertoldi & mckee 1996 , 1997 ) proposed a mechanism of photoionization - regulated star formation to explain the low star formation rate in the galaxy by combining two ideas : first , that magnetic fields support molecular clouds ; and second , that pdrs occupy a large fraction of molecular gas . the timescale for ambipolar diffusion is proportional to the ionization fraction ( @xmath23 ) and the fuv photons and cosmic rays are the sources of ionization . most molecular gas is sterile because it resides in a pdr , where the ionization is high enough to slow ambipolar diffusion . only highly shielded regions are likely to form stars . in addition , newly formed stars inject energy into the clouds , replenishing turbulence and slowing star formation in the rest of the cloud . in this picture , molecular clouds reach dynamic equilibrium when @xmath24 mag . by focusing on a peripheral region , we test the conditions in these regions , which should not be forming stars according to the theory . we have chosen a 1-d positional cut we call it the s140 cut in the peripheral region of the molecular cloud l1204 ( figure [ s140-iso ] ) for our pdr and chemistry study . s140 , at distance of @xmath25 pc ( crampton & fisher 1974 ) , is an region associated with the molecular cloud l1204 . in the rest of the paper , we will use s140 to refer to the whole cloud . the b0v star hd 211880 illuminates this cloud from the southwest side to create the visible region , an ionization front , and a bright pdr that has been studied extensively . for a picture that shows the co emission in relation to the ionization front and exciting star , see fig . 1 of zhou et al . northeast of the bright pdr lies the most prominent dense core ( hereafter referred to as the s140 dense core ) associated with the cloud . numerous studies have been carried out on this cloud , e.g. , blair et al . ( 1978 ) , tafalla , bachiller , & martin - pntado ( 1993 ) , plume , jaffe , & keene ( 1994 ) , emery et al . ( 1996 ) , and timmermann et al . ( the fuv radiation at the ionization front provided by hd 211880 is about 150 times habing s mean interstellar radiation field ( keene et al . 1985 ; habing 1968 ) . the s140 cut , specified in table [ isopos ] , is far enough ( @xmath26 ) from hd211880 to have a low fuv radiation field ( @xmath27 , [ sec4:fuv ] , [ sec4:kaufman ] ) and modest visual extinction ( @xmath28 , [ sec4:density ] ) . we have observed the s140 cut with the caltech submillimeter observatory ( cso ) and iso . from the cso , we collected data on transitions of and other co isotopes , the neutral carbon [ c i ] @xmath7p@xmath29p@xmath30 transition ( hereafter ) , and the line ( see details in [ sec4:cso ] ) . we use these lines to determine column density of the inner part of the cloud and to constrain the density . we used iso to observe the and lines along the cut ( see details in [ sec4:iso ] ) , which will help us to determine the and density of the cloud , as well as the incident fuv radiation intensity on the clouds . finally , we observed rotational transitions of with iso at selected positions ( [ sec4:iso ] ) . different off - cloud positions have been used for the iso and cso observations , based on the iras 100 map and the co emission ( table [ isopos ] ) . the iso off position had to be chosen farther away from the s140 cut than the cso off position , because of the greater extent of the continuum . the cso off position was used in position - switched observations , but the iso off position was observed in the same way as the on - cloud positions but with longer integrations , in order to measure the and lines from the general background and establish appropriate baseline intensities . various molecular lines and the line were observed using the cso , with parameters listed in table [ cso - obs ] . most of the data were taken with a re - imaging device installed on the telescope so that the effective aperture was 12 m , depending on the frequency and the run ( plume & jaffe 1995 ) . the beam size with the re - imager installed , listed in table [ cso - obs ] for each run , was 23and is referred to as the `` big beam '' . beam sizes of the big beam in table [ cso - obs ] are accurate to better than 10% . since our object is extended , the exact beam size is not critical . the line was observed with the full dish ( when the re - imager was unavailable ) , in the mode of `` on - the - fly mapping '' ; then the data were smoothed to 120 to match the big beam . the pointing and the main - beam efficiency were obtained by observing the available planets and the moon . for the full dish observations , the pointing uncertainty was less than 5@xmath31 ; with the re - imager , it was no more than 30@xmath31 . the cso facility sis receivers were used for all observations . the backend was a 1024-channel acoustic - optical spectrometer , with a bandwidth of 50 mhz . the typical single - sideband system temperature was 300 k for the 230 ghz band , 1300 k for the 345 ghz band , and 4000 k for the 490 ghz band . observations were done with position switching between the on- and the off - source positions . the data are presented on the t@xmath32 scale . the rms noise is not uniform for the cso data , because of different integration times , weather conditions , and instrument performance . at some positions , for some less abundant species , data were not taken because of extremely weak emission predicted from the more abundant isotopes and the usual abundance ratios . repeated observations showed that uncertainties of the results are less than 30% . parameters for the iso observations are in table [ iso - lines ] . information about iso can be found in kessler et al . we used the long - wavelength spectrometer ( lws ; clegg et al . 1996 ; swinyard et al . 1996 ) to observe the and lines in the lws02 mode at 16 positions along the s140 cut . for each line at every position , multiple scans were obtained ; each scan had 19 samples covering a width of 5 spectral resolution elements . each sample had an integration time of 0.4 sec . particle hits caused a number of bad scans , resulting in a significant reduction of usable data . three selected positions from the s140 cut were observed in the and lines with longer integrations . the iso off position of s140 was also observed in the and the lines . the data presented here have been processed through the off - line processor version 8.7 ( olp 8.7 ) , and the line fluxes were extracted by using the iso spectral analysis package version 1.6a ( isap 1.6a ) , calculated under the assumption that the sources were point sources . processing consisted of removing bad scans and obvious glitches and , in a few cases , dc offsetting the continuum to be closer to that of the other scans . the points were then averaged together using @xmath33 clipping and fitted with a line and baseline fit . the quoted errors are those given by the fit . because the standard calibration assumes a point source and , in the region we mapped , the line emission is more extended than the beam , we multiplied these fits by 0.84 for the 63 um line and 0.59 for the 157 um line , using table 4.10 of gry et al . we then converted to surface brightness , using beam solid angles listed in table 4.14 of gry et al . the relevant detectors are sw3 ( ) and lw4 ( ) , resulting in beam solid angles of 1.30 for and 6.65 for . the solid angle is smaller than the nominal beamsize of 16 but considerably larger than the diffraction limit . the beamsize is slightly larger than the diffraction limit at 158 @xmath34 m . the point source fluxes were multiplied by the following factors to get surface brightness in : 6.46 ; 8.87 . in addition to our own cut through the cloud , we have reprocessed the lws data taken toward the bright pdr near the far - infrared peak in s140 itself , originally taken for the program of emery et al . ( 1996 , tdt 09101820 in the iso data archive ) and for another program ( tdt s 82301120 , 122 , and 123 ) using the same corrections as for the data from our own program . because the lws has ten detectors , we also obtained undersampled continuum scans covering from about 43 to 175 from the detectors that were not on the lines . however , because of dark current uncertainty for the bands shorter than 100 and the fringing problem for the bands longer than 100 ( swinyard et al . 1996 ) , we could not make accurate measurements of the continuum from the iso scans . therefore in this work , we will use the iras measurements to constrain the emission . the low - lying pure - rotational h@xmath35 @xmath36 s(0 ) line at 28.218 @xmath34 m and the @xmath37 s(1 ) line at 17.035 @xmath34 m were observed with the iso - sws in the aot02 mode ( de graauw et al . typical integration times were @xmath38100 s per line , in which the 12 detectors were scanned once over the 28.0528.40 and 16.9617.11 @xmath34 m ranges around the lines . the @xmath39 s(3 ) 9.66 @xmath34 m and @xmath40 s(5 ) 6.91 @xmath34 m lines were measured in parallel with the s(0 ) and s(1 ) lines , respectively . the spectral resolving power for point sources is 2000 ( 150 km s@xmath41 ) at 28 @xmath34 m , 2400 ( 125 km s@xmath41 ) at 17 @xmath34 m , 2280 ( 130 km s@xmath41 ) at 9.7 @xmath34 m and 1550 ( 195 km s@xmath41 ) at 6.9 @xmath34 m . the sws apertures are @xmath42 at s(0 ) , @xmath43 at s(1 ) , and @xmath44 at the s(3 ) and s(5 ) lines . since the sws beam is considerably smaller than the lws beam , three positions separated by 30@xmath45 in ra were observed for each of the three lws positions and averaged to produce an intensity for comparison to the lws data . the h@xmath9 lines from translucent clouds are predicted to be weak and close to the sensitivity limit of the sws instrument . at this level , noise induced by charged - particle impacts on the detectors plays a large role . by good fortune , these particle hits were exceptionally low during the orbit in which the h@xmath9 data for s140 were taken so that the quality of the data ( @xmath380.2 jy rms at 28 @xmath34 m ) is comparable to that commonly obtained in much longer integration times . the h@xmath9 s(0 ) and s(1 ) lines were detected with the standard pipeline reduction , but special software designed to handle weak signals in combination with the standard interactive analysis package was used for improvements in the signal - to - noise ratio . this technique also yields a possible detection of the s(3 ) line at one position . the details and justification of the methods used in the software are described by valentijn & thi ( 2000 ) and thi et al . because the mid - infrared continuum emission is weak , @xmath46 3 jy , the data do not suffer from fringing effects caused by an inadequate responsivity function correction . however , the uncertainties are dominated by the removal of the fluctuating dark current ( see leech et al . 2001 ) , leading us to adopt a calibration uncertainty of @xmath47% , which is propagated in the analysis . observational results from the cso and iso are presented in tables [ s140cso - res][s140sws - res ] . spectra of , , , and the line at two selected positions are shown in figure [ csolines - sp ] . figure [ isolines - sp ] shows the spectra of the and the lines , averaged over the whole s140 cut , along with the s(0 ) and s(1 ) lines averaged over the three lws positions where sws observations were made . two velocity components of @xmath48 and @xmath49 are clearly seen in the cso data along the s140 cut . in table [ s140cso - res ] , results for each component are given when applicable . the line widths ( fwhm ) of the transition of both components are around 1.3 , implying a doppler parameter , @xmath50 , where @xmath51 is the @xmath52 half - width of the line . the first component is very close to the velocity observed at the s140 dense core . therefore the molecular gas with this velocity is likely to be the cloud extension of the s140 dense core . the second component , strongly peaking around ( 774 , 0 ) ( table [ s140cso - res ] ) for the line and the lines of and other co isotopes , is the same as that of core f identified in cs@xmath53 and nh@xmath54 ( 1,1 ) and ( 2,2 ) by tafalla et al . this dense core , apparently active in star formation with an associated iras source and outflow , is centered at ( 954 , @xmath55 ) ( relative to the reference point of this work in table [ isopos ] ) and extends about @xmath56 in nh@xmath54 ( 1,1 ) . we distinguish the emission in this component from the cloud by referring to it as core f ; the first component is referred to as the cloud . we define the mean emission in core f as the average of the 5 positions with offsets in @xmath57 between 576 and 972 , using only the @xmath58 component . these positions include all with obvious effects from core f. the mean emission from the cloud is the mean of emission from all positions in the @xmath59 component added to the mean emission from the @xmath58 component outside that range of @xmath57 . we add the two components for comparison to the iso observations , which are not spectrally resolved . table [ s140-hcop ] presents results of the line at selected positions . at ( @xmath60 , 0 ) , we detected a 0.2 k line of with the re - imager . following this detection , the full dish was used to map an area of @xmath61 centered at ( @xmath60 , 0 ) . the small - beam map revealed that the emission is clumpy . the strongest clump is close to the peak of nh@xmath54 ( 1,1 ) detected by tafalla et al.(1993 ) ( see figure [ s140-iso ] ) . in table [ s140-hcop ] we list the small - beam results at the peak position . big - beam observations of the line at other positions did not result in any detection up to the standard deviations given in table [ s140-hcop ] . measurements of the and lines are given in table [ s140lws - res ] , as is the ratio , / , and the iras continuum intensities at 60 and 100 . line measurements from the combined data set of all 16 positions along the s140 cut ( avg(16-pos ) ) , at the s140 pdr peak position , and at the iso off position are also given in the table . the 1@xmath62 uncertainties come from the gaussian fits performed by isap . these uncertainties are appropriate in evaluating the reliability of spatial variations in line strengths and line ratios . in comparisons of the absolute values of these quantities to theoretical models , the overall calibration uncertainty provides a more useful assessment of the significance of the results . the overall uncertainty results from a combination of model uncertainties for the primary calibrator ( uranus ) , variations in the detector dark current , and uncertainties in the value of the system solid angle at each wavelength ( swinyard et al . 1998 ) . considering the various correction factors and their uncertainties , we adopt a value of 30% for this overall uncertainty . figure [ s140cut - lines ] shows the positional variations along the s140 cut of the and lines ( of two velocity components separately ) , the and lines , and the intensity of the iras 100 continuum . the dotted lines for the panels of the and lines mark the line measurements of all the data combined of the 16 positions ( avg(16-pos ) in table [ s140lws - res ] ) . clearly , the and lines and the 100 continuum are quite flat over most of the s140 cut , while the line displays more than a factor of 10 variation in both velocity components , and the varies by a factor of 4 . the , , and show strong spatial variation similar to the line . in contrast , the average and fluxes of the 16 positions ( avg(16-pos ) ) agree with the measurements at each individual position within 40% . thus , core f is simply not apparent at all in the , , or emission . therefore , we will use the and data averaged over all 16 positions as representative of the cloud . core f is also not noticeable in the more limited data ( table [ s140sws - res ] ) . the average and intensities are @xmath63 and @xmath64 erg s@xmath41 @xmath65 sr@xmath41 respectively , and the average line ratio / is @xmath66 . if we subtract the emission at the iso - off position from the mean values , the intensities for and are 6.97 and 0.89 , with a ratio of / of 0.13 . these numbers are very different from the s140 bright pdr , where the ratio of / is @xmath67 , based on our reduction of the archival data , using an analysis like that used for our own data . for the peripheral regions , , rather than , is the primary coolant . the intensities of the lines are roughly constant at @xmath68 erg s@xmath41 @xmath65 sr@xmath41 at the three positions ( table [ s140sws - res ] ) . in addition , the excitation temperature between @xmath69 and @xmath70 is also constant within uncertainties at @xmath71 k. the constancy of intensity and temperature is rather remarkable because one position is toward core f and the other two are positions near the ends of the cut , without emission from lines tracing molecular gas of moderate density ( cf . table [ s140cso - res ] ) . clearly , the emission has the same pattern of constancy as the and lines and traces the gas in the pdr . the s(3 ) line is only tentatively detected at one position , with an excitation temperature between @xmath72 and @xmath73 of @xmath74 k. the non - detections at the other positions give a comparable upper limit on the temperature . in this section , we will attempt to reproduce the observations by using both published models and our own pdr models . the most important free parameters in the modeling are the fuv radiation intensity , the total extinction , and the volume density . therefore , we will first constrain the fuv intensity ( [ sec4:fuv ] ) and the extinction and density ( [ sec4:density ] ) . then we will compare the predicted and line intensities with the iso observations in [ sec4:kaufman ] , [ hst ] . because dust grains absorb most of the fuv photons incident on the cloud and most of that energy is re - emitted in the form of continuum , the intensity of the fuv radiation can be inferred from the continuum . hollenbach et al . ( 1991 ) presented model predictions of the continuum intensities at 100 and 60 for a range of incident fuv intensity . from table [ s140lws - res ] and figure [ s140cut - lines ] , the iras 100 and 60 intensities are quite uniform along most of the s140 cut except at positions 1516 for 60 and the positions 1316 for 100 , where the intensities are lower . therefore we use the intensities averaged over positions 1 - 13 and the intensity at position 16 to estimate a range of values for the ultraviolet intensity . the continuum intensities at the iso off position have been subtracted to yield @xmath75 mjy sr@xmath41 and @xmath76 mjy sr@xmath41 . assuming a @xmath77 dust emissivity , the total intensity is about @xmath78 ergs s@xmath41 @xmath65 sr@xmath41 . these intensities and a @xmath77 dust emissivity yield a dust temperature of @xmath79 k , near the constant value of 27 k reached for @xmath80 , when effects of transient heating of small grains / pahs by single photons are included ( hollenbach & tielens 1995 ) . therefore , the 60 continuum intensity in this region is probably strongly affected by small grains . based on fig . 18 of hollenbach et al . ( 1991 ) , we used the 100 intensity to estimate the strength of the ultraviolet field as @xmath81 . as noted by kaufman et al . ( 1999 ) , these values assume that the fuv impinges only on one surface of the cloud . if the cloud is heated from both sides , then the values of appropriate to one surface would be 20 30 . the assumption here is that the cloud is optically thin to the continuum . given the uncertainties in the fraction of the grain heating caused by fuv photons , we estimate in the range , 1560 . the enhancement factor of the radiation field @xmath82 with respect to the standard field given by draine ( 1978 ) is 1.71 times smaller than , @xmath83 ( draine & bertoldi 1996 ) . the mean integrated intensity of for the cloud component is 0.54 k . from table [ s140cso - res ] , the typical excitation temperature of is about 10 k. assuming that the line is optically thin and in lte , we calculate analytically the total column density of to be @xmath84 @xmath65 . the empirical relations of frerking , langer , & wilson ( 1982 ) on @xmath85 oph yield a visual extinction , @xmath86 mag , and a total column density of h nuclei of @xmath87 @xmath65 . assuming the cloud is spherical and the angular size of the cloud along the s140 cut is roughly 15 , the linear size of the depth of the cloud would be @xmath88 cm ( 4 pc ) at a distance of 910 pc . therefore , the average density along the s140 cut would be 800 . the assumption of lte is not valid at such a low density . models of excitation , including trapping with lvg codes , indicate that the populations at such low densities are far from lte . a more self - consistent solution is a density around 2000 and a column density of of 2.0 , yielding @xmath89 . the ratio of @xmath90 to @xmath91 lines of is roughly consistent with a density of 2000 , and the non - detection of @xmath92 outside core f limits the density to @xmath93 for typical abundances in translucent clouds and @xmath94 k. similar considerations applied to the position of core f ( 774,0 ) lead to estimates of @xmath95 mag and , using a size of 3 , a mean density of 4 . considering the uncertainties in these estimates , the likely uncertainties in column density and density are factors of 2 and 3 , respectively . a direct comparison between pdr models and our data will be given in [ hst ] , which supports these simple estimates . we first compare our results to the published grid of pdr models by kaufman et al . these calculations incorporated new collision rates for fine - structure lines and , new pah heating and chemistry , and lower gas - phase abundances for oxygen and carbon ( savage & sembach 1996 ) . they include regions of low fuv and low density , which seem to be appropriate for our situation . the calculations employed a turbulent broadening of 1.5 , about twice what we infer , but kaufman et al . argue that the results are not too sensitive to this parameter . we first consider the constraints from the two atomic species , and . from fig . 4 of kaufman et al . , the mean ratio of / of @xmath96 constrains @xmath97 . this is at the low end of the range inferred in [ sec4:fuv ] , even if we assume the cloud is heated from both sides . at this , the only solution for density is @xmath98 , lower than the values inferred in [ sec4:density ] . for lower , the solutions for @xmath99 bifurcate into lower and higher values . for example , at @xmath100 , there are solutions for log @xmath101 or 3.3 . the latter are more consistent with the lvg modeling of . the emission by itself ( fig . 3 of kaufman et al . ) , is less diagnostic , but roughly consistent with the ranges implied by the ratio . are these low values of consistent with the emission ? they are substantially below the estimates from [ sec4:fuv ] . our ratio of intensities of the sum of the and lines to the total intensity , 1.7 , is close to the highest values found by kaufman et al . in fact , the ratio of line to continuum of the cloud in s140 may exceed the maximum in their models , if indeed the cloud is heated from both sides . what about the data ? the mean for the cloud , summing the two components as for the data , and correcting for efficiency , is 1.3 . comparing to fig . 7 of kaufman et al . , this value favors very low . to get up to even 15 requires @xmath102 . for the conditions that match the other lines , the prediction of kaufman et al . is too high by a factor of 34 . given reasonable uncertainties , the mean cloud data require quite low values of and @xmath99 , with most likely values of @xmath0 and @xmath103 . these are both lower than our initial guesses , even if the cloud is heated from both sides . for this range of conditions , the surface temperature " of the pdr , plotted in fig . 1 of kaufman et al . , will lie between 100 and 200 k. this is the maximum temperature that should apply to the region of emission , if collisions dominate the excitation . the excitation temperatures of ( table [ s140sws - res ] ) are consistent with this range of temperatures , but the absolute intensities are difficult to reproduce ( [ sec4:h2interp ] ) . comparison of the kaufman et al . results with those of other models in the literature ( e.g. , le bourlot et al . 1993 , roueff , private communication ) and our own models ( e.g. , jansen et al . 1995 , spaans & van dishoeck 1997 , spaans , private communication ) , shows good agreement for the intensities but variations of factors of two in the intensities depending on the temperature structure , radiative transfer treatment and geometry of the source . thus , the range in inferred and @xmath99 could be somewhat larger , but the conclusion that both are low holds firm . the region of the core f emission is clearly denser than that of the cloud , yet core f is not apparent in either lines or continuum . the emission from the line may be quite independent of density up to about @xmath104 , in the allowed range of ; however , the line should increase with density . from fig . 4 of kaufman et al . ( 1999 ) , the ratio of / should increase to about 1 for @xmath105 and in the range that fits our data . to keep the ratio below 0.3 would require @xmath106 . that is , core f would have to be shielded from the ambient fuv field . interestingly , the emission does see core f in the @xmath58 line . either that component is not completely shielded or the enhanced is caused by some internal source that does not affect the line and continuum emission . to check the effects of radiative transport , different doppler broadening , and metal abundance at these low values of and @xmath99 , we ran our own models for the part of parameter space indicated in the last section . we used our own pdr code , an updated version of the code described by jansen et al . ( 1995 ) to calculate abundances of relevant species as a function of depth into the cloud . we then fed the results into a monte carlo code ( hogerheijde & van der tak 2000 ) to calculate excitation and radiative transport . einstein a values of and were taken from tielens & hollenbach ( 1985a ) and galavis et al . ( 1997 ) , respectively . in going from the pdr code to a spherical monte carlo code , one doubles the column density , so the pdr models were run for @xmath107 , yielding a total extinction through the cloud of 18 mag . we ran models for doppler parameters from 0.5 to 2.0 . the integrated intensities were generally insensitive to this parameter . two sets of values for metal abundances were used : normal abundances are those listed by jansen et al . ( 1995 ) in their table 2 for the @xmath108 oph diffuse cloud ; low metals " corresponds to decreasing all abundances other than c , n , and o by a factor of 10 . the most important species is s , which is thought to be quite depleted in molecular clouds compared to translucent clouds . the results showed that the radiative transport of is indeed quite sensitive to the input parameters because it is very subthermally excited at these low densities , but quite opaque ( @xmath109 for models that fit the data well ) . at very low values of and @xmath99 , the line predicted by the monte carlo cloud can be considerably stronger than that predicted by the pdr code alone , after accounting for the doubling of the column density . the differences seem to be partially in the handling of the radiative transport and partly in the geometry . calculations that calculate the temperature self consistently in spherical geometry produce even larger effects ( m. spaans , personal communication ) . among a grid of models , the best fit was obtained for a model with @xmath110 and @xmath111 and low metals . this model reproduced the line strength , indicating that the extinction estimate from [ sec4:density ] was about right . the ratio of / was about 50% lower than the observations in this model , while models with @xmath112 gave ratios higher than observed and lines about a factor of 2 stronger than observed . the line produced by this model is about 1.5 times our rms noise , still consistent with the observations , while the model with @xmath112 would produce a 10 @xmath62 detection . the abundance and thus the line are considerably enhanced in the case of low metals . taken together , these results suggest that @xmath111 is about right , but perhaps a bit low . the best model gave a line in good agreement with observations and an line about 40% too strong . higher values of or @xmath99 greatly overproduced the line . these are in rough agreement with kaufman et al . , but the line predicted by our models is about a factor of 4 weaker than the line predicted by kaufman et al . , and our prediction agrees with the observations . kaufman et al . note that their lines are stronger than in previous calculations because of inclusion of pah chemistry in their models . while charge transfer from pahs to to create is included in our models , the details may be different . we use the results of the model with @xmath110 and @xmath111 in the next section . figure 5 shows the distribution of electron fraction ( @xmath113 , with @xmath114 ) with extinction from the best fitting models from [ hst ] . the model with low metals fits the data better and has slightly lower . the timescale for ambipolar diffusion is : @xmath115 where is the electron abundance ( mckee et al . since the lifetime of molecular clouds is roughly 30 myrs ( bash , green , & peters 1977 ) , if @xmath116 exceeds this time , the region is effectively sterile as long as it remains magnetically sub - critical : it will be dissipated before it can form stars . by setting to 30 myr , we find a threshold ionization fraction for star formation of @xmath117 . in figure 5 , the threshold @xmath117 is reached at @xmath118 mag for the low metal model and is not reached by the end of the high metal model at @xmath107 mag . a cloud with a column density less than 610 mag will have little well - shielded ( @xmath119 ) gas to form stars . thus the ionization fraction in the region we are studying suggests that little star formation would be expected in the extended cloud , if photoionization - regulated star formation applies ( mckee 1989 ) . based on the above analysis , the localized core f along the s140 cut , with @xmath120 mag , has substantial gas that is sufficiently well shielded ( @xmath119 ) to allow star formation , which has been confirmed by the outflow around this core ( tafalla et al . this is consistent with mckee s theory of photoionization - regulated star formation . in comparison , most of the molecular gas in the extended cloud , with @xmath121 mag , has a @xmath122 and is unlikely to form stars , if subjected to fuv radiation from both sides of the cloud . infrared studies to look for evidence of star formation in the peripheral region of s140 ( away from core f ) would be able to test this prediction and mckee s theory . since the average extinction of molecular clouds in the galaxy is about 7.5 mag ( larson 1981 ) , according to the photoionization - regulated star formation model , most of the molecular gas will not form stars , leading to the low observed galactic star formation rate . the sterility of much of the molecular gas is supported by the apparent absence of distributed populations of embedded stars in ngc 2023 ( li , evans , & lada 1997 ) . the extinction in the peripheral region of that cloud , estimated from emission is less than 10 mag , while that in the core is about 14 mag , just enough for shielding to decrease the ambipolar diffusion time . more recently , carpenter ( 2000 ) has used 2mass data to set upper limits on the distributed population in several clouds , again consistent with the result that star formation is largely confined to clusters forming in gas of higher than average extinction and density . of course , a long @xmath116 is only a necessary condition for the picture of photoionization - regulated star formation , not a sufficient one . the clouds must be magnetically subcritical as well . crutcher ( 1999 ) has summarized the data on this topic , concluding that clouds were slightly supercritical , but he noted that the uncertainties allowed the clouds to be subcritical . shu et al . ( 1999 ) have noted some corrections that make the clouds even closer to the critical boundary , and they suggested that observed clouds may be self - selected to be close to marginally critical . the h@xmath9 pure - rotational lines observed with the iso - sws provide a new test of the pdr models . early iso - sws results have shown that h@xmath9 lines up to @xmath40 s(5 ) are readily detected in traditional high density ( @xmath123 @xmath124 ) , high ( @xmath125 ) pdrs such as s140 at the position of the interface with the h ii region ( timmermann et al . 1996 , see review by wright 2000 ) . those data give excitation temperatures for @xmath126 of 400700 k , much higher than expected from pre - iso pdr models ( bertoldi 1997 , draine & bertoldi 1999 ) . because of the high densities , the h@xmath9 level populations are thermalized so that the h@xmath9 intensities directly trace the temperature structure in the warm pdr layers . to remedy these discrepancies , bertoldi ( 1997 ) suggested that either the photoelectric heating efficiency needs to be increased or that the h@xmath9 formation rate on grains needs to be larger at high temperatures , shifting the h@xmath127h@xmath9 transition zone closer to the warm edge . weingartner & draine ( 1999 ) have calculated increased photoelectric heating efficiencies based on an enhanced dust - to - gas ratio in the pdr due to gas - grain drift , and draine & bertoldi ( 2000 ) have shown that such models can reproduce the h@xmath9 observations of ngc 2023 . note that these comparisons include a factor of 510 enhancement of the models due to limb - brightening / geometry . in the low - density , low regions of s140 studied here , the h@xmath9 populations should be controlled by ultraviolet pumping rather than collisions , at least in the outer layers of the cloud . moreover , enhanced efficiencies due to gas - grain drift should not play a role , nor should there be any geometrical effects in this face - on cloud . thus , the iso - sws data should provide a test of the h@xmath9 excitation in a different pdr regime ( thi et al.1999 ) . for the pdr parameters derived in the previous sections , @xmath128 @xmath124 and @xmath129 , standard low density pdr models such as those by hollenbach et al . ( 1991 ) or le bourlot et al . ( 1993 ) and our own models give h@xmath9 excitation temperatures derived from the s(0 ) and s(1 ) lines of @xmath38100200 k , consistent with our iso data . however , these models underproduce the absolute h@xmath9 intensities by at least an order of magnitude . the same conclusion was reached by habart et al . ( 2000 ) and kemper et al.(1999 ) for somewhat denser pdrs exposed to low . habart et al . proposed to increase the h@xmath9 formation rate on grains by more than a factor of 5 compared with the standard h@xmath9 formation rate of @xmath130 cm@xmath7 s@xmath41 at 100 k ( jura 1975 ) to reproduce their observations . if the h@xmath9 formation rate in our s140 models is increased by a similar factor of 5 , the comparison with observations is improved considerably , although the models still fall short by a factor of a few . contributions from the far - side of the cloud could help to further reduce the discrepancies . such an arbitrary , large increase of the h@xmath9 formation rate is inconsistent , however , with the mechanism of the formation of h@xmath9 by diffusion of physisorbed h atoms on silicate grains , which is effective only at very low temperatures ( pirronello et al . a mechanism in which h atoms are chemically attached to the grains and in which the second h atom lands directly on the attached h atom , the so - called eley - rideal mechanism ( herbst 2000 ) is one possible mechanism to explain a large formation rate at high temperatures . porous ( carbonaceous ) grains and/or large pah molecules appear to be the best candidate materials , but quantitative experiments and calculations on interstellar grain analogs at high temperatures are needed . in any case , an increase in the surface to volume ratio over the standard grain models may be needed to allow the high formation rates that seem to be necessary to explain our observations . an alternative mechanism to increase the h@xmath9(@xmath132 ) populations has been proposed by spaans ( 1995 ) , joulain et al . ( 1998 ) and falgarone et al . ( 2000) using the energy released by the dissipation of turbulence or by weak ( few km s@xmath41 ) shocks . the observed h@xmath9 intensities toward s140 are within a factor of a few of those predicted by joulain et al . for a cloud with @xmath133 @xmath124 . at densities typical for the periphery of s140 , however , the turbulent model intensities should be reduced significantly . the models of joulain et al . do not take ultraviolet pumping into account , which will enhance the populations of the higher @xmath132 levels . further development of these models and constraints from other observed species will be needed to test them fully . it should be noted that the strengths of the s(0 ) and s(1 ) lines observed in the outskirts of s140 are comparable to those inferred from ultraviolet observations of h@xmath9 in diffuse and translucent clouds . consider as an example the recent fuse results for the translucent cloud toward hd 73882 ( @xmath134=2.4 mag ) by snow et al . the observed column densities in @xmath135 , 3 , and 5 translate into intensities of @xmath136 , @xmath137 , and @xmath138 erg s@xmath139 @xmath140 sr@xmath41 for the s(0 ) , s(1 ) , and s(3 ) lines , respectively . the s(1 ) intensity is comparable to our observed values , but that of s(0 ) is lower and of s(3 ) higher . indeed , the corresponding excitation temperatures for the @xmath1417 levels of @xmath142300 k are larger than those of @xmath143 k found here for the @xmath1443 levels , although they are consistent with that found for the @xmath735 levels at one position . more sensitive searches for the higher rotational lines of h@xmath9 are needed to constrain its excitation and determine the relative roles of ultraviolet pumping , collisions , and the h@xmath9 formation mechanism in establishing the level populations . fuse studies of the h/ transition in well - characterized , low column density regions are also interesting to provide further constraints on the empirical formation rate in different environments using the analysis of jura ( 1975 ) . in summary , the h@xmath9 excitation temperatures and the widespread nature of the emission suggest a pdr origin , in which the h@xmath9 lines arise from the outer warm layers . however , the absolute model intensities fall short by an order of magnitude , unless the h@xmath9 formation rate is enhanced significantly . other explanations invoked for high - density pdrs , such as geometry or gas - grain drift , do not apply to these regions . together with the results from ultraviolet observations , our results suggest that our understanding of the basic chemical processes involving h@xmath9 is still incomplete . the 158 @xmath34 m line and 63 @xmath34 m lines are the dominant coolants of the interstellar gas in galaxies . a significant fraction of the emission from late - type galaxies arises in dense pdrs with some additional contributions from atomic gas not associated with molecular clouds . in the milky way , the emission fraction is about equally divided between pdrs associated with molecular clouds and neutral , purely atomic clouds ( petuchowski & bennett 1993 , mochizuki & nakagawa 2000 , but see heiles 1994 for a different view ) . pdrs contribute a larger fraction of the emission from galaxies with more active star formation ( crawford et al . 1985 , stacey et al . 1991 ) while , particularly in the outer parts of galaxies with lower star formation rates , purely atomic gas may contribute to the emission ( madden et al . the continuum emission from galaxies comes primarily from the same pdrs , although , in the lowest luminosity spirals , optical emission from late - type stars may also contribute to the heating of dust in the molecular ism ( smith et al . 1994 ) and something like half the emission may come from purely atomic clouds ( mochizuki & nakagawa 2000 ) . it should be possible , in principle , to derive information about the physics of the neutral ism in galaxies from the continuum and fine - structure line measurements ( wolfire , tielens & hollenbach , 1990 , stacey et al . 1991 , carral et al . 1994 , lord et al . 1996 , malhotra et al . 2001 ) . recent studies of starburst galaxies ( luhman et al . 1998 ) and of a range of lower luminosity spirals ( malhotra et al . 1997 , 2001 ) have made much of an apparent `` deficit '' in the emission . in many of the starbursts and in a fraction of the spirals , the /fir luminosity ratio lies as much as a factor of 10 below the canonical value ( 3@xmath14510@xmath146 ) , more typical for late - type galaxies . galactic photodissociation regions with high @xmath99 or high @xmath147 can have /fir ratios as low or lower than those of galaxies with the biggest deficits ( e.g. genzel , harris , & stutzki 1989 ) . the [ o i ] line then provides a much larger fraction of the cooling power ( cf . the values for the s140 cut and the s140 irs 1 position in table [ s140lws - res ] ) , and dust temperatures are higher . however , malhotra et al . ( 2001 ) argue that the _ total _ fine structure emission ( plus ) is decreased substantially only if @xmath148 is increased . this combination leads to highly charged grains and inefficient photoelectric heating ( tielens & hollenbach 1985a , bakes & tielens 1994 ) . malhotra et al . ( 2001 ) considered but rejected a variety of explanations , concluding that high values of @xmath148 in the emitting regions provide the best explanation for their sample of `` normal '' spirals . studies of line emission from other galaxies have made use of the same models as adopted here , which assume a uniform incident radiation field and a single pdr density for the emitting material ( kaufman et al . 1999 , stacey et al . 1991 , wolfire et al . in addition to using simple models , we can gain insight into the nature of the regions in galaxies responsible for fine - structure line emission by comparing galaxy pdr emission to emission from galactic sources with an appropriate range of properties . the iso observations of fine - structure line emission from other galaxies take in entire giant molecular cloud complexes and the nuclear regions of galaxies . they therefore contain contributions not only from the high excitation regions near ob stars but also from large regions with lower uv fields . with our measurements of the cut through the outer part of the s140 cloud , a comparison of galactic and extragalactic fine - structure line emission can include a range of galactic pdr conditions . figure 6(a - c ) reproduces the plots shown in figures 1a , 5 , and 8 of malhotra et al . ( 2001 ) and shows the variation of log@xmath149}})/l(fir))$ ] ( fig . 6a ) , @xmath150}})/l({\mbox{[c\,{\scriptsize ii}]}})$ ] ( fig . 6b ) , and log@xmath151}})+l({\mbox{[o\,{\scriptsize i}]}}))/l(fir))$ ] ( fig . 6c ) versus @xmath152 for the 45 normal , star - forming galaxies in table 7 of malhotra et al . ( 2001 ) . figure 6 compares these data to the average values from our s140 cut and to typical values for high galactic pdr sources ( labeled for brevity ) , taken from genzel , harris , & stutzki ( 1989 ) to be @xmath153}})/l(fir)= 8.5\times10^{-5}$ ] , @xmath150}})/l({\mbox{[c\,{\scriptsize ii}]}})=5.8 $ ] , and @xmath154 . figure 6 clearly shows that the galaxies inhabit a part of parameter space that is different from either the s140 cut or the high pdr sources . because the galaxies generally lie between the points for the s140 cut and the high pdrs , we attempted to reproduce the galaxy observations by mixtures of the s140 cut and the high pdrs , indicated by the mixing line connecting these two points . the line represents parameters calculated by varying the contributions of the two connected galactic components in proportion to the fraction of the emission contributed by each . it is not possible , however , to produce the observed positions of the galaxies in the figures with any linear combination of the s140 cut and the high pdrs . the result is similar if we use the emission from ngc 7023 , representing reflection nebulae , instead of the regions . at all values of @xmath152 , the mixture produces too much fine - structure line emission ( in particular , too much ) , when compared to the galaxy sample . at a given value of @xmath152 , for example , the value of log@xmath151}})+l({\mbox{[o\,{\scriptsize i}]}}))/l(fir))$ ] observed for the galaxies lies on average 0.3 below the constructed mixture . the results in figure 6 show that one can not produce the observed emission ratios in galaxies by simply combining regions like the s140 cut with regions of active star formation in our galaxy . malhotra et al . ( 2001 ) argued that models with high @xmath148 can explain individual galaxies where the [ cii ] and [ oi ] is weak compared to the far - ir continuum . the systematically lower values of the galaxy ratios compared to positions along our mixing line , however , suggests that the conditions in heterogeneous clouds would not lead to appropriately weak line fluxes at a given value of @xmath155 . rather than focussing solely on the line to continuum ratio , it may be worthwhile to examine possible reasons why the @xmath155 ratio might vary . if that flux ratio increased less rapidly along the mixing line , the data could be matched . mixing very small amounts of high pdr emission has a big effect on @xmath152 , while the effect on the line emission is more modest , as indicated by the tick marks on the mixing line that indicate where 90% and 50% of the emission comes from the s140 cut . moderating the increase in @xmath152 as one mixes in small amounts of high pdr emission properties could allow the mixing line to pass through most of the points . this would require a component of emission from cooler dust . lowering the 60100 @xmath34 m color temperature of the high pdrs from @xmath156 k to @xmath157 k ( for a @xmath158 emissivity law ) could account for the difference . even colder dust is common in the molecular clouds near the center of our galaxy ( lis et al . 2001 ) . there are at least two possible causes for the cooler dust . the dust properties in the inner few kiloparsecs of galaxies , where most of the iso emission arises , might differ significantly from those in nearby high galactic pdrs . alternatively , there might be a contribution from regions heated by cooler stars or from very opaque regions in molecular clouds . these regions would need to increase the fraction of the emission in the 100 @xmath34 m iras band from 25% to 40% of the total luminosity . whichever reason is correct , observations of some regions in the inner part of our own galaxy ( g0.68 - 0.2 , iso data archive ) find a line to continuum ratio lower than for all the galaxies in the malhotra sample , and yet the @xmath152 is only 0.2 . a mixing line connecting g0.68 - 0.2 to the region / high pdr s would run below the bulk of the galaxies in figures 6a and 6c . more complex 3-part mixtures of cold cloud edges like the s140 cut , cold inner galaxy clouds like g0.68 and warm regions could presumably fill the part of parameter space occupied by the galaxies from the malhotra sample . understanding the properties of inner galaxy regions like g0.68 may provide further insight into the conditions in extragalactic pdrs . we have studied the peripheral region of the molecular cloud l1204/s140 , where the fuv radiation and density are relatively low , using iso to observe , , and lines , and the cso to observe the submillimeter line of atomic carbon and millimeter emission lines of and other co isotopes . we analyzed the results with published pdr models and our own pdr models , coupled to a monte carlo simulation code for radiative transfer . the important results from this work are : 1 . we detected wide - spread , , and emission in the region . the and lines display much less spatial variation than and the lines from and other co isotopes , which reveal the presence of a denser , shielded core that does not emit significantly in the and lines . 2 . the average and intensities of the s140 cut are @xmath159 and @xmath160 ergs s@xmath41 @xmath65 sr@xmath41 ; the average line ratio ( / ) is @xmath66 . therefore the line is the dominant cooling line in this region . this is quite different from what has been observed in the s140 bright pdr by iso , where the ratio / is @xmath161 . 3 . using the model predictions in hollenbach et al . ( 1991 ) and the 100 iras flux , the fuv intensity for the region would be 4060 , in the situation of single - sided fuv radiation on the cloud . if the fuv impinges on both sides , the inferred @xmath162 . based on emission and simple lvg modeling , the density of the gas , @xmath103 , and @xmath163 . conditions in the shielded core are different : @xmath164 , and @xmath165 . the pdr models ( both published and our own ) indicate that the observed intensity and the ratio / are best reproduced with @xmath0 and @xmath103 . our models , coupled to a monte carlo simulation of the radiative transfer , reproduce the observed emission for @xmath166 mag . these conditions are also consistent with the constraints on density from line ratios of and . our models also reproduce the observed line , while the published models of kaufman et al . ( 1999 ) overproduce this line . in this regime of @xmath99 and @xmath147 , the intensity is very difficult to model because it is extremely subthermal and also very opaque ( the optical depth is 914 ) . trapping is very important in determining the emergent intensity . quantitative analysis of the ionization structure and timescale for cloud collapse for the peripheral regions of s140 leads to conclusions consistent with a picture of photoionization - regulated star formation . the detection of widespead rotational emission is not readily explained with standard pdr models and formation rates . possible solutions include a substantial enhancement in the the formation rates or localized heating by weak shocks associated with turbulent decay . 7 . while the emission in and continuum from other galaxies lies between the characteristics of the emission from the peripheral region studied here and regions of higher and @xmath99 , simple mixtures of these regions are not able to reproduce the characteristics of the emission from other galaxies . we thank the referee for a careful reading and for suggestions that improved the paper . we would like to extend our sincere thanks to the wonderful service by the iso supporting staff at ipac ( s. lord , s. unger , d. levine , l. hermans especially ) and at sron groningen ( e. valentijn and f. lahuis in particular ) . we also thank the cso staff members and e. gregersen and k. mochizuki for helping with the cso observations . m. spaans and d. jansen provided useful discussions and help with pdr codes ; e. roueff provided data for comparison of pdr models . this research has made use of nasa s astrophysics data system abstract service , the simbad database , operated at cds , strasbourg , france , and the online services of irsky and ibis at ipac , and the skyview online service . wl was partially supported by a continuing fellowship and a david bruton , jr . , fellowship of the university of texas , a frank n. edmonds , jr . memorial fellowship and a david alan benfield memorial scholarship of the department of astronomy , the university of texas . work with the cso re - imager was supported by nsf grant ast-9530695 . the research was supported by nasa grants nag2 - 1055 , nag5 - 3348 , and the state of texas , nwo grant 614.41.003 , a nwo spinoza grant , and a nwo bezoekersbeurs . bertoldi , f. , & mckee , c.f . 1997 , rev . aa ( serie de conferencias ) , 6,195 black , j. h. & dalgarno , a. 1977 , , 34 , 405 . blair , g. n. , evans , n. j. , ii , vanden bout p. a. , & peters , w. l. , iii 1978 , , 219 , 896 homepage . carpenter , j. m. 2000 , , 120 , 3139 carral , p. , hollenbach , d.j . , lord , s.d . , colgan , s.w.j . , haas , m.r . , rubin , r.h . , & erickson , e.f . 1994 , apj , 423 , 223 clegg , p. e. et al . 1996 , , 315 , l38 . crampton , d. & fisher , w. a. 1974 , pub . , 14 , 12 . crawford , m.k . , genzel , r. , townes , c.h . , & watson , d. , 1985 , apj , 291,755 crutcher , r. m. 1999 , , 520 , 706 de graauw , t. et al . 1996 , , 315 , l49 . draine , b. t. 1978 , , 36 , 595 . draine , b. t. & bertoldi f. 1996 , , 468 , 269 . draine , b.t . , bertoldi , f. 2000 , in molecular hydrogen in space , ed . f. combes and g. pineau des forts ( cambridge university press ) , p. 131 . draine , b. and bertoldi , f. 1999 , in the universe as seen by iso , eds . p. cox , m.f . kessler , esa sp-427 , p. 553 . emery , r. , et al . 1996 , , 315 , l285 . evans , n. j. , ii 1991 , _ frontiers of stellar evolution _ , ed . d. l. lambert ( san francisco : astronomical society of the pacific ) , p. 45 . evans , n. j. , ii , , 37 , 311 falgarone , e. , verstraete , l. , hily - blant , p. , pineau - des - forts , g. 2000 , in molecular hydrogen in space , ed . f. combes and g. pineau des forts ( cambridge university press ) , p. 225 . federman , s. r. , cardelli , j. a. , van dishoeck , e. f. , lambert , d. l. , & black , j. h. 1995 , , 445 , 325 . frerking , m. a. , langer , w. d. , & wilson , r. w. 1982 , , 262 , 590 . galavis , m. e. , mendoza , c. , & zeippen , c. j. 1997 , , 123 , 159 . genzel , r. , harris , a.i . , & stutzki , j. 1989 , in proc . 22nd eslab symposium , b. kaldeich ed . , esa sp-290 , 115 gry , c. , et al . 2001 , the iso handbooks , vol . iv ( lws ) habart , e. , boulanger , f. , verstraete , l. , pineau des forts , g. , falgarone , e. , abergel , a. 2000 , in iso beyond the peaks , esa sp-456 , p. 103 habing , h. j. 1968 , bull . netherlands , 19 , 421 . heiles , c. 1994 , apj , 436 , 720 herbst , e. 2000 , astrochemistry : from molecular clouds to planetary systems , iau symposium 197 , eds . minh and e.f . van dishoeck ( san francisco : asp ) , p. 147 hogerheijde , m. r. , jansen , d. j. , & van dishoeck , e. f. 1995 , , 294 , 792 . hogerheijde , m. r. , & van der tak , f. f. s. 2000 , , 362 , 697 hollenbach , d. j. , takahashi , t. , & tielens , a. g. g. m. 1991 , , 377 , 192 . hollenbach , d. j. & tielens , a. g. g. m. 1999 , rev mod . 71 , 173 hollenbach , d. j. & tielens , a. g. g. m. 1995 , _ the physics and chemistry of interstellar molecular clouds _ , ed . g. winnewisser & g. c. pelz ( berlin : springer ) , p. 164 jansen , d. j. , van dishoeck , e. f. , black , j. h. , spaans , m. c. , sosin , c. 1995 , , 302 , 223 . jansen , d. j. , van dishoeck , e. f. , keene , j. , boreiko , r. t. , & betz , a. l. 1996 , , 309 , 899 . joulain , k. , falgarone , e. , pineau des forts , g. , flower , d.1998 , a&a , 340 , 241 kaufman , m. j. , wolfire , m. g. , hollenbach , d. j. , & luhman , m. l. 1999 , , 527 , 795 . keene , j. , blake , g. a. , phillips , t. g. , huggins , p. j. , & beichman , c.a . 1985 , , 299 , 967 . kemper , c. , spaans , m. , jansen , d.j . , hogerheijde , m.r . , van dishoeck , e.f . , tielens , a.g.g.m . 1999 , apj , 515 , 649 kessler , m. f. et al . 1996 , , 315 , l27 . le bourlot , j. , pineau des forts , g. , roueff , e. , & flower , d. r. 1993 , , 267 , 233 . leech , k. et al . 2001 , the iso handbooks , vol . vi ( sws ) li , w. , evans , n. j. , ii , & lada , e. a. 1997 , , 488 , 277 . d. c. , serabyn , e. , zylka , r. , & li , y. 2001 , , 550 , 761 liseau , r. et al . 1999 , , 344 , 342 lord , s.d . , hollenbach , d.j . , haas , m.r . , rubin , r.h . , colgan , s.w.j . , & erickson , e.f . 1996 , apj , 465 , 703 madden , s. c. , geis , n. , genzel , r. , herrmann , f. , jackson , j. , poglitsch , a. , stacey , g. j. , & townes , c. h. 1993 , , 407 , 579 malhotra , s. et al . 1997 , apj 491 , 27 malhotra , s. et al . 2001 , apj , 561 , 766 mckee , c. f. , zweibel , e. g. , goodman , a. a. , & heiles , c. 1993 , protostars and planets iii , 327 mckee , c. f. 1989 , , 345 , 782 . mochizuki , k. , & nakagawa , t. 2000 , apj , 535 , 118 petuchowski , s. j. & bennett , c. l. 1993 , , 405 , 591 pirronello , v. , biham , o. , liu , c. , shen , l. , vidalu , g. 1997 , apj , 483 , l131 plume , r. , & jaffe , d. t. 1995 , , 107 , 488 . plume , r. , jaffe , d. t. , & keene , j. 1994 , , 425 , l49 . pringle , j. e. , allen , r. j. , & lubow , s. h. 2001 , , 327 , 663 savage , b. d. & sembach , k. r. 1996 , , 34 , 279 shu , f. h. , allen , a. , shang , h. , ostriker , e. c. , & yun , z - y . 1999 , the origin of stars and planetary systems , ed . c. j. lada & n. d. kylafis ( dordrecht : kluwer ) , 193 smith , b.j . , harvey , p.m. , colome , c. , zhang , c.y . , di francesco , j. , & pogge , r.w . 1994 , apj , 425 , 91 snow , t.p . 2000 , apj , 538 , l65 steiman - cameron , t. y. , haas , m. r. , tielens , a. g. g. m. , & burton , m. g. 1997 , , 478 , 261 . sternberg , a. & dalgarno , a. 1989 , , 338 , 197 . swinyard , b. m. et al . 1996 , , 315 , l43 . swinyard , b.m . et al . 1998 , spie , 3354 , 896 tafalla , m. , bachiller , r. , & martin - pntado , j. 1993 , , 403 , 175 . tauber , j. , tielens , a. g. g. m. , meixner , m. , & goldsmith , p. 1994 , , 422 , 136 . thi , w. f. et al . 2001 , , 561 , 1074 tielens , a. g. g. m. & hollenbach , d. j. 1985a , , 291 , 722 . tielens , a. g. g. m. & hollenbach , d. j. 1985b , , 291 , 747 . timmermann , r. , bertoldi , f. , wright , c. m. , drapatz , s. , draine , b. t. , haser , l. , & sternberg , a. 1996 , , 315 , l281 . valentijn , e. a. , & thi , w.f . 2000 , experimental astronomy , 10 , 215 van dishoeck , e. f. & black , j. h. 1986 , , 62 , 109 . van dishoeck , e. f. & black , j. h. 1988 , , 334 , 771 . van dishoeck , e. f. & black , j. h. 1989 , , 340 , 273 . weingartner , j.c . , draine , b.t . 1999 , in the universe as seen by iso , eds . p. cox & m.f . kessler , esa sp-427 ( noordwijk : estec ) , p. 783 wolfire , m. g. , tielens , a. g. g. m. , & hollenbach , d. j. 1990 , its the interstellar medium in external galaxies : summaries of contributed papers p 37 - 39 ( see n91 - 14100 05 - 90 ) , 37 zhou , s. , butner , h. m. , evans , n. j. , ii , gsten , r. , kutner , m. l. , & mundy , l. g. 1994 , , 428 , 219 zuckerman , b. , & palmer , p. 1974 , , 12 , 279 cc ref . coord . & ( 221742 , 632315@xmath31 ) + end position 1 & ( 1368@xmath31 , 0 ) + end position 2 & ( @xmath167 , 0 ) + step & 99@xmath31 + cso off pos . & ( 221322 , 632315@xmath31 ) + iso off pos . & ( 2205467 , 632315@xmath31 ) + lrrccccc june 1995 & s140 & & 230.53797 & 0.23 & rimg & 120 & 0.80 + & & & 220.39870 & 0.24 & rimg & 120 & 0.80 + & & & 219.56032 & 0.23 & rimg & 120 & 0.80 + & & & 330.58800 & 0.12 & rimg & 110 & 0.80 + june 1996 & s140 & @xmath168 & 492.16070 & 0.08 & fd & 14 & 0.45 + april 1997 & s140 & & 267.55762 & 0.12 & rimg & 160 & 0.80 + & & & 267.55762 & 0.12 & fd & 26 & 0.70 + lrrrrrcc @xmath7p@xmath169p@xmath9 & 63.183705 & 227.72 & 0 & 4.7(5 ) & 8.87(-5 ) & lws & 1.3 + @xmath4p@xmath170p@xmath6 & 157.7409 & 91.22 & 0 & 2.8(3 ) & 2.4(-6 ) & lws & 6.65 + 20 & 28.218 & 509.88 & 0 & 5.4(1 ) & 2.9(-11 ) & sws & 1.3 + 31 & 17.035 & 1015.12 & 170.48 & 1.1(3 ) & 4.8(-10 ) & sws & 8.9 + 53 & 9.662 & 2503.82 & 1015.12 & 1.9(5 ) & 9.8(-9 ) & sws & 6.6 + crcccccccccccccc 1 & ( @xmath171 , 0 ) & ... & 1.1 & & 0.6 & ... & & ... & ... & & ... & ... & & ... & ... + 2 & ( @xmath172 , 0 ) & 2.7 & 1.8 & & 0.8 & ... & & ... & ... & & ... & ... & & ... & ... + 3 & ( 81@xmath31 , 0 ) & 3.9 & 3.4 & & 1.2 & ... & & ... & ... & & 0.1 & ... & & ... & ... + 4 & ( 180@xmath31 , 0 ) & 6.7 & 3.2 & & 3.9 & ... & & 1.4 & ... & & 0.4 & ... & & 1.6 & ... + 5 & ( 279@xmath31 , 0 ) & 4.5 & 5.4 & & 6.2 & ... & & 1.3 & 0.5 & & 0.4 & ... & & 2.8 & ... + 6 & ( 378@xmath31 , 0 ) & 4.1 & 5.8 & & 6.5 & ... & & 0.8 & ... & & 0.3 & ... & & 3.2 & 1.5 + 7 & ( 477@xmath31 , 0 ) & 7.2 & 5.1 & & 4.3 & 2.2 & & 1.1 & 0.9 & & 0.2 & 0.1 & & 3.7 & 1.5 + 8 & ( 576@xmath31 , 0 ) & 7.2 & 9.2 & & 6.4 & 5.6 & & 1.5 & 1.6 & & 0.3 & 0.4 & & 3.7 & 4.3 + 9 & ( 675@xmath31 , 0 ) & 7.7 & 9.5 & & 6.1 & 10.6 & & 1.4 & 2.6 & & 0.5 & 1.4 & & 2.7 & 3.8 + 10 & ( 774@xmath31 , 0 ) & 8.6 & 11.8 & & 7.3 & 15.7 & & 2.5 & 5.7 & & 1.2 & 2.8 & & 2.6 & 6.3 + 11 & ( 873@xmath31 , 0 ) & 6.5 & 11.3 & & 4.4 & 14.8 & & 2.5 & 6.2 & & 0.4 & 2.5 & & ... & 6.0 + 12 & ( 972@xmath31 , 0 ) & 6.3 & 10.2 & & 3.6 & 8.0 & & 1.7 & 0.8 & & 0.1 & 1.4 & & ... & 3.0 + 13 & ( 1071@xmath31 , 0 ) & 6.6 & 7.9 & & 3.0 & 2.4 & & ... & ... & & ... & 0.2 & & ... & 3.9 + 14 & ( 1170@xmath31 , 0 ) & 5.6 & 3.6 & & 1.3 & 1.3 & & ... & ... & & ... & ... & & ... & 1.5 + 15 & ( 1269@xmath31 , 0 ) & 3.7 & 4.1 & & ... & 2.7 & & ... & ... & & ... & ... & & ... & 1.6 + 16 & ( 1368@xmath31 , 0 ) & 3.3 & 2.4 & & ... & 1.5 & & ... & ... & & ... & ... & & ... & 1.2 + cloud " & & 5.6 & 4.0 & & 4.0 & 2.0 & & 1.6 & 0.70 & & 0.39 & 0.15 & & 2.9 & 1.9 rrccccc 1 & ( @xmath177117,0 ) & [email protected] & [email protected] & 0.12 & 57.3 & 219 2 & ( @xmath17718,0 ) & [email protected] & [email protected] & 0.11 & 57.2 & 221 3 & ( 81,0 ) & [email protected] & [email protected] & 0.21 & 58.8 & 220 4 & ( 180,0 ) & [email protected] & [email protected] & 0.14 & 66.1 & 244 5 & ( 279,0 ) & [email protected] & [email protected] & 0.12 & 61.5 & 236 6 & ( 378,0 ) & [email protected] & [email protected] & 0.17 & 56.2 & 232 7 & ( 477,0 ) & [email protected] & [email protected] & 0.10 & 52.9 & 226 8 & ( 576,0 ) & [email protected] & [email protected] & 0.15 & 52.5 & 226 9 & ( 675,0 ) & [email protected] & [email protected] & 0.13 & 54.3 & 228 10 & ( 774,0 ) & [email protected] & [email protected] & 0.16 & 58.8 & 230 11 & ( 873,0 ) & [email protected] & [email protected] & 0.21 & 63.4 & 227 12 & ( 972,0 ) & [email protected] & [email protected] & 0.20 & 63.0 & 224 13 & ( 1071,0 ) & [email protected] & [email protected] & 0.15 & 58.3 & 208 14 & ( 1170,0 ) & [email protected] & [email protected] & 0.16 & 51.3 & 191 15 & ( 1269,0 ) & [email protected] & [email protected] & 0.19 & 44.8 & 164 16 & ( 1368,0 ) & [email protected] & [email protected] & 0.19 & 39.6 & 145 & avg(16-pos ) & [email protected] & [email protected] & [email protected] & & & iso - off & [email protected] & [email protected] & 0.27 & 9.6 & 47.7 & irs1 & 44@xmath1782 & 68@xmath1783 & @xmath179 & & 1 & ( @xmath180,0 ) & [email protected] & [email protected] & @xmath181 & @xmath182 & @xmath183 `` f '' & ( 743,0 ) & [email protected] & [email protected] & @xmath181 & @xmath184 & @xmath183 14 & ( 1170,0 ) & [email protected] & [email protected] & @xmath185 & @xmath186 & @xmath187	we have carried out an in - depth study of the peripheral region of the molecular cloud l1204/s140 , where the radiation and the density are relatively low . our observations test theories of photon - dominated regions ( pdrs ) in a regime that has been little explored . knowledge of such regions will also help to test theories of photoionization - regulated star formation . 158 and 63 lines are detected by iso at all 16 positions along a 1-dimensional cut in right ascension . emission from rotational transitions 20 and 31 , at 28 and 17 , was also detected at several positions . the , , and intensities along the cut show much less spatial variation than do the rotational lines of and other co isotopes . the average and intensities and their ratio are consistent with models of pdrs with low radiation ( ) and density . the best - fitting model has @xmath0 and density , @xmath1 . standard pdr models underpredict the intensity in the rotational lines by up to an order of magnitude . this problem has also been seen in bright pdrs and attributed to factors , such as geometry and gas - grain drift , that should be much less important in the regime studied here . the fact that we see the same problem in our data suggests that more fundamental solutions , such as higher formation rates , are needed . also , in this regime of low density and small line width , the line is sensitive to the radiative transfer and geometry . using the ionization structure of the models , a quantitative analysis of timescales for ambipolar diffusion in the peripheral regions of the s140 cloud is consistent with a theory of photoionization - regulated star formation . observations of in other galaxies differ both from those of high pdrs in our galaxy and from the low regions we have studied . the extragalactic results are not easily reproduced with mixtures of high and low regions .
consider a classical constrained theory @xcite , with configurations @xmath0 , conjugate momenta @xmath1 , classical brackets @xmath2\|}$ ] and constraints @xmath3 . _ observables _ @xcite are then objects forming zero brackets with the constraints , _ o _ 0 . [ basic - obs ] there are multiple such notions due to ` zero ' , ` brackets ' and ` the constraints ' taking a variety of different precise meanings @xcite . some such notions of observables can be more physically useful than just any functions ( or functionals ) of @xmath0 and @xmath1 . this is because of their greater physical content ; moreover , at least some versions of such notions of observables would be expected to contain physical information only . at the very least this property is required in phrasing final answers to physical questions about a theory . the zero can for instance be the usual notion of zero alias _ strongly zero _ in this context or dirac s notion @xcite of _ weakly zero _ , denoted @xmath4 , meaning zero up to terms linear in constraints . the bracket can for instance , at the classical level , be the usual poisson bracket ; we restrict attention to this case in the rest of this article . the constraints can include all the constraints , or a subset thereof . three particular cases of this which are well - known enter the current article . \1 ) _ classical dirac observables _ @xcite are quantities @xmath5 that ( usually weakly poisson ) brackets - commute with _ all _ of a given theory s first - class constraints , _ d _ 0 . [ c - d ] various interpretations proposed for these entities gave further colourful names for these , such as ` evolving constants of the motion ' @xcite and ` perennials ' @xcite . the names ` true ' and ` complete ' observables have also been used in similar contexts @xcite . conceptually , these are ` _ maximally constrained _ ' observables ; the current article will also give mathematical cause to call these ` _ zeroth observables _ ' . \2 ) ` _ unconstrained observables _ ' @xmath6 , on the other hand have no constraint - related restrictions . ` _ unital observables _ ' is also a mathematical name for these for reasons which will become clear as this article develops . ` partial observables ' @xcite are a well - known particular notion of unconstrained observables ( though these play no further part in the current article ) . \3 ) introduced @xcite a further notion of observables , which subsequently indeed came to be termed _ observables_. these are quantities @xmath7 which ( again usually weakly poisson ) brackets - commute with all of a given theory s first - class linear constraints , _ k _ 0 . [ flin - k ] using observables reflects treating first - class quadratic constraints , @xmath8 , distinctly from first - class linear ones @xmath9 @xcite . the problem of time literature @xcite is replete with reasons why this may be desirable ; moreover also observables are more straightforward to find than dirac ones . indeed , the _ problem of observables _ that it is hard to construct a sufficiently large set of these to describe all physical quantities , especially for gravitational theory is one of the many facets of the problem of time . observables are more straightforward to construct than dirac ones , and classical ones are more straightforward to construct than quantum ones . the problem of observables remains little understood . strategies proposed in this regard include the following . strategy 1 ) insist upon constructing dirac observables , @xmath5 . strategy 2 ) make use of unconstrained observables , @xmath6 @xcite . in strategy 2 ) case one is disregarding all the information content of the constraints . however , on some occasions strategy 2 ) is used as a stepping stone to constructing dirac - type observables . strategy 3 ) consider observables , @xmath7 , to suffice . some examples are also useful at this stage . in unconstrained theories , all three notions coincide . in theories whose sole constraint is first - class quadratic , unconstrained and observables are coincident , trivial and distinct from dirac observables , which are nontrivial . this applies to minisuperspace models and to some of the simplest relational mechanics @xcite models . in electromagnetism and yang mills theory , on the other hand , all the constraints are first - class linear , so dirac and observables coincide and are distinct from unconstrained observables . finally , in more advanced relational mechanics theories @xcite and in gr , there are first - class linear constraints and a first - class quadratic constraint . here , none of the above three notions of observables coincide . in particular , for gr the first - class linear constraints are the vector - valued gr momentum constraint @xmath10 , whereas the first - class quadratic constraint is the scalar - valued gr hamiltonian constraint @xmath11 . the observables then obey _ i k _ 0 , [ k_gr ] where @xmath12 is specifically the poisson bracket . for relational mechanics , @xmath13 which has the mathematical form of a quadratic energy constraint takes over the role of @xmath11 . here many examples observables are known explicitly @xcite : they are shapes , or shapes and scales , for various different notions of shape @xcite , and conjugate quantities . for electromagnetism , the observables forms are also well - understood @xcite , however for gr they remain only formally understood @xcite . the next consideration of note is dirac s conjecture , by which first - class linear constraints would always coincide with gauge constraints . unfortunately , this is false @xcite . due to this , a fourth notion of observable is unearthed : the @xmath14-observables @xmath15 which commute with all of a theory s gauge constraints : g _ g _ 0 ; @xmath14 is that gauge theory s underlying gauge group . of course , for some theories @xmath9 and @xmath16 coincide , in which case @xmath7 and @xmath15 coincide as well ; e.g. this is the case in relational mechanics , electromagnetism and yang mills theory . strategy 4 ) is then to consider @xmath14-observables observables to suffice . the next development follows from perceiving that a notion of observables being meaningful is contingent on the subset of constraints it commutes with a fortiori closing as a subalgebraic structure @xcite ( lemma 1 of appendix [ observables ] ) . in particular , in some cases the @xmath9 do not close , and thus indeed the notion of observables @xmath7 does not exist for that theory . in problem of time considerations , this means that many approaches to the problem of observables can only be addressed once the constraint closure problem has been resolved @xcite . i.e. one needs the constraints algebraic structure prior to consideration of associated observables algebraic structures . it is also required that the notion of bracket involved in defining the @xmath17 matches that under which the @xmath3 close . additionally , by appendix [ observables ] s lemma 2 , the @xmath17 themselves form a closed algebraic structure . [ it is , moreover , interesting that observables algebraic structures exhibit some conceptual similarities with theory of casimirs : appendix [ associated ] , which proved to be a central ingredient in understanding ( much simpler ) algebraic structures : lie groups and their reps . ] along the above lines , focus shifts from observables to the observables associated with whichever closed subalgebraic structures each theory s constraint algebraic structure happens to possess . i term these a - observables , where the ` a ' stands for ` algebraic substructure ' . the subalgebraic structures of a gives algebraic structure then turn out to form a bounded lattice ( appendix [ lattices ] ) under the inclusion operation . the association by ` forms zero brackets with ' then produces in a further lattice , now of notions of observables . this pair of bounded lattice structures form the current article s main underlying result . fig [ latt-1 ] gives preliminary versions of this involving @xmath6 , @xmath5 , @xmath7 and @xmath15 . the current article illustrates this with nontrivial examples of notions of a - observable arising in relational mechanics , slightly inhomogeneous cosmology and supergravity , in secs 2 , 3 and 4 respectively . we conclude in sec 4 with a more general account of the lattice of constraint subalgebraic structures and of the associated lattice of notions of a - observables . discussion of the article s range of algebraic concepts is difered to a multi - part appendix . the constraint algebraic structures for these are outlined appendix [ cas ] . some of the constraint subalgebras are then given in fig [ latt-2 ] , alongside the associated notions of a - observables . in this model @xcite @xmath18 , and one splits modewise ( each mode corresponds to collection of @xmath19-harmonics labels ) and into scalar , vector and tensor ( svt ) parts . this splits @xmath11 into @xmath20 , @xmath21 and @xmath22 pieces , and @xmath10 into @xmath23 and @xmath24 pieces . modewise and svt - wise splits give decoupled equations at the unreduced level , to lowest nontrivial perturbative order . it then turns out that @xcite taking out @xmath23 and @xmath24 fails to take out all of the @xmath25 information left in the model . the amount by which it fails to do so depends on whether one is considering the single minimally coupled scalar field matter version @xcite or the vacuum version @xcite . in the vacuum case , it is out by 1 configuration space degree of freedom , whereas it is out by 2 in the scalar field case . the vacuum case also has a more decoupled partly reduced system , by which it is clear @xcite that removing @xmath21 suffices to take one down to @xmath25 invariance , i.e. to this model s counterpart of @xmath26 . it is this vacuum case that the current article considers in detail ; the scalar field counterpart remains only partly understood . in this vacuum case , then , the constraint algebraic structure is as in fig [ latt-3].a ) and the corresponding notions of observables are as in fig [ latt-3].b ) . firstly , note that observables and @xmath14-observables for @xmath27 are distinct in this example . i also term the latter superspace observables . secondly , note that there are additional notions of a - observables , for which i suggest the names s - observables and v - observables , meaning commutation with just @xmath23 and just @xmath24 respectively . taking out just @xmath24 has a scalar field case parallel which underlies wada s even more partial reduction in @xcite ; he then proceeded to reduce out @xmath23 as well @xcite . introducing a local lorentz frame formulation to accommodate the model s fermionic species causes a local lorentz frame constraint @xmath28 ( and its conjugate ) to appear . the canonical formulation of supergravity then has the expected @xmath11 , @xmath10 and @xmath28 constraints alongside a specifically supersymmetric constraint @xmath29 . all we need to know for this article about @xmath28 and @xmath29 are that each of these constraints is linear in the momenta , and the schematic form of the subsequent constraint brackets , as per appendix [ cas ] . the particular feature this article concentrates upon is that in supergravity a subset of the linear constraints the supersymmetry constraints have the supergravity counterpart of the quadratic @xmath30 as _ their _ integrability ; this follows from eq ( [ s - s->h ] ) . this means that observables are not well - defined for supergravity , since the quadratic @xmath11 now arises as an integrability of the linear @xmath29 . eq ( [ s - s->h ] ) forms the second integrability of the ` two - way ' pair , to the first integrability being of form ( [ dirac - algebroid ] ) . some constraint subalgebraic structures of note are then provided in fig [ latt-3 ] . in particular , the non - supersymmetric linear constraint combination of @xmath28 and @xmath10 closes , as does the gr - like combination of @xmath11 and @xmath10 , and the non - supersymmetric combination of @xmath28 , @xmath10 and @xmath11 . i term the corresponding notions of a - observables non - supersymmetric , gr - likes and non - supersymmetric dirac @xcite . there is a notion of observables associated with each constraint subalgebraic structure that a given theory s constraint algebraic structure happens to possess . i term these , collectively , a - observables ( fig [ latt-5 ] ) . both the constraint subalgebraic structures of a given theory s constraint algebraic structure , and the theory s notion of a - observables form bounded lattice structures . the dirac observables @xmath5 correspond to the zero of the latter lattice , arising from considering the whole constraint algebraic structure , which is the unit of the former lattice . on the other hand , the unconstrained observables @xmath6 correspond to the unit of the latter lattice , arising from considering the trivial subalgebraic structure ( i d ) , which is the zero of the former lattice . once this context has been framed , it is clear that the notion observables commuting with first - class linear constraints is only one some occasions well - defined . this becomes seen , rather , as an early example of nontrivial a - observable ( i.e. neither unconstrained nor dirac ) , which is realized in some theories such as gr but not others such as supergravity . thus approaches to quantum gravity which depend on the quadratic to linear distinction between constraints , and the subsequent notion of observables , can be fragile to change of theory . this is to be contrasted with how @xmath5 and @xmath6 are defined for _ all _ theories . moreover , other interesting closed subalgebraic structures arise , each with an associated notion of a - observable . in the current article , we have discussed a number of such , and coined names for them . \a ) frame observables , non - supersymmetric observables , gr - like observables and non - supersymmetric dirac observables in the case of supergravity . \b ) superspace observables as distinct from observables in slightly inhomogeneous cosmology , alongside scalar - only and vector - only observables . \c ) a wide range of geometrically meaningful a - observables in the context of relational theories of mechanics . the introduction s list of strategies toward dealing with the problem of observables can then be updated to include the following . strategy 5 ) conceive in terms of the lattice of notions of a - observables @xmath31 , such that a _ c _ ` = ' 0 , for @xmath32 whichever ( closed ) subalgebraic structure of the constraint algebraic structure . this strategy covers \1 ) making whichever such choice . \2 ) the possibility of having further selection principles among the various candidate theories whose constraint algebraic structures admit a variety of nontrivial proper subalgebraic structures . \3 ) using one or more notions of a - observables as intermediate stages in computing more restrictive types of observables . note moreover that the dirac observables strategy is both _ always available _ and _ requires no choice or further selection principle_. this is a direct consequence of taking the whole constraint algebraic structure , as follows . the same two maxims also apply at the opposite extreme of the lattice of subalgebraic structures : the unconstrained observables . i end by pointing out two future research directions . \1 ) increase the scope of the examples used to include more supersymmetric relational mechanics as well as superconformal supergravity . \2 ) quantum mechanical counterparts and quantization process from classical to quantum versions remain to be presented @xcite . * acknowledgements * i thank those close to me . i thank the fqxi and damtp for travel money and hosting in 2014 - 2015 ( slightly inhomogeneous cosmology ) and 2015 - 2016 ( supergravity ) . i also thank jeremy butterfield , chris isham , malcolm maccallum , don page , enrique alvarez , john barrow , marc lachieze - rey and reza tavakol for discussions or helping me with my career . @xmath14 @xcite are simultaneously groups and differentiable manifolds ; additionally their composition and inverse operations are differentiable . working with the corresponding infinitesimal ` tangent space ' around @xmath14 s identity element the lie algebra @xmath33 is more straightforward due to vector spaces tractability , while very little information is lost in doing so . for instance , the representations of @xmath33 determine those of @xmath14 . more formally , a _ lie algebra _ is a vector space equipped with a product ( bilinear map ) @xmath34\| } : \frg \times \frg \longrightarrow \mbox{\frg}$ ] that is antisymmetric and obeys the leibniz ( product ) rule and the _ jacobi identity _ g_1 g_2 g_3 + = 0 [ jacobi - id ] @xmath35 . this an example of _ algebraic structure _ : equipping a set with one or more further product operations . particular subcases of lie brackets include the familiar poisson brackets and quantum commutators . moreover , a lie algebra s generators ( lie group generating infinitesimal elements ) @xmath36 obey _ _ = c^ _ _ , [ str - const ] where @xmath37 are the _ structure constants _ of that lie algebra . it readily follows that the structure constants with all indices lowered are totally antisymmetric , and also obey _ [ c^ _ ] = 0 . [ firstjac ] next suppose that a hypothesis is made about some subset @xmath38 of the generators , @xmath39 , being significant . these are linear and quadratic constraints in the context of constraints in gravitational theory . denote the rest of the generators which form @xmath40 by @xmath41 . on now needs to check to what extent the algebraic structure in question actually complies with this assignation of significance . such checks place limitations on how generalizable some intuitions and concepts which hold for simple examples of algebraic structures are . in general , the split algebraic structure is of the form k _ k_^ = c^^_^k_^ + c^_^ h _ , [ lie - split-1 ] k _ h _ = c^^_k_^ + c^^_h_^ , [ lie - split-2 ] h _ h_^ = c^_^ k _ + c^^_^h_^ . [ lie - split-3 ] denote the second to fifth right - hand - side terms by 1 ) to 4 ) . 1 ) and 4 ) being zero are clearly subalgebra closure conditions . 2 ) and 3 ) are ` interactions between ' @xmath40 and @xmath38 . the following cases of this are then realized in this article . \i ) _ direct product _ if 1 ) to 4 ) are zero , then @xmath42 : the _ direct product group _ : the cartesian product of the sets of @xmath38 and @xmath40 , with group operation @xmath43 . in this case , one can trivially make use of the representation theory of @xmath38 and @xmath40 to build that of @xmath44 . \ii ) _ semi - direct product _ if 2 ) alone is nonzero , then @xmath45 . now , @xmath46 is given by @xmath47 for @xmath48 , @xmath49 a subgroup of @xmath14 and @xmath50 a group homomorphism . in this case , _ mackey theory _ an advanced type of induced representations method @xcite can be used to construct the representation theory of semidirect product groups . \iii ) ` _ thomas integrability _ ' . if 1 ) is nonzero , then @xmath38 is not a subalgebra : attempting to close it leads to some @xmath39 are discovered to be integrabilities . i denote this by @xmath51 . a simple example of this occurs in splitting the lorentz group s generators up into rotations and boosts : the group - theoretic underpinning @xcite of thomas precession schematically , decompose the familiar lorentz algebra bracket into j j ~j , j k ~k , k k ~k + j , the key bracket being the last one by which the boosts are not a subalgebra . thomas precession then refers to the rotation arising thus from a combination of boosts . \iv ) ` _ two - way integrability _ ' . if 1 ) and 4 ) are nonzero , neither @xmath38 nor @xmath40 are subalgebras , due to their imposing integrabilities on each other . i denote this by @xmath52 . in this case , any wishes for @xmath38 to play a significant role by itself are almost certainly dashed by the actual mathematics of the algebraic structure in question . one enters the set of constraints in one s possession into the brackets in use to form a constraint algebraic structure . this may enlarge one s set of constraints , or cause of one to adopt a distinct bracket . if inconsistency is evaded , the eventual output is an algebraic structure for all of a theory s constraints , symbolically _ _ ^ _ = c^^_^_^ . in some cases , the @xmath53 are a lie algebra s structure constants , whereas in other cases in particular for gr , they are a lie algebroid s structure functions . see appendix [ algebroids ] for a bit more about algebroids . for now , we note that the preceding sec s split classification extends to algebroids as well . all of the above carries over to supersymmetric versions as well . examples of ii ) relational mechanics provide a number of examples of semi - direct product constraint algebras . there are versions with constraints in correspondence with each of the following : the euclidean group @xmath54 , the similarity group @xmath55 , the equiareal , ( or equivoluminal or higher-@xmath56 counterpart ) group @xmath57 , and the affine group @xmath58 . @xmath59 correspond to the zero total momentum constraint @xmath60 , @xmath61 to the zero total angular momentum constraint @xmath62 ( or arbitrary-@xmath56 generalization thereof ) , and @xmath63 to the zero total dilational momentum constraint @xmath64 . [ we are here using @xmath65 as an index running over the particle labels . ] @xmath66 corresponds to @xmath67 alongside zero total shear momentum constraints and zero total ` procrustes stretch ' ( @xmath56-volume preserving stretch ) momentum constraints @xcite , whereas @xmath68 corresponds to all of these alongside @xmath69 . in each case , a suitably compatible form of quadratic energy constraint @xmath13 can be adjoined , but the preceding structures persist as subalgebras of first - class linear constraints . i.e. f ~f , e ~0 , e ~0 . the first bracket indicates a closed subalgebra , the second that @xmath13 is a good @xmath70 object ( a scalar ) , and the third indicates that @xmath13 forms its own subalgebra rather than implying any of the @xmath70 as integrabilities . thus adjunction of @xmath13 is a first example of i ) further examples of i ) include a further relational mechanics a theory @xcite with @xmath71 invariance which models the absense of absolute axes of rotation and absolute scale whilst retaining an absolute origin provides an example of direct product constraint algebra of @xmath67 and @xmath69 . of course , the standard model also involves a direct product lie group @xmath72 , so @xmath73 is well - exemplified within conventional gauge theory . examples of iii ) at the level of constraint algebraic structures include ( as outlined in @xcite ) the relational mechanics that are based on each of the conformal group @xmath74 and on all of the supersymmetric counterparts of the groups in example ii ) . the key bracket here is an integrability of the form p ~m + d , [ k - p ] by which the conformal algebra is @xmath75 thomas . i.e. a translation and a special conformal transformation compose to give both a rotation ( ` conformal precession ' ) _ and _ an overall expansion . then another salient example of iii ) is the dirac algebroid formed by the constraints of gr @xcite , schematically _ i m_i ~m_i , _ i h ~h , h ~m . [ dirac - algebroid ] it is the last of these brackets which contains structure functions , with teitelboim pointing to the interpretational complications of this @xcite . moreover , the physics of the problem in hand requires this enlargement and complication of algebraic structure , to model the myriad of possible foliations of spacetime by spatial hypersurfaces . also note that the first of these brackets shows that the momentum constraint @xmath10 closes by itself al a subalgebraic structure ; these are first - class linear constraints corresponding to @xmath76 . it is , moreover a bona fide ( if infinite-@xmath56 ) lie algebra , of the spatial diffeomorphisms @xmath76 on the corresponding topological manifold @xmath77 . finally , the second of these brackets means that @xmath11 is a good @xmath76 object : a scalar density . example of iv ) the constraint superalgebroid of supergravity . at the level of algebraic structures , each of supersymmetry and split space - time gr can be separately envisaged as thomas - type effects . upon considering both at once , these integrabilities furthermore go in opposite directions , so the more complicated ` two - way ' integrability case arises . the schematic form of the key new relation for supergravity is s ~h ( + m + j ) . [ s - s->h ] the integrability pointing in the opposite direction remains of the schematic gr form ( [ dirac - algebroid ] ) . more details of this superalgebroid can be found at e.g. @xcite . superalgebroids are treated in outline in e.g. @xcite ; the above outline readily continues to carry over upon replacing ` lie algebra ' by ` lie superalgebra ' ; supergravity s constraint superalgebroid is probably theoretical physics most salient example of superalgebroid . example 1 ) the case in which the @xmath97 themselves are formed from the generators themselves is well - known . in this case , the association amounts to producing the centre @xmath98 of the _ universal enveloping algebra _ @xmath99 of @xmath33 . this is named thus due to its encapsulation of features universal to all representations of the @xmath33 in question . in this case , the resulting @xmath97 are known as _ casimirs _ these play a prominent role in representation theory , with @xmath100 s total angular momentum operator @xmath101 being the best - known such . in this example of association procedure , each given lie algebra leads to a corresponding algebra of casimirs . . suppose @xmath17 commutes solely with a set of @xmath3 which is not closed : it does not include some of the @xmath102\|}$ ] . however , the jacobi identity with one @xmath103 and two @xmath104 as entries and making two uses of ( [ basic - obs ] ) gives a _ poset _ ( _ partially ordered set _ ) is a set @xmath109 alongside an ordering relation @xmath110 which is reflexive , antisymmetric and transitive . this is less stringent than the more familiar total ordering , for which every two elements have to be related ; this permits a linear ranking representation , by which total orderings are also known as _ posets elements need not all be related , by which posets in general contain multiple branching chains . this is representable using a _ hasse diagram _ ; all the figures in the current paper are based on hasse diagrams . an element 1 of @xmath111 is a _ unit _ if @xmath112 , @xmath113 , and an element 0 of @xmath111 is a _ null element _ if @xmath112 , @xmath114 . a lattice possessing these is called a _ bounded lattice _ ; this applies to all examples in the current article . subset inclusion @xmath115 is one common realization of @xmath110 . this carries over to the case of sets with further structures , in which case one is dealing with subspace inclusion . in this setting , intersection of two subspaces plays the role of meet and smallest subspace containing a pair of spaces that of their join . e.g. for subgroups of a group @xmath14 , the intersection of two subgroups is also a subgroup , and the join of two subgroups is defined as the subgroup generated by their union . 0 = i d : the trivial group , and 1 = @xmath14 itself . constraint subalgebraic structures of an algebraic structure @xmath116 follow suit in this regard , with 0 = i d : the trivial algebra , and 1 = @xmath116 itself . on the other hand , for observables subalgebraic structures of an observables algebraic structure @xmath117 associated a with constraint algebraic structure @xmath116 , each constraint present in the latter induces an extra relation in the former . hence the association map @xmath118 is an order - reversing lattice morphism @xmath119 . this can be recast as order - preserving by use of @xmath120 in place of @xmath115 . @xmath117 still possesses a 0 and a 1 , with , by the order - reversing , @xmath121 the space @xmath122 of unconstrained observables @xmath6 and @xmath123 the space @xmath124 maximally constrained i.e . observables @xmath5 . hence the lattice of observables is also a bounded lattice . note that @xmath118 maps all constraint subalgebraic structures to observables algebraic substructures , since one can always pose the problem of what functions commute with a given constraint subalgebraic structure . adding constraints can never increase the freedom in the observables algebraic structure , but on occasion it may have no effect if the constraint being added imposes no further restriction on the allowed functional form of the observables . e.g. an angle quantity is already a ratio and a combination of dot products . i.e. @xmath118 is capable of being many to one , as occurs in the example of conformal relational mechanics . in such a case , @xmath118 is a coarse - graining of the ordering information in moving between the two lattices ; also note that in all the other examples considered in this article , @xmath118 is 1 : 1 and fully preserving of the ordering information . see @xcite for a bit more about lattices with basic and theoretical physics examples , and @xcite for somewhat more theory . with many conceptually and technically interesting properties of lattices having been worked out , it will be very interesting to see which of these apply to lattices of a - observables . kucha , time and interpretations of quantum gravity " , in _ proceedings of the 4th canadian conference on general relativity and relativistic astrophysics _ ed . g. kunstatter , d. vincent and j. williams ( world scientific , singapore , 1992 ) , reprinted as int d20 * 3 ( 2011 ) . canonical quantum gravity and the problem of time " , in _ integrable systems , quantum groups and quantum field theories _ ed . ibort and m.a . rodrguez ( kluwer , dordrecht 1993 ) , gr - qc/9210011 . kucha , canonical quantum gravity " , in _ general relativity and gravitation 1992 _ , ed gleiser , c.n . kozamah and o.m . moreschi m ( institute of physics publishing , bristol 1993 ) , gr - qc/9304012 . c. rovelli , `` is there incompatibility between the ways time is treated in general relativity and in standard quantum mechanics ? '' , p. 126 in _ conceptual problems of quantum gravity _ ed . a. ashtekar and j. stachel ( birkhuser , boston , 1991 ) . e. anderson , the problem of time and quantum cosmology in the relational particle mechanics arena " , arxiv:1111.1472 . e. anderson , problem of time in quantum gravity " , annalen der physik , * 524 * 757 ( 2012 ) , arxiv:1206.2403 . e. anderson , where to apply relationalism " , write - up of invited seminar at the conferences at fredericton 2014 , arxiv:1411.4316 . r. gilmore , _ lie groups , lie algebras , and some of their applications _ ( dover , new york 2006 ) . v. gorbounov , f. malikov and v. schechtman , gerbes of chiral differential operators . iii " , in _ the orbit method in geometry and physics : in honor of a.a . , ed . c. duval , l. guieu , v. ovsienko ( a.m.s . , providence , rhode island 2001 ) .	the observables notion is shown to apply only to a limited range of theories . relational mechanics , slightly inhomogeneous cosmology and supergravity are used as examples that require further notions of observables . a suitably general notion of a - observables is then given to cover all of these cases . ` a ' here stands for ` algebraic substructure ' ; a - observables can be defined by association with each closed algebraic substructure of a theory s constraints . both constrained algebraic structures and associated notions of a - observables form bounded lattices . * edward anderson * _ damtp , centre for mathematical sciences , wilberforce road , cambridge cb3 owa . _
understanding how galaxies formed and evolved is a central challenge of modern astronomy . in the past decade , spectacular progress in instrumentation and observing techniques has enabled major advances as the high - redshift universe has been opened up for research . our current view has been largely influenced by the discovery of an abundant population of actively star - forming galaxies at redshifts @xmath29 selected by the efficient lyman - break technique @xcite . these lyman - break galaxies ( lbgs ) are among the best - studied classes of high - redshift objects : large samples have been spectroscopically confirmed at @xmath29 ( e.g. * ? ? ? * ) and extensive investigations have focussed notably on their stellar populations , star formation histories , chemical abundances , kinematics , and clustering properties ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? lbgs dominate the ultraviolet ( uv ) luminosity density at high redshift and their number density is comparable to that of @xmath30 galaxies locally , making them a major constituent of the early universe . in the context of the widely favoured hierarchical scenarios of galaxy formation , they are thought to be the progenitors of present - day massive galaxies in groups and clusters ( e.g. * ? ? ? yet , by construction the lyman - break technique relies on a strong lyman discontinuity in the rest - frame far - uv and is necessarily biased towards relatively unobscured galaxies with intense recent or on - going star formation activity . the typical stellar ages derived for lbgs are of a few @xmath31 with moderate extinction @xmath32 @xcite . older and more quiescent systems at @xmath33 that have formed the bulk of their stars at @xmath34 or , alternatively , more obscured galaxies may have escaped detection in optical surveys ( e.g. * ? ? ? * ) . with the advent of @xmath35-class telescopes and the development of sensitive near - infrared ( nir ) instruments equipped with large - format detectors , it has become possible to extend very deep surveys to longer wavelengths and access the rest - frame optical emission of sources at @xmath36 . compared to the rest - frame uv , the rest - frame optical spectral energy distribution ( sed ) of galaxies is less affected by the light of young massive stars and by dust extinction , and better traces longer - lived stars that dominate the stellar mass . colour criteria involving nir bandpasses can now be applied to identify new populations at @xmath37 . the long - standing debate over the formation of massive early - type galaxies has driven considerable interest in the discovery of , and subsequent systematic searches for , red objects such as the `` extremely red objects '' ( eros ) generally defined by @xmath38 or @xmath39 . in our own faint infrared extragalactic survey ( fires ; * ? ? ? * ) , based on very deep nir imaging of the hubble deep field south ( hdf - s ) and of the field around the @xmath5 cluster @xmath4 , we have identified significant numbers of fairly bright ( @xmath40 ) candidate @xmath1 galaxies selected from their @xmath0 colours . analysis of the hdf - s sample suggests that this population makes a comparable contribution to the stellar mass density at @xmath33 as lbgs @xcite and hence may be a substantial component in terms of stellar mass . remarkably , there are far fewer such bright red objects per unit area in the hdf - north ( e.g. @xcite ; see also @xcite ) . cosmic variance may however largely explain these differences since both hubble deep fields are small and the @xmath0 population may be strongly clustered @xcite . candidate high - redshift galaxies with unusually red @xmath28 colours have been reported by other authors as well ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? ? * ; * ? ? ? however , the focus has been on those objects with the most extreme colours , which are rarer and mostly at the faintest magnitudes . an immediate question is what causes the red colours of these objects . very red nir colours can be produced by an evolved population at @xmath41 due to the stellar photospheric balmer/4000 break redshifted into the @xmath42 band and beyond or by high levels of extinction in galaxies possibly at lower redshifts . the presence of a highly obscured active galactic nucleus ( agn ) is another possibility ( e.g. , * ? ? ? * ; * ? ? ? while it appears that a combination of stellar population aging and extinction effects is required to explain the observed colours ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , the results so far remain inconclusive due to the scarcity of systematic studies of large samples and to insufficient observational constraints . in this paper , we take advantage of our fires data set to address the above issue from the @xmath12 broad - band seds of 34 objects with @xmath0 selected from both fields surveyed . the addition of the @xmath4 objects triples the original hdf - s sample . we focus on the ensemble properties of the @xmath28 red galaxies . @xcite present results of our follow - up optical and nir spectroscopy of a bright sub - sample in the @xmath4 field , which are particularly relevant to this work . we briefly describe the data in [ sect - data ] . we discuss the selection criteria applied to construct the @xmath28 selected samples in [ sect - sample ] . in [ sect - prop ] , we analyze the properties of the seds to investigate the stellar populations and dust obscuration . we model the seds using evolutionary synthesis in [ sect - models ] to constrain quantitatively the stellar ages and masses , the extinction , and the star formation rates . in [ sect - discussion ] , we interpret our results and compare the derived properties to those of lbgs and of @xmath43-band - selected objects at similar redshifts in the fires fields . we summarize the paper and main implications in [ sect - conclu ] . for convenience , we adopt the term `` distant red galaxies , '' or `` drgs , '' introduced by @xcite to designate candidate @xmath1 galaxies with @xmath0 colours . all magnitudes are expressed in the vega - based photometric system except when `` ab '' indicates reference to the ab system @xcite . throughout , we assume a @xmath44-dominated cosmology with @xmath45 , @xmath46 , and @xmath47 . the data were obtained as part of fires @xcite , a public nir survey of the hdf - s and @xmath4 fields carried out at the european southern observatory ( eso ) very large telescope ( vlt ) . the observations , data reduction , and source catalogues are described in detail by @xcite for hdf - s and n.m . frster schreiber et al . ( 2004 , in preparation ) for the @xmath4 field . briefly , the nir observations were taken in the @xmath48 , @xmath49 , and @xmath27 bands with the vlt infrared spectrograph and array camera ( isaac ; * ? ? ? * ) . for hdf - s , a total of 103 h integration time was spent in a single @xmath2 pointing covering the hubble space telescope ( hst ) wfpc2 main field . we complemented the nir data with the publicly available deep optical wfpc2 imaging in the @xmath50 , @xmath51 , @xmath52 , and @xmath53 bands @xcite . for the @xmath4 field , 77 h of isaac integration time was obtained in a @xmath3 mosaic of four pointings . at optical wavelengths , we used existing hst wfpc2 mosaics in the @xmath52 and @xmath53 bands @xcite and collected additional imaging in the bessel @xmath54 , @xmath55 , and @xmath26 bands with the vlt fors1 instrument . with hdf - s , fires achieves the currently deepest ground - based nir imaging and the deepest @xmath43-band map to date , even from space . with the @xmath4 field , the area surveyed is nearly five times wider , down to @xmath56 brighter magnitudes . the sources were detected in the @xmath27 band using version 2.2.2 of the sextractor software @xcite . for consistent photometry across all bands , the fluxes were measured on the maps convolved to a common spatial resolution of full - width at half maximum fwhm = 0.48 for hdf - s and 0.69for the @xmath4 field , matching to the map of poorest seeing in each data set ( @xmath49 and @xmath54 band , respectively ) . from the curve - of - growth analysis of the average stellar profile in each band , the fractional enclosed flux agrees to within 3% at diameters @xmath57 for hdf - s and 2% at @xmath58 for the @xmath4 field , where the diameters range from the smallest ( @xmath59 of the matched psfs ) to the largest apertures relevant for the colour measurements . colours and seds used in this work are based on measurements in custom isophotal apertures defined from the detection map , hereafter referred to as `` colour '' fluxes or magnitudes . total magnitudes in the @xmath27 band were computed in apertures based on auto - scaling apertures @xcite for isolated sources and adapted isophotal apertures for blended sources . the photometric uncertainties were derived empirically from simulations on the maps . the total @xmath60 limiting magnitudes for point sources are @xmath61 for hdf - s and 23.1 mag for the @xmath4 field . we will refer to the objects by their number in catalogue versions v3.0e for hdf - s and v3.1b for the @xmath4 field , prefixed with `` @xmath62 '' and `` @xmath63 . '' wherever appropriate , we corrected for the galactic extinction along the hdf - s and @xmath4 lines of sight ( @xmath64 and 0.11 mag ; * ? ? ? * ) . we also accounted for the gravitational lensing by the foreground @xmath4cluster based on the lensing model constructed by @xcite from weak lensing analysis . the average magnification factor for all @xmath27-band selected objects considered in this paper is 1.18 ( or 0.18 mag ) . we emphasize that the colours and sed shapes are unaffected since lensing is achromatic . photometric redshifts @xmath65 were determined by applying an algorithm involving linear combinations of redshifted spectral templates of galaxies of various types , as described by @xcite . monte - carlo simulations were used to estimate the @xmath65 errors accounting for uncertainties in the fluxes as well as template mismatch , and reflecting not only the confidence interval around the best solution but also the presence of secondary solutions . comparison with available spectroscopic redshifts @xmath66 imply an accuracy of @xmath67 for both fields . for the @xmath41 regime of interest in this work ( see [ sub - sample ] ) , @xmath68 based on the 20 @xmath27-band selected sources with @xmath66 determinations . to identify red candidate high - redshift galaxies , or drgs , we applied the criterion @xmath69 discussed by @xcite . this criterion was specifically designed to target the stellar photospheric balmer and 4000 breaks at redshifts above 2 , which are the most prominent features in the rest - frame optical continuum sed of galaxies . the balmer discontinuity at 3650 is strongest in a - type stars and , more generally , is indicative of stellar ages @xmath70 . at older ages , the 4000 break characteristic of cooler stars with types later than about g0 , and strongest in giants and supergiants , becomes dominant . it is due to the sudden onset of metallic and molecular opacity bluewards of 4000 produced notably by h@xmath71k , , , and cn lines . based on synthetic spectral templates computed with the @xcite models ( see [ sect - models ] ) , unextincted single - age stellar populations older than @xmath72 or passively evolving galaxies formed at @xmath73 satisfy the criterion at @xmath74 , where the balmer/4000 break is redshifted into the nir regime . galaxies forming stars at a constant rate have a @xmath28 colour redder than 2.3 mag only if they are @xmath75 old and moderately obscured at @xmath76 , or younger and highly obscured with @xmath77 at @xmath78 . as argued by @xcite , our follow - up spectroscopy is consistent with the @xmath0 criterion selecting @xmath1 galaxies with a high efficiency of @xmath79 . the @xmath28 colour cut at 2.3 mag is slightly bluer than used in other recent studies ( in particular * ? ? ? * ; * ? ? ? our objective is not to focus on the rare extreme objects but to allow inclusion of systems with rest - frame colours characteristic of normal nearby galaxies , which may be missed by rest - frame uv - selected surveys . observed colours of @xmath69 correspond to rest - frame colours of @xmath80 or @xmath81 at @xmath82 . these rest - frame colours encompass those of the majority of present - day luminous galaxies , from elliptical to intermediate - type spiral galaxies ( e.g. * ? ? ? * ; * ? ? ? as discussed by @xcite , contamination by foreground cool stars is expected to be negligible since most known m , l , and t dwarfs have @xmath28 colours bluer than 2.3 mag ; others such as extreme carbon stars or mira variables that could be redder are rare and unlikely to be present in pencil beam surveys at high galactic latitudes such as our fires fields ( see also * ? ? ? * ; * ? ? ? * ) . from nir spectroscopy of six drgs in the @xmath4 field and the chandra deep field south ( cdf - s ) , @xcite argued that strong emission lines in the @xmath27 band are unlikely to account for the observed red colours of most of the population . the main contaminants are more likely very dusty ( @xmath77 ) galaxies at @xmath83 , probably at the @xmath84 level @xcite . another potential concern is that samples selected with the @xmath0 criterion suffer from contamination by agn . however , the majority of the bright drgs we identified in the hdf - s and @xmath4 fields ( [ sub - sample ] ) are well resolved in our @xmath27-band imaging data , implying that their nir broad - band fluxes are not strongly influenced by continuum emission from an agn . emission lines from type ii agn could also influence the nir colours . the rest - frame optical spectra of @xcite show that this effect is minor for the confirmed type ii agn in our samples , and we are pursuing a nir spectroscopy program to confirm this result . of type ii agn in the spectroscopic sample @xcite is likely to result from the spectroscopic selection bias towards galaxies that are bright in the observer s optical and with bright emission lines . ] for the moment , we conclude it is unlikely that emission from agn affects the nir fluxes and colours in a substantial way . for the purpose of sed analysis and modeling , our goal was to construct a sample of @xmath28 selected candidate @xmath41 galaxies with robust nir colours , reliable seds , and spanning a redshift range that would allow consistent comparisons with the well - studied lbgs . we required that the @xmath69 objects have a signal - to - noise ratio @xmath85 on their @xmath27-band colour flux and a minimum of 30% of the total exposure time ( or relative weight @xmath86 ) in all bands . mosaic , the nir isaac fields have somewhat different total integration times and the highest overall weights are reached in the small overlapping regions . to ensure that the selection area covers fairly the mosaic , we considered the weights for the nir data normalized to the total exposure time in each of the isaac fields individually . for sources in the overlapping regions , we required that @xmath86 in at least one of the field . ] this yielded initial samples of 14 objects at @xmath87 in hdf - s and 31 at @xmath88 in the @xmath4 field . the effective sky - plane areas are 4.48 and @xmath89 , respectively . the photometric redshifts of the selected sources lie in the range @xmath90 . spectroscopic redshifts are available for four of them in the @xmath4 field ( @xcite ; s. wuyts et al . 2004 , in preparation ) . three have @xmath66 between 2.42 and 2.61 , and their @xmath65 agrees with @xmath66 within @xmath91 . the fourth has @xmath92 and turned out to be an interloper at @xmath93 . we note that the large uncertainty towards low redshifts for this source reflects the presence of a long tail of @xmath65 solutions with comparable likelihood . figure [ fig - z ] compares the @xmath65 s with @xmath66 s and shows the redshift distributions of the samples , where we used @xmath66 whenever available instead of @xmath65 . for the present study , we focused on the @xmath28 selected sources with redshift between 2 and 3.5 . we chose this interval because we are primarily interested in galaxies at @xmath41 and it matches well the range for lbgs selected from space - based hst wfpc2 data . ground - based selected lbgs typically lie at somewhat higher redshift with less overlap with the range covered by our drg samples . this is mainly because the selection criteria rely on a different set of filters with , notably , a redder `` @xmath54 '' bandpass ( for a discussion , see , e.g. , * ? ? ? the samples finally adopted , with @xmath0 and @xmath9 , consist of 11 objects at @xmath87 in hdf - s and 23 at @xmath88 in the @xmath4 field . the redshift distribution for the hdf - s ( @xmath4 field ) sample has a mean @xmath94 ( 2.44 ) , median @xmath95 ( 2.42 ) , and dispersion @xmath96 ( 0.30 ) . these values are computed using the @xmath66 whenever available , and are essentially unchanged when considering only the @xmath65 s ( the largest difference is @xmath97 for the @xmath4 field ) . there are four objects just below the @xmath98 limit whose @xmath91 @xmath65 error would allow a @xmath1 and , conversely , five objects just above whose @xmath91 @xmath65 error would allow @xmath99 ; we verified that these small sample variations would not alter the conclusions of our sed analysis and modeling . table [ tab - sample ] gives relevant characteristics of the adopted drg samples . we note that not all the @xmath4 field objects for which spectroscopy is presented by @xcite are included in our sample . the initial selection for the spectroscopic follow - up was based on a preliminary version of the reduced data and photometry . in the final photometric catalogue ( v3.1b ) , the two galaxies with spectral signatures indicating an agn are slightly bluer than the @xmath28 cutoff ( @xmath100 and @xmath101 , numbered 1195 and 1458 in * ? ? ? * ) . the bright @xmath102 and red @xmath103 object @xmath104 ( number 184 in * ? ? ? * ) satisfies all criteria except for the minimum integration time in the nir bands , with relative weights @xmath105 , @xmath106 , and @xmath107 . however , it has a spectroscopic redshift ( included in figure [ fig - z]_a _ ) and , at @xmath108 , lies in the range considered for our analysis . @xmath104 constitutes a valuable addition that we will discuss where relevant but keep out of the formal sample . figure [ fig - numcounts ] shows the raw cumulative number counts per unit area as a function of total @xmath27 magnitude of the drgs , for both cases of @xmath41 and of no redshift restriction . to assess the variations of surface density with limiting magnitude and between the two fields , the counts are extended to the @xmath60 @xmath109 limits for point sources of 23.8 mag in hdf - s and 23.1 mag in the @xmath4 field . we did not account here for the lensing by the @xmath4 cluster . the corrections would be small , with an average dimming by @xmath110 and a mean reduction by @xmath111 from image- to source - plane area for the redshift range covered by the sources . these effects are opposite so that to first order , the surface density to a given apparent magnitude is insensitive to lensing ( the exact effect depends on the slope of the luminosity function , which is unknown ) . per unit area , the @xmath4field is richer in sources brighter than @xmath112 while hdf - s has higher surface densities for fainter limits down to 23 mag ( where the completeness levels still are @xmath113 for both fields ) . these differences are not due to the lensing magnification by @xmath4 . for @xmath41 , there are five drgs at @xmath114 in the @xmath4 field and none in hdf - s ; at fainter limits , hdf - s has higher surface densities by a factor of two on average . the redshift cut has little impact : without it , there are seven drgs at @xmath114 in the @xmath4field and still none in hdf - s , and the surface densities at fainter limits are 1.7 times higher on average in hdf - s . the two fields cover small areas and cosmic variance could easily account for a large part of the differences in the surface densities , the more so in view of the possible strong clustering of the drg population @xcite . this might also explain the deficiency in hdf - north , with similar survey area as hdf - s , where we find no source with @xmath0 colours at magnitudes brighter than @xmath115 using ground - based photometry ( from * ? ? ? band because the corresponding f110w bandpass of nicmos onboard hst is significantly bluer and about twice as broad as the @xmath42 or @xmath48 bandpasses ( the colour term is negligible between the latter two ) . consequently , f110w photometry is not useful to select galaxies red in their @xmath28 colours . ] figure [ fig - numcounts_col ] plots the raw cumulative counts as a function of @xmath28 cutoff for all @xmath27-band selected sources at @xmath87 in hdf - s and @xmath116 in the @xmath4 field ( the magnitude limits ensuring @xmath85 on the @xmath27-band colour fluxes of the drgs ) . for @xmath9 , the drg samples are not strongly sensitive to the exact colour cut : varying the criterion between 2.2 and 2.4 mag implies @xmath117 objects for hdf - s and @xmath118 for the @xmath4 field , or @xmath119 variations . imposing no redshift limits , the same colour cut variations result in @xmath120 objects for hdf - s and @xmath121 for the @xmath4 field ; the fraction of galaxies at @xmath99 increases rapidly for @xmath28 colours bluer than 2.3 mag . the median @xmath91 uncertainty on the @xmath28 colours among the adopted drg samples is 0.13 mag and all those with uncertainties larger than 0.2 mag ( the maximum is 0.43 mag ) have colours redder than 3.0 mag . we also constructed a sample of lbgs taken in hdf - s . we used the original hdf - s hst wfpc2 photometric catalogue from @xcite and selected the lbgs using the criteria introduced by @xcite . of the 139 selected lbgs at @xmath122 , 91 are cross - identified in our fires @xmath27-band selected catalogue and have @xmath86 in our full set of maps . all lbgs in this sub - sample are brighter than @xmath123 . to make up the comparison lbg sample , we then selected those satisfying the same @xmath109 limit as for the hdf - s drgs and with redshift @xmath9 . the requirement on the @xmath27-band magnitudes ensures that the selected lbgs have reliable nir photometry and rest - frame @xmath26-band luminosities similar to those of the selected drgs . the adopted sample consists of 33 lbgs at @xmath87 . their redshift distribution has @xmath124 , @xmath125 , and @xmath96 . the mean and median redshifts are close to those of the @xmath9 drgs , especially from hdf - s . the average ( and median ) @xmath28 colour is 1.68 mag with dispersion @xmath126 , and only two of these lbgs are redder than @xmath127 . for this lbg sample , we will hereafter use the corresponding @xmath128 photometry from the fires hdf - s catalogue ( v3.0e ) . although selected at different wavelengths , the use of seds from the same photometric catalogue ensures uniformity in the data sets for both drgs and lbgs and more directly comparable results in the analysis and modeling . the requirement that our lbg sample be also detected in the @xmath27 band makes our selection slightly different than the classical purely optical selection . we did not construct a similar lbg sample for the @xmath4 field because of its ground - based @xmath129 data , which would select lbgs in a somewhat higher redshift range than spanned by the drg samples . the large numbers of drgs that we identified from @xmath69 in the fires survey suggest they may represent a significant component of the high - redshift galaxy population @xcite . it is thus important to understand their nature , in particular their stellar populations and dust content . before undertaking the detailed theoretical modeling , we here examine the broad - band properties of the sample to assess the relative importance of aging and extinction in a qualitative but less model - dependent manner . the observed seds of drgs in both fires fields exhibit a wide variety in their shapes , especially in the optical regime . figure [ fig - seds ] presents the seds of selected representative objects in each of the hdf - s and @xmath4 field . the galaxies are sorted according to increasing @xmath130 colour , and the seds are normalized to the flux density in the @xmath27 band ( the overplotted sed fits will be discussed in [ sect - models ] ) . at one extreme , some of the drgs appear overall very red with little if any flux in the optical bands . at the other , some have blue observed optical colours indicative of on - going star formation activity . most exhibit a rather abrupt transition redwards of the @xmath48 band reminiscent of a spectral break in the continuum . a few show a rather smooth and gradual increase in flux with wavelength suggestive of a featureless continuum reddened by dust obscuration . the rest - frame optical colours of the drgs lie within the range covered by present - day normal galaxies , in striking contrast with the much bluer lbgs . figure [ fig - rfseds ] shows the rest - frame seds of our samples of drgs and lbgs at @xmath9 . all seds are dereddened for galactic extinction . we accounted for the average intergalactic attenuation due to line blanketing by ly@xmath131 ( @xmath132 ) and higher - order lyman lines ( @xmath133 ) following the prescriptions of @xcite . the seds are compared with the empirical uv - optical template spectra of nearby galaxies of types e , sbc , scd , and i m from @xcite . the spectra shown here were extended beyond their original @xmath134 coverage by @xcite using the evolutionary synthesis code of ( * ? ? ? * gissel98 version ) with appropriate model assumptions . the seds and spectral templates are normalized to a flux density @xmath135 of unity at @xmath136 using linear interpolation between the adjacent rest - frame data points . at @xmath137 , the seds of all drgs fall within the envelope of normal galaxies . none of the drgs has rest - frame optical colours as red as the earliest - type nearby ellipticals , perhaps not surprisingly so because the universe is only a few times @xmath138 old at @xmath9 . at shorter wavelengths , the seds span a wide range in rest - frame uv slopes encompassing those of normal galaxies and extending to much redder slopes , suggesting a range in level of star formation activity and in interstellar extinction . in contrast , the lbg seds exhibit less scatter especially in the uv and are systematically bluer than late - type spirals , as first demonstrated by @xcite . these authors showed that the seds of their hdf - n lbgs fit well in the envelope defined by the local starburst templates of @xcite and are consistent with the presence of a young and modestly obscured stellar population . figure [ fig - rfseds_sim ] illustrates the effects of extinction , where we used the im - type template of @xcite and applied the extinction law of @xcite assuming a simple uniform foreground screen of obscuring dust . the template seds inherently include some intrinsic reddening so that our simulations represent additional amounts of extinction . interestingly enough , the i m template combined with added extinction of @xmath139 between 0 and 3 mag brackets the range of rest - frame sed shapes of our drg sample . large extinction offers a natural interpretation for the many objects with very red rest - frame uv emission and consequently could also account in part for the red rest - frame optical colours . figures [ fig - rfseds ] and [ fig - rfseds_sim ] do not allow us to assess the relative importance of evolved stellar populations and extinction effects in producing the observed seds of drgs . however , they demonstrate clearly that drgs and lbgs are distinct in their rest - frame uv - optical seds , presumably because of very different dominant stellar populations and/or extinction levels . in figure [ fig - rfseds ] , the seds of the three spectroscopically - confirmed drgs among our sample along with @xmath104 ( see [ sub - sample ] ) are plotted with different symbols . the comparison indicates that the galaxies in the spectroscopic sample do not deviate strongly in their rest - frame seds from the full photometric sample . on the other hand , it also shows that the spectroscopic sample does not probe the entire range of properties seen for the ensemble of drgs . the wavelength sampling and resolution of our data are too coarse to delineate accurately a balmer/4000 break . even when strongest , the break represents a moderate flux jump by a factor of @xmath140 , a subtle feature compared to the lyman break at 912 . for the brightest drg of the spectroscopic sample , the presence of a break is confirmed by the detection of a significant drop in continuum level in its nir spectrum collected with nirspec at the keck telescope @xcite . here , we examine the observed nir properties of our ensemble of drgs using diagnostic colour - colour diagrams , which allow us to establish that a break is required to explain the colours of an important fraction of them . specifically , we test whether the red observed nir colours of drgs ( 1 ) can be reconciled with a very young ( @xmath141 ) stellar population that is highly obscured by dust , or if ( 2 ) the presence of a more evolved ( @xmath142 ) stellar population whose sed has a prominent balmer/4000break is necessary . in exploring various colour combinations , we relied on the sharpness of the balmer/4000 break versus the smoother wavelength dependence of dust attenuation . extinction makes all colours redder while a prominent break produces a larger contrast between colours because of the steep drop in flux at the break . as a stellar population ages , the continuum slope intrinsically reddens because cooler stars dominate the optical light but the break remains comparatively sharp . for redshifted populations , these differences in colour contrast will be reflected in the variations with @xmath143 in colour - colour space . we found that @xmath15 versus @xmath14 and @xmath14 versus @xmath13 provide the most sensitive diagnostic diagrams . the use of these adjacent bandpasses traces as closely as possible the balmer/4000 break in the range @xmath144 while minimizing reddening due to dust and to cool dominant stars . figure [ fig - nircol ] compares in these diagrams the colour distributions of the drgs with the redshift evolution at @xmath145 of selected model seds : 250 myr and 1 gyr old single - age stellar populations ( ssps ) , a passively evolving population formed at @xmath146 , and populations with constant star formation rate ( csf ) and ages of 10 myr , 100 myr , and 1 gyr . we computed the synthetic spectra and colours using the codes of @xcite and @xcite as for the sed modeling and with the same standard assumptions ( see [ sect - models ] ) . we also plot tracks for the 100 myr old csf model with @xmath139 in the range @xmath147 , applying the extinction law of @xcite and a uniform foreground screen geometry . the model tracks in the top panels of figure [ fig - nircol ] show the expected trend that the older the age , the larger the amplitude of the colour variations with @xmath143 as the stronger break moves redwards across the successive bandpasses . the effect is maximized for ssps or passively evolving populations , where the break is not diluted by the light from younger stars with shallower or no absorption features and with bluer continuum . csf models are consistently bluer in all colours and occupy narrower regions than ssps of the same age . extinction effects shift the tracks along paths that run diagonally across the diagrams but preserve their amplitude at fixed age . as shown schematically in the bottom panels , the locus of csf models with varying extinction gets wider with older ages , because of the stronger break . the effects of aging and extinction can be distinguished in that the loci of dusty model tracks broaden with increasing age in the direction perpendicular to the extinction path , the key feature we here exploit . the immediate result from these diagrams is that the colours of the ensemble of drgs can not be reconciled with very young and highly obscured populations . dusty csf models need the broadening due to significant aging to better match the observed spread perpendicular to the extinction vector . to quantify the fraction of drgs not overlapping with the dusty models , figure [ fig - nircol_proj ] presents histograms where the colour distributions have been rebinned as a function of the deviation in @xmath14 and @xmath13 around the line best representing the dusty 100 myr old csf tracks ( using robust linear regression ) . the maximum extent of the tracks about this line is indicated in the plots . the central loci and maximum widths for the 10 myr and 1 gyr old dusty csf models are plotted as well . about two - thirds of the drg sample has colours which are inconsistent with those of a 10 myr old obscured star - forming population . the fraction drops to @xmath148 for an age of 100 myr , and @xmath149 for 1 gyr ( depending on the colour considered ) . interestingly , the @xmath15 versus @xmath14 diagram shows that if drgs are actively star - forming , then the @xmath0 colour cut will miss even those with a strong optical break unless they are highly obscured . in other words , the criterion preferentially selects not only for the most evolved star - forming systems but for the most obscured ones as well . the need for high extinction is alleviated for scenarios where the star formation rate is declining with time ( as we discuss further in [ sect - models ] ) , but the @xmath28 colour cut still selects for evolved galaxies . figures [ fig - nircol ] and [ fig - nircol_proj ] demonstrate that the red nir colours of the ensemble of drgs can not be attributed to extinction effects alone . for a significant fraction of the objects , the colours require the presence of a strong balmer/4000 break produced by stars with ages @xmath150 . in this section , we describe our model ingredients and fitting methodology . we present the results for our set of `` standard '' parameters with fixed initial mass function ( imf ) , metallicity , extinction law , and two choices of star formation history ( sfh ) . we also test the sensitivity of the results to changes in the sfh , metallicity , and extinction law . we modeled the broad - band seds of our drg sample following the approach adopted in recent similar studies of high redshift galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? we fitted redshifted synthetic spectra of stellar populations to the seds to constrain simultaneously the age and extinction , keeping @xmath143 fixed for each object . we derived the stellar masses ( @xmath151 ) and star formation rates ( sfrs ) by scaling those of the best - fit input model so as to reproduce the observed fluxes . we determined the stellar mass - to - light ( m / l ) ratios using the attenuated rest - frame @xmath26-band luminosities ( @xmath152 ) computed as described by ( * ? ? ? * their appendix c ) . although the extrinsic ( i.e. dust - attenuated ) luminosities from our best - fit models agree to better than 10% on average , we preferred the less model - dependent estimates of @xcite . we generated the synthetic spectral templates with the most recent version of the evolutionary synthesis code developed by g. bruzual and s. charlot @xcite . this `` bc03 '' version features improved stellar evolutionary tracks with observationally motivated prescriptions for the thermally - pulsing asymptotic giant branch phase . it also incorporates updated and newly available stellar libraries for two sets of spectral resolution and coverage . we used the lower resolution set covering @xmath153 to 160 relying on the basel 3.1 library . we selected the `` padova 1994 '' evolutionary tracks , which are preferred by @xcite over the more recent `` padova 2000 '' tracks because the latter may be less reliable and predict a hotter red giant branch leading to worse agreement with observed galaxy colours . we used the solar metallicity set of tracks . the only constraints available for the chemical abundances of drgs come from the [ ] @xmath154 / h@xmath131 line ratio measured in two of the galaxies observed spectroscopically , which suggests high metallicities of @xmath155 @xcite . these drgs appear more metal - rich than the five lbgs at @xmath33 studied by ( * ? ? ? * @xmath156 ) and similar to the seven uv - selected star - forming `` bx / md '' objects at @xmath157 for which @xcite inferred solar , and possibly super - solar , metallicities . in all cases , however , the determinations rely on limited samples and suffer from large uncertainties . the computed grid of spectral templates covers ages between @xmath158 and @xmath159 . the sfh most appropriate for our objects is unknown and broad - band photometry alone is not effective at constraining it ( see [ sub - paramvar ] ; also , e.g. , * ? ? ? * ; * ? ? ? we considered two star formation rate parametrizations : constant in time with @xmath160 and exponentially declining in time as @xmath161 with timescale @xmath162 ( hereafter `` csf '' and `` @xmath163 '' models ) . these exact choices are somewhat arbitrary . the csf model is intended to represent the case of significant on - going or recent star formation activity . evidence for this scenario in part of the drg population includes the detection of h@xmath131 and/or [ ] @xmath164 lines in the non - agn sources of the spectroscopic sample @xcite and the blue rest - frame uv sed of several others among the full sample ( see figures [ fig - seds ] and [ fig - rfseds ] ) . the @xmath163 model allows for the possibility of quiescent systems that underwent a period of enhanced star formation in their past . we adopted a salpeter imf ( @xmath165 ) between 0.1 and 100@xmath16 . the imf is unconstrained for our objects , and a steep rise down to the lower mass cutoff likely is unrealistic in view of the turnover below 1@xmath16 inferred for the local imf ( e.g. , * ? ? ? * ; * ? ? ? the most critical impact is on the derived masses and mass - to - light ratios ( @xmath166 ) , which depend strongly on the shape and cutoff of the low - mass imf . for instance , for a fixed rest - frame @xmath26-band luminosity , the derived masses and m / l ratios would be about a factor of two lower with the @xcite or @xcite imfs ( e.g. , figure 4 of @xcite ; see also @xcite for a discussion of the effects of different imfs in their sed modeling of lbgs ) . we have used the mass of stars still alive @xmath151 instead of the total mass of stars formed @xmath167 . @xmath167 is simply the integral of the sfr and corresponds to the total gas mass consumed since the onset of star formation . @xmath151 is computed by subtracting from @xmath167 the mass returned to the ism by evolved stars via stellar winds and supernova explosions ( see * ? ? ? * for details ) . we preferred @xmath151 in our analysis because it represents the mass of the stars producing the light . for the typical ages @xmath168 of interest here and with the adopted imf , the cumulative mass - loss effects are modest and comparable for the csf and @xmath163 models , reaching @xmath169 of @xmath167 . even in the limit of an ssp where the effects are largest , they amount to a similar fraction of @xmath170 . both interstellar extinction by dust within the objects and attenuation due to intergalactic h opacity were applied to the bc03 templates while performing the fits . we explored a range of extinction with @xmath171 . higher @xmath139 values are physically possible but unlikely to be actually measured from rest - frame uv / optical data because the corresponding optical depths @xmath172 largely exceed unity . the extinction law and the geometry for the obscuring dust and emitting sources are other uncertain parameters . we adopted the law of @xcite derived empirically from observations of local uv - bright starburst galaxies under the formalism of a foreground screen of obscuring dust . as emphasized by these authors , this extinction law and model geometry may not be valid in dust - rich systems where the optical depths are large and the sources and dust are more or less mixed spatially . the average lyman line absorption by the intergalactic medium ( igm ) was accounted for based on the prescriptions of @xcite . lyman continuum absorption was approximated by setting the flux of the templates equal to zero at @xmath173 . we performed the fits using the publicly available hyperz photometric redshift code , version 1.1 @xcite . we fixed the redshift to the @xmath65 determined independently ( see [ sect - data ] ) or to the @xmath66 when available . the code was allowed to fit simultaneously the spectral template ( age ) and the extinction , for each of the sfh adopted . interstellar extinction is first applied to the synthetic spectra , which are then redshifted and further attenuated for igm absorption . the fitting algorithm relies on least - squares minimization , with the chi - squared calculated as @xmath174 ^ 2 $ ] . the sum is performed over the @xmath175 photometric bands . the template fluxes @xmath176 are computed from the reddened , redshifted , igm - absorbed template spectra using the total effective transmission functions in each bandpass . the scaling factor @xmath177 is determined from the full sed constructed from the colour fluxes , corrected by the same amount to the total aperture based on the @xmath43-band measurements ; all seds have also been corrected for galactic extinction and , for the @xmath4 field sample , for the lensing magnification prior to the fitting ( [ sect - data ] ) . we treated any flux with @xmath178 as a non - detection , with upper limit equal to the @xmath91 uncertainty . the assigned @xmath91 uncertainties and upper limits correspond to those of the flux measurements . we assumed a minimum error of 0.05 mag on the photometry to avoid the fits being driven by a few data points with very small errors . this also accounts for uncertainties in the absolute flux calibration . our results are not strongly influenced by this assumption . if we do not impose a minimum photometric error , the median of the best - fit extinction values remains the same for the two fields and two standard sfhs considered . the median @xmath151 s vary by @xmath179 , and the median sfrs by @xmath180 . the median of the best - fit ages decreases from 1.7 to 1.0 gyr for the hdf - s drgs and csf , and increases from 1.0 to 1.4 gyr for the @xmath4field drgs and the @xmath163 model , not affecting our findings that drgs harbour old stellar populations . hyperz resamples the age grid from the bc03 models from 221 to 51 fixed steps . we required that the template ages considered for the fitting do not exceed that of the universe at the redshift of each source . we emphasize that the best - fit age , which we will denote @xmath181 , is strictly speaking the time elapsed since the onset of star formation . an alternative , perhaps more meaningful definition is the sfr - weighted mean age @xmath182 corresponding more closely to the age of the stars contributing the bulk of the stellar mass . for csf models , @xmath183 . for exponentially decaying models , @xmath184 / [ 1 - \exp(-t_{\rm sf}/\tau)]$ ] , which becomes @xmath185 when @xmath186 . moreover , @xmath181 should not be over - interpreted as the age of the entire galaxy . it is important to remember that all the properties derived from the sed modeling apply only to the stellar population that dominates the fluxes in the observed wavelength range . this does not preclude , e.g. , the presence of a pre - existing stellar population that has faded out too much ( a scenario explored for lbgs by * ? ? ? * ) or of a younger population that is too obscured to be detected by our observations . we estimated the confidence intervals for individual objects from monte - carlo simulations . for each source , we performed 500 synthetic realizations of the data by varying the fluxes by an amount randomly drawn from the ( gaussian ) distribution of the measurement uncertainties ( including the minimum error of 0.05 mag where appropriate ) . for sources with only a @xmath65 determination , we also varied the redshift by drawing randomly from the associated probability distribution @xmath187 . we then repeated the fitting procedure for each realization . we determined the 68% confidence intervals from the distributions of best - fit values obtained with the simulated data . the behaviour of the confidence intervals for the drg samples studied here are very similar to those of the spectroscopic sample of @xcite and of the lbgs modeled by @xcite and @xcite , i.e. often asymmetric around the best - fit values , reflecting strong degeneracies notably between @xmath181 and @xmath139 , and implying that @xmath151 is the best constrained quantity for the majority of objects ; we refer to the above papers for a more detailed discussion . ) standard model are @xmath188 ( @xmath189 ) for the age @xmath181 , @xmath190 ( @xmath191 ) for @xmath151 , @xmath192 ( @xmath193 ) for @xmath194 , and @xmath195 ( 50% to a factor of 2 ) for the instantaneous sfr . for the hdf - s lbg sample also modeled in this work ( [ sub - drgs_lbgs ] ) , the uncertainties are @xmath196 ( @xmath197 ) for the age @xmath181 , @xmath198 ( @xmath199 ) for @xmath151 , @xmath200 ( @xmath201 ) for @xmath194 , and @xmath202 ( @xmath203 ) for the instantaneous sfr . for drgs and lbgs , the uncertainties on @xmath139 are @xmath204 with both the csf and @xmath163 models . ] the derived uncertainties are smaller or comparable to the differences in best - fit values obtained by varying the assumed sfh , metallicity , or extinction law ( see [ sub - paramvar ] ) . the main goal of this work is to characterize the ensemble properties of drgs ( from the median and the global distributions of best - fit values ) , and accounting for the uncertainties of individual sources does not alter our general conclusions . we will therefore not discuss them further , except where most relevant . both the csf and @xmath163 models provide acceptable fits to the seds of our drg samples . , for the best - fit csf models are 1.6 for the hdf - s sample and 2.1 for the @xmath4 field sample . for the @xmath163 models , the @xmath205 values are 0.6 and 1.6 , respectively . @xmath205 s higher than the expected value of unity for good fits could indicate that the models do not provide a realistic representation of the objects , the photometric uncertainties underestimate the real measurement errors , the distributions of uncertainties of the modeled properties are not gaussian , or , most likely , that a combination of the factors is at play . ] the best - fitting models for selected sources are shown in figure [ fig - seds ] . table [ tab - res_std2 ] and figure [ fig - res_std2 ] summarize the derived properties . the results are similar for the hdf - s and @xmath4 field samples . with the csf model , the median values for hdf - s ( @xmath4 field ) are @xmath206 ( 2.0 ) gyr , @xmath207 ( 2.4 ) mag , @xmath208 , @xmath209 , and instantaneous @xmath210 . with the @xmath163 model , the median age and extinction are 1.0 gyr and 0.9 mag for both fields ; the median for the other properties in hdf - s ( @xmath4 field ) are @xmath211 , @xmath212 , and @xmath213 . the median ages of the bulk of stars @xmath214 and the median time - averaged sfrs @xmath215 for the two sfhs and two fields differ less than the median @xmath181 and instantaneous sfrs , with values between 0.75 and 1.0 gyr , and 90 and @xmath216 . the dispersion among individual drgs is large for all properties . this stems partly from the diversity in seds noted in [ sect - prop ] . for instance , with a given sfh , quite different combinations of age and extinction are derived for the objects that are bluest in the rest - frame uv compared to the reddest ones . part of the ( small ) differences between the hdf - s and @xmath4 samples likely reflects the brighter @xmath27-band magnitude limit and larger proportion of the brightest sources , the redder @xmath28 colours , and the lower redshifts for the @xmath4objects ( see table [ tab - sample ] ) . the different sets of optical photometric bands may play a role too by providing somewhat different observational constraints . for the ensemble of drgs , either sfh thus implies old ages , and high stellar masses and m / l ratios . high sfrs and large amounts of extinction are derived for the csf model while the @xmath163 model leads to lower sfrs and extinction ( we comment on the validity of the adopted @xcite extinction law in view of the large extinction derived for the drgs in [ sub - paramvar_avlaw_z ] ) . similar results were obtained in our initial study of hdf - s @xcite and for the spectroscopic sample @xcite . in particular , for the two spectroscopically - confirmed non - agn drgs in the @xmath4 field , the ages are supported by estimates from the h@xmath131 equivalent widths and the stellar masses are consistent with the dynamical masses from the h@xmath131 or [ ] @xmath217 line widths . as we show in [ sub - paramvar ] , the stellar masses are the most robust properties against variations in sfh , metallicity , and extinction law ( neglecting changes in the imf ) . with our definition , the @xmath218 ratios also are because the luminosities involved are extrinsic . on the other hand , the instantaneous sfrs are very sensitive to the model assumptions , especially to the sfh ( again ignoring the imf ) . to test the sensitivity to the photometric redshift uncertainties , we modeled the drg samples as above , adopting in turn the lowest and highest redshift allowed by the 68% confidence interval around @xmath65 for each object . although the effects on the median values of the best - fit properties can be appreciable ( by up to 0.7 gyr for the age , 0.9 mag for @xmath139 , a factor of 2 for @xmath151 , and a factor of 5 for sfr , depending on the field and sfh ) , our conclusions are qualitatively unaffected . we also modeled the samples of drgs including the objects outside of the @xmath219 range . the impact on the median properties is significantly smaller ( @xmath220 , @xmath221 , @xmath180 , and @xmath222 for the ages , @xmath139 s , @xmath151 s , and sfrs , respectively ) . the sed fits of figure [ fig - seds ] illustrate well the difficulty in discriminating between the two sfhs ( in particular , e.g. , for @xmath223 and @xmath224 ) . based on the formal @xmath225 values , the fits are statistically indistinguishable for all but three galaxies , for which the csf model could in principle be ruled out at the 95% confidence level . for two of these , @xmath226 and @xmath227 , the higher flux in the @xmath49 band relative to the @xmath27 band strongly influences this result . although their @xmath15 colour is fairly blue , they are not the most extreme of our sample . for the other , @xmath228 , the optical data are better reproduced with the @xmath163 model . for this spectroscopically studied drg @xcite , correcting for line emission would affect mostly the observed @xmath55-band flux ( downwards by 10% for ly@xmath131 ) and would improve the agreement with the csf model . the distinction between the models is not very robust and the @xmath225 computation relies on uncertainties that do not account for those of @xmath65 or possible emission line contributions . in view of this , we chose not to discriminate between sfh models for individual sources in our analysis . we explored the effects of variations in three of the model assumptions : the sfh , the extinction law , and the metallicity . figure [ fig - paramvar ] summarizes the results for the ages , extinction , stellar masses , and sfrs derived for different sets of input parameters . while the effects may differ among individual objects , we are here interested in the ensemble properties of drgs and so the discussion refers to the median values . we find that the sfh has the largest impact on the derived quantities . the age , extinction , and instantaneous sfr are the most affected ones whereas the stellar mass and m / l ratio are the least sensitive ones and change only by factors of @xmath229 . we considered exponentially declining sfrs with a range of @xmath24-folding timescales @xmath230 between 10 myr and 1 gyr . they are compared with the csf and also an ssp model , equivalent to @xmath231 and @xmath232 , respectively . with our broad - band data , it is not possible to discriminate among any of the sfhs for the large majority of our drgs . the overall trend with decreasing timescale is of younger ages by up to an order of magnitude . the extinction values become lower but do not vary monotonically with @xmath230 , exhibiting a minimum around 300 myr . the maximum difference amounts to 1.8 mag , or a factor of @xmath233 in attenuation at @xmath26 in the rest - frame . it is intriguing that the lowest extinction values are not derived for the models with shortest timescales , as might be expected . this may be due to worse template mismatch at @xmath234 . the sfrs drop by several orders of magnitude , to some extent because of lower extinction but mainly as a consequence of the functional form of the assumed sfr itself combined with the ages , which are such that @xmath235 . the stellar masses tend to decrease with smaller @xmath230 but remain within a factor of 2 of the masses for csf . we thus reach similar conclusions about the stability of derived @xmath151 for drgs as @xcite and @xcite in their modeling of lbgs also based on broad - band optical - to - nir photometry . as discussed by these authors , this is in part due to the nir data probing the rest - frame optical regime , which is less sensitive to @xmath166 variations due to age or extinction than the rest - frame uv . in addition , the age and dust degeneracies tend to cancel out in terms of the effect on the ratio of stellar mass to observed ( attenuated ) light . the same behaviour is seen in the case of our drgs ( see * ? ? ? * for a discussion ) . while age and extinction can vary strongly with the sfh , the combinations of best - fit values result in much tighter constraints on @xmath151 and @xmath236 . the largest extinction is derived for the longest timescales . the need for high extinction here may be artificially enhanced by the choice of sfh along with the constraint that the best - fit ages do not exceed that of the universe at each object s redshift . in the limit of csf , the population of young massive stars rapidly levels off on the timescale of their main - sequence lifetime ( @xmath237 ) while the population of lower - mass longer - lived stars grows progressively . to match the typically red seds of drgs , the uv light from the luminous massive stars continuously formed has to be suppressed with sufficient extinction . the maximum age constraint limits the relative contribution of evolved versus young stars to the integrated light , further enhancing the need for extinction . over the range of sfhs explored here , the strong coupling between age and @xmath230 also reflects the need for the build - up of a large enough population of evolved stars . models with very short timescales @xmath238 where the extinction is allowed to vary lead to ages @xmath239 and fairly high @xmath240 . arguably , such young dusty burst models may not be appropriate for the ensemble of drgs ( though they may be for some individual objects ) in view of the evidence for a prominent spectral break in their rest - frame continuum seds ( see [ sect - prop ] ) . choices of @xmath241 lead to ages consistent with the appearance of a strong balmer/4000 break in both cases of very quiescent or still actively star - forming galaxies . although models with @xmath242 can not be statistically ruled out in our fits , we regard them as least plausible for drgs . to test the effects of possible variations in the extinction law and in metallicity , we fitted two alternative suites of models to the seds . in one variant , we modified the prescription for the extinction law to that of @xcite for the milky way ( mw ) , keeping the solar metallicity . in the other , we used the set of bc03 models for metallicity @xmath243 and adopted for consistency the extinction law for the small magellanic cloud ( smc ; * ? ? ? * ; * ? ? ? the main differences between the @xcite and mw extinction laws lie in the ratio of total - to - selective absorption @xmath244 ( 4.05 versus 3.1 , respectively ) and in the @xcite law lacking the 2175 bump characteristic of mw dust mixtures . otherwise , their wavelength dependence are fairly similar . the smc law with @xmath245 also lacks the 2175 bump . in addition , it rises more steeply with decreasing wavelengths in the near - uv than the other two laws ; in other words , the @xcite and mw laws are much `` greyer '' at near - uv wavelengths . for the models with @xmath246 and the mw law , the ages for a given sfh remain the same or become mostly younger , by up to a factor of 3 . the best - fit @xmath139 s also remain the same or decrease , by at most 0.6 mag . the effects are more important for the @xmath243 and smc law models , with generally older ages by up to a factor of 5 and @xmath139 s lower by up to 2 mag . in both model sets , the instantaneous sfrs are strongly affected towards shorter timescales @xmath230 , especially for the low - metallicity case . the stellar masses are systematically lower for both variants , by a factor of @xmath140 on average . the near - systematic increase in median age of the drgs for the lower metallicity models can be attributed to the combined effects from the extinction law and stellar emission . since the smc law is steeper in the rest - frame uv than both the @xcite and mw laws , lower @xmath139 values will reproduce the observed sed slopes in this regime but also imply bluer rest - frame uv - to - optical colours such that older ages are required to match the colours of drgs . in addition , at @xmath247 and for a fixed age , the stellar evolutionary tracks and model atmospheres predict higher effective temperatures , bluer continua , and shallower metallic and molecular absorption features , further driving up the best - fit model ages . as emphasized by @xcite , it is important here to keep in mind that theoretical stellar tracks and stellar atmosphere models for non - solar metallicities are still not fully tested against empirical data ( see the discussion by * ? ? ? the variations in derived parameters we describe above provide nevertheless indications of the magnitude of metallicity effects . we further note that the high obscuration inferred for the models with star formation timescales @xmath248 raises concerns about the applicability to drgs of the adopted @xcite extinction law and foreground dust screen geometry . the @xcite law provides a good representation for local uv - bright starbursts with @xmath139 up to about 2.5 mag and where the foreground dust approximation is observationally supported . for the csf model , for instance , the @xmath139 values obtained for half of the drgs indicate extinction levels at the limit of , or beyond the range of validity . this may suggest that the @xcite law is inappropriate , and that the dust and sources geometry is more complex . the picture that emerges from our sed analysis and modeling is that drgs constitute an old and massive galaxy population at high - redshift . obscuration by interstellar dust plays an important role in their observed properties . depending on their actual sfh , they may also still be very actively forming stars . in this section , we interpret our results in a broader context by comparing our drgs with lbgs . we also consider the ensemble of all @xmath27-band selected galaxies in both fires fields at @xmath9 . we modeled these samples in the exact same way as the drgs . an immediate implication of the model results presented in [ sect - models ] is that drgs appear to be older , more obscured , and with higher stellar masses than lbgs . our sample of optically - selected lbgs at @xmath87 and @xmath9 in hdf - s allows a consistent comparison , based on the same data and with identical model ingredients and assumptions . with our standard csf model , we derive median values for the lbgs of @xmath249 , @xmath250 , @xmath251 , @xmath252 , and @xmath253 . with the @xmath163 model , we find lower median age of 360 myr and sfr of @xmath254 , but the same median extinction , stellar mass , and m / l ratio . we can not statistically discriminate between the fits for different sfhs for any of our lbgs . compared to the hdf - s lbgs and with csf , our drgs are typically @xmath255 times older , @xmath256 times more obscured , have roughly an order of magnitude higher stellar mass , and form stars at @xmath257 times higher instantaneous rates ( based on the median properties ) . among the various declining sfhs we considered , the @xmath163 model leads to the lowest extinction for the drgs ( figure [ fig - paramvar ] ) . one might expect that this sfh reduces most the inferred differences between the two populations . the drgs , however , remain distinct from lbgs in their median properties except for the sfr , being 3 times older , 1.5 times more extincted , @xmath258 times more massive , and @xmath259 times as actively star - forming . as for the drgs , the lbgs show an appreciable dispersion in best - fit properties with some overlap with drgs . however , the differences in the median values and combinations thereof ( especially age , @xmath139 , and @xmath151 ) indicate that the ensemble properties of the two populations are different for either sfh . the distinction between the drgs and lbgs is best illustrated in figure [ fig - mlv_av ] , which compares the distribution of the two populations in the @xmath260 plane ( where the median uncertainties from the 68% confidence intervals are @xmath261 in @xmath194 and @xmath204 in @xmath139 for the drg and lbg samples , and our two standard models ) . in this parameter space , the overlap is very small for the csf model . for the @xmath163 model , about two - thirds of the drgs shift to a lower extinction interval @xmath262 , in common with the lbgs , while the rest uniformly populates the higher @xmath139 range explored . the ensemble of drgs and lbgs remain however clearly separated in the ranges of @xmath236 ratios covered , with very little overlap . this is driven by the stellar masses since the drg and lbg samples are similarly luminous in the rest - frame @xmath26 band , with nearly identical median values of @xmath263 ( within 20% ) . the weakness of our hdf - s lbg sample is that @xmath66 s are available for seven out of 33 galaxies . however , we arrive at similar conclusions if we consider the model results for the spectroscopically confirmed lbg samples of @xcite and @xcite , also based on optical to nir seds . the typical ( median or geometric mean ) ages inferred are @xmath264 , with extinction @xmath32 and @xmath265 ( for @xmath246 , a salpeter imf down to 0.1@xmath16 , and our adopted cosmology ) . our results are within a factor of a few of those of @xcite and @xcite , similar to the difference between the latter two . it is however difficult to compare in detail the results among all three studies . the data sets are different in terms of depth , quality , and photometric bands involved . the sample of @xcite was selected in the hdf - n from the space - based hst data while the ground - based sample of @xcite was selected over a much wider area and is brighter . in their modeling , @xcite assumed exponentially declining sfhs with the timescale as a free parameter . @xcite did not attempt to fit the sfh and discuss mainly the csf scenario . in addition , the models of @xcite and @xcite relied on earlier versions of the bruzual & charlot code and partly different model ingredients . in view of this , it is probably fair to say that the properties derived for all three lbg samples are in broad agreement . the conclusions about the differences between drgs and lbgs remain qualitatively the same . we investigated relationships between the modeled quantities and the observed nir properties , the key feature of our drg selection . for this purpose , we complemented the drg samples with the hdf - s lbgs and all other @xmath27-band selected galaxies at @xmath9 in both the hdf - s and @xmath4 fields down to the same @xmath109 limits of 22.5 and 21.7 mag , respectively . figure [ fig - resall1 ] plots , for our standard @xmath163 model , the derived stellar masses and m / l ratios as a function of observed @xmath28 colour and @xmath27-band magnitude . the results with the csf are qualitatively similar . there is a clear correlation between the derived @xmath151 and the observed @xmath28 colour , which is even tighter for the @xmath266 ratio . the redder the objects , the higher the stellar masses and mass - to - light ratios . the drgs lie at the massive end and extend smoothly the relationships for the @xmath9 objects with bluer @xmath28 colours , including the lbgs . because of the different limiting magnitudes between the hdf - s and @xmath4fields , it is difficult to draw conclusions on the variations of @xmath151 and @xmath267 versus @xmath109 . the data suggest that our @xmath27-band selection does not miss a significant fraction of massive galaxies at @xmath9 that would be faint because of extinction effects . figure [ fig - resall2 ] plots the variations of the derived extinctions and instantaneous star formation rates per unit stellar mass with observed @xmath28 colours . results for both the standard csf and @xmath163 models are shown . there are general trends of increasing @xmath139 and of decreasing @xmath268 with redder @xmath28 colour . the variations in derived ages ( not shown in figure [ fig - resall2 ] ) are essentially the reverse of those in the @xmath268 ratio , which provides a measure of @xmath269 ( with a direct proportionality for the csf model , and a non - linear monotonic relation for the @xmath163 model ) . these plots further show that if drgs might be exceptional in their high absolute sfrs , they are not at all in terms of their sfr per unit mass . we here speculate about the evolution of the @xmath9 drgs in the fires fields , based on the results for the csf and @xmath163 models . figure [ fig - cosmic_sfr ] plots the relative variations with redshift of the average sfr of the drg samples and , for comparison , of the hdf - s lbg sample at @xmath9 . we computed the curves from the model sfr @xmath270 used as input in the evolutionary synthesis . for each galaxy , we scaled the model @xmath270 curves to match the derived instantaneous sfr at the epoch of observation . for the purpose of this discussion , we in fact assumed that all galaxies lie at @xmath271 , close to the mean and median redshift of both our drg and lbg samples ( see [ sect - sample ] ) . we then normalized the resulting average sfr curves to unity at @xmath271 . for the csf model , the average instantaneous sfr of the drg samples has not varied strongly since @xmath272 up to @xmath98 , increasing by a factor of @xmath140 with time . on the other hand , the average instantaneous sfr of the hdf - s lbg sample has increased by more than an order of magnitude over the same redshift interval . for the @xmath163 model , the variations are much noisier but indicate , broadly speaking , roughly comparable activity levels in the past out to @xmath273 , where the mean instantaneous sfr of lbgs drops abruptly . the relative variations in average sfr backwards in time are dominated by the age distribution among each population , and the steeper drop for the lbgs reflects their younger ages compared to the drgs . this extrapolation of the star formation histories suggests that @xmath9 drgs have started forming stars at very early epochs , and earlier than @xmath9 lbgs have . a possible scenario for drgs is that they are the evolved descendants of systems that experienced a `` lyman - break phase '' at @xmath34 ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? . this evolutionary picture may be consistent with the increased reddening in rest - frame uv - to - optical colours inferred by @xcite for lyman - break selected galaxies between @xmath274 ( `` @xmath55-dropouts '' ) and @xmath33 ( `` @xmath54-dropouts '' ) . we emphasize that the above discussion applies to the stellar populations that dominate the rest - frame uv - optical emission of the galaxies , as traced by the observed seds . as discussed by @xcite , the stellar masses of drgs are already very high at @xmath275 and comparable to those of early - type galaxies locally . this may indicate that drgs have already assembled most of their stellar mass at these redshifts and evolve thereafter mostly passively . using our csf model results , we estimated the stellar masses predicted at @xmath276 if star formation is maintained at the same rate as inferred at the redshift of observation . the median present - day stellar masses would be @xmath277 , which correspond to the most massive galaxies in local galaxy clusters . while some of the drgs might evolve into such systems , it seems rather unlikely we have detected such large numbers of their progenitors in our two small fields . more plausibly , most of the drgs probably become predominantly quiescent systems at @xmath278 , and may possibly grow further in mass during brief merger events . assuming that drgs stop forming stars around their redshift of observation and evolve passively thereafter , the csf models predict colours of @xmath279 and @xmath280 at @xmath281 , even if all the dust is removed at the instant when star formation is quenched ( such red colours are also predicted using the @xmath163 models instead ) . this may suggest an evolutionary connection between drgs at @xmath1 and eros at @xmath282 . in absolute terms , the instantaneous sfrs of @xmath283 derived from the sed modeling assuming csf lie at the high end inferred for the local galaxy population , in the regime of luminous and ultra - luminous infrared galaxies . such a high sfr rate seems supported by the possible detection of one of the non - agn spectroscopically - confirmed drg in the @xmath4 field at sub - millimeter wavelengths from deep scuba observations ( @xcite ; k. kraiberg knudsen et al . 2004 , in preparation ) . the csf hypothesis is unverifiable with the data at hand , and may be a valid representation for a few objects only . conversely , while the @xmath163 model implies substantially lower instantaneous sfr at the redshift of observation , the high stellar masses lead to extremely high sfrs at earlier epochs ( @xmath284 ) . obviously , the sfh of drgs may well be more complex than our simplistic assumptions . the less model - dependent estimate from @xmath215 of @xmath283 seems to suggest elevated levels of star formation activity averaged over the lifetime of the stellar populations traced by our seds . using the results presented in this paper , we can crudely assess the relative importance of the drg and lbg populations at high redshift . in the following , we consider our samples of drgs , lbgs , and @xmath27-band selected sources lying at @xmath9 and to the @xmath27 total magnitude limits of 22.5 for hdf - s and 21.7 for the @xmath4 field . in hdf - s , the lbgs outnumber the drgs by a factor of three . they contribute about 2.5 times more than drgs to the total observed @xmath43-band light and extrinsic rest - frame @xmath26-band luminosity @xmath285 from all sources , with fractions of @xmath286 and @xmath287 , respectively . the relative contributions are reversed for the stellar mass , where drgs account for @xmath286 of the total compared to @xmath288 for lbgs ( using the individual @xmath151 s derived from either the csf or @xmath163models , and assuming the same model parameters for all populations ) . in the @xmath4 field and to the corresponding brighter magnitude limit , the drgs produce about 35% of the total observed @xmath27-band light and @xmath285 , and make up @xmath289 of the integrated @xmath151 . this provides a first - order indication that drgs may represent a significant constituent of the @xmath275 universe in terms of stellar mass . admittedly , our estimates are rough . a detailed analysis of the contribution of drgs to the stellar mass budget in both fires fields will be presented by g. rudnick et al . ( in preparation ; see also @xcite for first estimates in hdf - s ) . we have presented the analysis and theoretical modeling of the @xmath12 seds of 34 drgs at @xmath9 . the galaxies were selected by their @xmath0 colours in the hdf - s and @xmath4 fields observed as part of the fires survey @xcite , at @xmath290 and @xmath116 , respectively . the addition of the @xmath4 field increases the area surveyed by a factor of five to an @xmath291 brighter limit and triples the original sample from hdf - s . this allows for the first time a robust assessment of the ensemble properties of drgs . we find large numbers of @xmath28 red objects in both fields , with surface densities of @xmath292 at @xmath293 . since no @xmath0 source is found in the hdf - n down to @xmath115 , cosmic variance is substantial . observations of wider areas and multiple lines of sight will be required for a more reliable determination of the surface density of the @xmath28 red population . using new diagnostic diagrams involving the @xmath13 , @xmath14 , and @xmath15 colours , we have shown that the red nir colours of the drgs can not be attributed to extinction effects alone and require the presence of a prominent balmer/4000 break characteristic of evolved stars for a large fraction of them . in the rest - frame optical , the seds of drgs fall within the envelope of normal nearby galaxies and , in the rest - frame uv , indicate a wide range in star formation activity and/or dust obscuration . the rest - frame uv - to - optical seds of drgs are in stark contrast with those of lbgs , which are much bluer and more uniform . these differences can be easily understood in terms of selection effects . the lbg criteria preferentially select galaxies that are actively star - forming and modestly obscured . our @xmath0 criterion primarily targets evolved systems , ranging from those with quiescent or no star formation and little extinction to those that are highly obscured and vigorously forming stars . we have applied evolutionary synthesis models to the seds of our drgs to constrain quantitatively the age of the stellar populations , the extinction , the stellar masses , and the star formation rates . we have also investigated the effects of variations in the assumed star formation history , metallicity , and extinction law on the model results . the derived properties are similar for the hdf - s and @xmath4 field samples . for constant star formation , a salpeter imf between 0.1 and 100@xmath16 , solar metallicity , and the @xcite extinction law , we obtain for each field a median age @xmath294 , extinction @xmath295 , stellar mass @xmath296 , stellar mass - to - light ratio @xmath297 , and instantaneous star formation rate @xmath298 . the stellar masses and m / l ratios are the most robust properties against variations in the model parameters ( neglecting those of the imf ) while the instantaneous sfrs are strongly model - dependent . models with exponentially declining star formation rates and @xmath24-folding timescales in the range @xmath299 generally lead to lower ages , extinction values , and instantaneous sfrs but similar stellar masses and m / l ratios within a factor of two . the smallest median @xmath139 s are derived with @xmath162 . for this @xmath163 model and the assumptions above for the imf , metallicity , and extinction law , the median age and @xmath139 are 1 gyr and 0.9 mag for both fields , and the median sfrs @xmath300 . an important implication of our work is that drgs appear to be dominated by significantly older , more obscured , and more massive stellar populations than lbgs at similar redshifts and similar rest - frame @xmath26-band luminosities . using our sample of optically - selected lbgs in hdf - s , we have explicitely shown that the distinction holds for constant star formation and @xmath163models , the two star formation histories that bracket the range of median @xmath139 s derived for our drgs ( neglecting variations in model parameters between the two classes ) . for both drgs and lbgs , there is a large spread in derived properties among individual objects and some of the drgs have ages , extinction , and stellar masses characteristic of lbgs , and vice - versa . however , the overlap is small , especially for the stellar masses and m / l ratios . comparison with the lbg studies of @xcite and @xcite is less straightforward because of differences in the optical / nir data sets , in the details of the sample selection , and in the modeling but , broadly speaking , it supports the same conclusions . more generally , we find that the derived age , extinction , and stellar mass all increase with redder @xmath28 colour among the full sample of @xmath9 @xmath27-band - selected galaxies in the fires fields , down to the same magnitude limits as the drg samples . the relationships are particularly well - defined for the stellar mass and m / l ratio versus @xmath28 colour . if drgs represent the most evolved and most massive galaxies at @xmath301 , it is tempting to speculate that they might be the precursors of the most massive present - day galaxies . such a final fate has been proposed for lbgs at @xmath33 in the context of hierarchical models of galaxy formation ( e.g. * ? ? ? * ) . for drgs , with stellar masses already similar to those of local early - type galaxies , one may reason in a simplistic way that there is little room for substantial growth through subsequent merger events . extrapolating the star formation histories backward in time for both drgs and lbgs at @xmath9 , we infer that the drg population has been forming stars at a roughly constant rate and possibly back to @xmath302 , whereas the lbgs started to form later at @xmath274 . this is consistent with a scenario in which the drgs are the evolved descendants of systems undergoing a `` lyman - break phase '' at @xmath303 . the drgs may evolve into the ero population at @xmath281 . another intriguing possibility is a connection with the high - redshift population detected at sub - millimeter wavelengths . one of the drgs in the @xmath4 field likely is the counterpart of a sub - millimeter source detected with scuba ( @xcite ; k. kraiberg knudsen et al . 2004 , in preparation ) . in addition , although detailed information is currently available for a few objects only , recent work indicates that some of the bright sub - millimeter galaxies at @xmath304 host evolved stellar populations , show evidence of substantial chemical enrichment , and have dynamical / stellar masses @xmath305 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the results presented here constitute a major improvement over our previous studies of @xmath28 selected objects at @xmath1 . significant uncertainties remain , however , and further investigations are clearly needed . our analysis still relies on a limited sample selected from two small fields and the model parameters are largely unconstrained . deep photometric surveys of wider areas and disjoint fields as well as extensive follow - up spectroscopy will be crucial to assess ( and average over ) field - to - field variations , secure the redshifts , derive robust number densities , and constrain more accurately the physical properties ( e.g. , extinction , metallicity , dynamical masses ) . the current dearth of constraints on the nature of drgs further limits our ability to interpret their observed properties . in that respect , it will be very important to establish firmly the frequency of agn among drgs and their contribution to the measured optical / nir seds , in particular from spectroscopy . it will also be important to obtain independent constraints on the current and past star formation rates , with observations at infrared and submillimeter wavelengths ( e.g. , with scuba and the spitzer space telescope ) complementing rest - frame optical line diagnostics . ultimately , a better understanding of drgs and their relationship to other galaxy populations at high and low redshift may provide important constraints on the assembly history of massive galaxies . we would like to thank the eso staff for their assistance and their efforts in obtaining the high quality isaac and fors1 data for fires and in making them available to us . we are grateful to henk hoekstra for kindly computing for us the lensing magnifications in the @xmath4 field . we thank the referee , casey papovich , for very useful comments and suggestions that helped improve the paper . n.m.f.s . appreciates the critical discussions and encouraging comments of matt lehnert . n.m.f.s . acknowledges the department of astronomy at yale university for its hospitality and generous support during a working visit , and the leids kerkhoven - bosscha fonds for travel support . allen c. w. 1976 , in astrophysical quantities , university of london eds . , the athlone press , p.264 baugh , c. m. , cole , s. , frenk , c. s. , & lacey , c. g. 1998 , , 498 , 504 bertin , e. , & arnouts , s. 1996 , , 117 , 393 bolzonella , m. , miralles , j .- m . , & pell , r. 2000 , , 363 , 476 bouchet , p. , lequeux , j. , maurice , e. , prvot , l. , & prvot - burnhon , m. l. 1985 , , 149 , 330 bruzual , a.g . , & charlot , s. 1993 , , 405 , 538 bruzual , a.g . , & charlot , s. 2003 , , 344 , 1000 buta , r. , mitra , s. , de vaucouleurs , g. , & corwin , h. g. , jr . 1994 , , 107 , 118 calzetti , d. , armus , l. , bohlin , r. c. , kinney , a. l. , koornneef , j. , & storchi - bergmann , t. 2000 , , 533 , 682 casertano , s. , et al . , 2000 , , 120 , 2747 chabrier , g. 2003 , , 115 , 763 coleman , g.d . , wu , c .- c . , & weedman , d.w . 1980 , , 43 , 393 daddi , e. , cimatti , a. , pozzetti , l. , hoekstra , h. , rttgering , h. j. a. , renzini , a. , zamorani , g. , & mannucci , f. 2000 , , 361 , 535 daddi , e. , et al . 2003 , , 588 , 50 dickinson , m. , et al . , 2000 , , 531 , 624 elston , r. , rieke , g. h. , & rieke , m. j. 1988 , , 331 , l77 erb , d. k. , shapley , a. e. , steidel , c. c. , pettini , m. , adelberger , k. l. , hunt , m. p. , moorwood , a. f. m. , & cuby , j .- g . 2003 , , 591 , 101 ferguson , h. c. , dickinson , m. , & papovich , c. 2002 , , 569 , l65 fernndez - soto , a. , lanzetta , k. m. , & yahil , a. 1999 , , 513 , 34 franx , m. , et al . 2000 , the messenger , 99 , 20 franx , m. , et al . 2003 , , 587 , l79 genzel , r. , baker , a. j. , tacconi , l. j. , lutz , d. , cox , p. , guilloteau , s. , & omont , a. 2003 , , 584 , 633 giavalisco , m. , & dickinson , m. 2001 , , 550 , 177 giavalisco , m. , steidel , c. c. , adelberger , k. l. , dickinson , m. e. , pettini , m. , & kellogg , m. 1998 , , 503 , 543 hall , p. b. , et al . , 2001 , , 121 , 1840 hoekstra , h. , franx , m. , & kuijken , k. 2000 , , 532 , 88 hu , e. m. , & ridgway , s. e. 1994 , , 107 , 1303 jansen , r. a. , franx , m. , fabricant , d. , & caldwell , n. 2000 , , 126 , 271 kinney , a. l. , calzetti , d. , bohlin , r. c. , mcquade , k. , storchi - bergmann , t. , & schmitt , h. r. 1996 , , 476 , 38 koekemoer , a. m .. , et al . , 2004 , , 600 , l123 kron , r. g. 1980 , , 43 , 305 kroupa , p. 2001 , , 322 , 231 labb , i. , et al . 2003 , , 125 , 1107 madau , p. 1995 , , 441 , 18 mccarthy , p. j. , et al . , 2001 , , 560 , l131 moorwood , a. f. m. , et al . , 1998 , the messenger , 94 , 7 neri , r. , et al . 2003 , , 597 , l113 norman , c. , et al . , 2002 , , 571 , 218 oke , j. b. 1971 , , 170 , 193 papovich , c. , dickinson , m. , & ferguson , h. c. 2001 , , 559 , 620 papovich , c. , et al . 2004 , , 600 , l111 pettini , m. , shapley , a. e. , steidel , c. c. , cuby , j .- g . , dickinson , m. , moorwood , a. f. m. , adelberger , k. l. , & giavalisco , m. 2001 , , 554 , 981 prvot , m. l. , lequeux , j. , prvot , l. , maurice , e. , & rocca - volmerange , b. 1984 , , 132 , 389 rudnick , g. , et al . , 2001 , , 122 , 2205 rudnick , g. , et al . 2003 , , 599 , 847 saracco , p. , giallongo , e. , cristiani , s. , dodorico , s. , fontana , a. , iovino , a. , poli , f. , & vanzella , e. 2001 , , 375 , 1 saracco , p. , et al . , 2003 , , submitted ( astro - ph/0310131 ) schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 scodeggio , m. , & silva , d. r. 2000 , , 359 , 953 shapley , a. e. , erb , d. k. , pettini , m. , steidel , c. c. , & adelberger , k. l. 2004 , , in press ( astro - ph/0405187 ) shapley , a. e. , steidel , c. c. , adelberger , k. l. , dickinson , m. , giavalisco , m. , & pettini , m. 2001 , , 562 , 95 shapley , a. e. , steidel , c. c. , pettini , m. , & adelberger , k. l. 2003 , , 588 , 65 smith , g. p. , et al . , 2002 , , 330 , 1 steidel , c. c. , adelberger , k. l. , giavalisco , m. , dickinson , m. , & pettini , m. 1999 , , 519 , 1 steidel , c. c. , & hamilton , d. 1993 , , 105 , 2017 steidel , c. c. , giavalisco , m. , dickinson , m. , & adelberger , k. l. 1996a , , 112 , 352 steidel , c. c. , giavalisco , m. , pettini , m. , dickinson , m. , & adelberger , k. l. 1996b , , 462 , l17 tecza , m. , et al . 2004 , , in press ( astro - ph/0403264 ) thompson , d. , et al . , 1999 , , 523 , 100 totani , t. , yoshii , y. , iwamuro , f. , maihara , t. , & motohara , k. 2001 , , 558 , l87 van dokkum , p. g. , et al . 2003 , , 5887 , l83 van dokkum , p.g . , franx , m. , fabricant , d. , illingworth , g.d . , & kelson , d.d . 2000 , , 541 , 95 van dokkum , p. g. , et al . 2004 , , in press ( astro - ph/0404471 ) yan , l. , mccarthy , p. j. , weymann , r. j. , malkan , m. a. , teplitz , h. i. , storrie - lombardi , l. j. , smith , m. , & dressler , a. 2000 , , 120 , 575 lcccc & median & mean & rms + + @xmath109 & ( mag ) & 21.39 & 21.25 & 0.55 + @xmath306 & ( mag ) & 26.33 & 26.69 & 0.96 + @xmath28 & ( mag ) & 2.60 & 2.75 & 0.43 + @xmath143 & & 2.50 & 2.60 & 0.42 + + + + & median & mean & rms + + @xmath109 & ( mag ) & 20.95 & 20.84 & 0.74 + @xmath306 & ( mag ) & 26.86 & 26.61 & 0.94 + @xmath28 & ( mag ) & 2.79 & 2.85 & 0.43 + @xmath143 & & 2.42 & 2.44 & 0.30 + llcccccc & & & + & median & mean & rms & median & mean & rms + + age @xmath181 & ( gyr ) & 1.7 & 1.5 & 0.86 & 1.0 & 1.1 & 1.0 + sfr - weighted age @xmath182 & ( gyr ) & 0.85 & 0.74 & 0.43 & 0.75 & 0.93 & 0.92 + @xmath139 & ( mag ) & 2.7 & 2.0 & 0.78 & 0.90 & 1.20 & 0.96 + @xmath151 & ( @xmath307 ) & 8.1 & 15 & 13 & 7.5 & 7.9 & 5.7 + @xmath152 & ( @xmath308 ) & 5.0 & 5.7 & 3.6 & 5.0 & 5.7 & 3.6 + @xmath236 & ( @xmath309 ) & 1.2 & 2.7 & 2.2 & 1.2 & 1.5 & 0.89 + @xmath310 & ( @xmath311 ) & 91 & 120 & 96 & 91 & 150 & 160 + @xmath312 & ( @xmath311 ) & 120 & 150 & 120 & 6.7 & 96 & 140 + @xmath313 & ( @xmath314 ) & 7.6 & 19 & 26 & 1.5 & 18 & 25 + @xmath315 & ( @xmath316 ) & 30 & 33 & 30 & 1.2 & 21 & 32 + + + + & & & + & median & mean & rms & median & mean & rms + + age @xmath181 & ( gyr ) & 2.0 & 1.7 & 0.84 & 1.0 & 1.3 & 0.91 + sfr - weighted age @xmath182 & ( gyr ) & 1.0 & 0.83 & 0.42 & 0.75 & 1.1 & 0.85 + @xmath139 & ( mag ) & 2.4 & 2.2 & 0.56 & 0.90 & 1.3 & 0.80 + @xmath151 & ( @xmath307 ) & 16 & 21 & 16 & 10 & 15 & 11 + @xmath152 & ( @xmath308 ) & 5.1 & 7.4 & 6.4 & 5.1 & 7.4 & 6.4 + @xmath236 & ( @xmath309 ) & 2.3 & 3.3 & 1.9 & 1.8 & 2.4 & 1.7 + @xmath310 & ( @xmath311 ) & 140 & 150 & 110 & 94 & 170 & 150 + @xmath312 & ( @xmath311 ) & 170 & 190 & 130 & 23 & 69 & 110 + @xmath313 & ( @xmath314 ) & 6.5 & 15 & 19 & 1.5 & 6.9 & 10 + @xmath315 & ( @xmath316 ) & 31 & 32 & 22 & 2.7 & 11 & 17 +	we investigate the nature of the substantial population of high - redshift galaxies with @xmath0 colours recently discovered as part of our faint infrared extragalactic survey ( fires ) . this colour cut efficiently isolates galaxies at @xmath1 with red rest - frame optical colors ( `` distant red galaxies , '' or `` drgs '' ) . we select @xmath0 objects in both fires fields , the @xmath2 hubble deep field south ( hdf - s ) and the @xmath3 field around the @xmath4 cluster at @xmath5 ; the surface densities at @xmath6 are @xmath7 and @xmath8 , respectively . we here discuss a sub - sample of 34 drgs at @xmath9 : 11 at @xmath10 in hdf - s and 23 at @xmath11 in the @xmath4 field . this sample enables for the first time a robust assessment of the population properties of drgs . we analyze the @xmath12 spectral energy distributions ( seds ) constructed from our very deep near - infrared ( nir ) and optical imaging collected at the eso very large telescope and from the hubble space telescope . we develop diagnostics involving the @xmath13 , @xmath14 , and @xmath15 colours to argue that the red nir colours of our drg sample can not be attributed solely to interstellar dust extinction and require for many the presence of an evolved stellar population with a prominent balmer/4000 break . in the rest - frame , the optical colours of drgs fall within the envelope of normal nearby galaxies and the ultraviolet colours suggest a wide range in star formation activity and/or extinction . this is in stark contrast with the much bluer and more uniform seds of lyman - break galaxies ( lbgs ) . from evolutionary synthesis modeling assuming constant star formation ( csf ) , we derive for the drgs old ages , large extinctions , and high stellar masses , mass - to - light ratios , and star formation rates . for solar metallicity , a salpeter imf between 0.1 and 100@xmath16 , and the calzetti et al . extinction law , the median values for the hdf - s ( @xmath4 field ) sample are 1.7 ( 2.0 ) gyr , @xmath17 ( 2.4 ) mag , @xmath18 ( 1.6 ) @xmath19 , @xmath20 ( 2.3 ) @xmath21 , and @xmath22 ( 170 ) @xmath23 . models assuming exponentially declining sfrs with @xmath24-folding timescales in the range @xmath25 generally imply younger ages , lower extinction , and lower sfrs , but similar stellar masses within a factor of two . compared to lbgs at similar redshifts and rest - frame @xmath26-band luminosities , drgs are older , more massive , and more obscured for any given star formation history . for the entire sample of @xmath27-band selected galaxies in the fires fields at @xmath9 and to the above magnitude limits , we find that the derived ages , extinctions , and stellar masses increase with redder @xmath28 colours . although the rest - frame optical colours of the drgs are similar to those of local normal galaxies , the derived properties are quite different ; detailed studies of this new @xmath1 population may significantly enhance our understanding of how massive galaxies assembled their stellar mass .
hot cores are compact ( 0.1 pc ) , warm ( @xmath5 100300 k ) , dense ( 10@xmath6 h nuclei ) clouds of gas and dust near or around sites of recent star formation ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the hot - core phase is thought to last about 10@xmath7 yr @xcite to 10@xmath8 yr @xcite and represents the most chemically rich phase of the interstellar medium , characterized by complex molecules like , , , and . the complex chemical and physical processes occurring in the hot - cores are not fully understood . until recently hot cores were thought to be associated with high - mass protostars ( m@xmath9 8 @xmath10 ) only and to represent an important phase in their evolution toward ultracompact and compact h ii regions . the central energizing source for a number of hot cores have been identified using high angular resolution mid - infrared ( mir ) and millimeter continuum observations ( e.g. * ? ? ? * ; * ? ? ? these detections strengthen the idea of hot molecular cores ( hmcs ) as representing a stage in the evolutionary sequence of massive protostars . however , there are non - negligible examples of hot cores that are in the vicinities of uc h ii regions and appear to be only externally heated . for these sources it may well be argued that the hot cores can also arise as chemical manifestations of the effect of the uc h ii regions on their environments , rather than being only the precursors of uc h ii regions . the chemical models are still far from providing a unique interpretation of the hot core chemistry , and would benefit from high angular resolution continuum and spectroscopic observations suitable to understand the temperature and density distributions of the cores . in particular , the strong sensitivity of surface and gas - phase chemistry to dust temperature and gas density highlights the importance of the study of the abundances of complex molecules in different regions with varied physical characteristics . furthermore , all the existing chemical models assume the hot cores to be internally heated and have radially varying density and temperature profiles @xcite ; the models do not yet provide a consistent treatment of the externally heated hot cores . we present high angular resolution ( 1 ) interferometric observations with the berkeley - illinois - maryland association ( bima ) array of the well - studied hot core associated with the uc h ii region g34.26 + 0.15 located at a distance of 3.7 kpc @xcite . the h ii region is a prototypical example of cometary morphology , which may be due to the bow - shock interaction between an ambient molecular cloud and the wind from an energetic young star moving supersonically through the cloud @xcite . observations with the vla @xcite show that the highly compact molecular cloud appears to be wrapped around the head of the cometary h ii structure , with the ionization front advancing into the cloud . g34.26 + 0.15 has been extensively studied in radio continuum @xcite and radio recombination lines @xcite . at radio continuum frequencies , it exhibits several components : two uc h ii regions called a & b , a more evolved h ii region with a cometary shape named component c , and an extended ring - like h ii region with a diameter of 1 called component d @xcite . molecular gas has been mapped in , hco@xmath11 , so , and co @xcite . the hot core associated with g34.26 + 0.16 has been the target of chemical surveys using single - dish telescopes @xcite in which complex molecules characteristic of hot cores were detected . molecular line observations suggest that the hot core does not coincide with the h ii region component c ; it is offset to the east by at least 2 and shows no sign of being internally heated @xcite . based on narrow - band mid - infrared imaging of the complex , @xcite concluded that the same star is responsible for the ionization of the h iicomponent c and heating the dust but is not interacting with the hot core seen in molecular emission . at a resolution of 12 , @xcite also found the peak of the 350 emission to be coincident with the component c of the uc h ii region . in this paper we use the bima observations to study the energetics , chemistry and kinematics of the molecular gas contributing to the hot core emission associated with g34.26 + 0.15 . observations of the source g34.26 + 0.15 were acquired with the ten - element bima interferometer between 1999 december and march 2000 at three frequency bands centered approximately at 87 , 107 and 109 ghz using the a & b configurations of the array . due to technical difficulties only nine antennas could be used for the observations at 87 ghz . table [ obsdetails ] presents a log of the observations , including the typical system temperatures in the different configurations presented here . the primary fwhm of the array is between 132 and 106 at frequencies between 87 and 109 ghz . the correlator was configured to split each frequency into four windows per sideband , two of which had bandwidths of 100 mhz and 64 channels each and the remaining two had bandwidths of 50 mhz and 128 channels each . the sources uranus , 1830 + 063 and 1771 + 096 were observed as the primary flux calibrator , the phase calibrator and the secondary calibrator , respectively . however owing to the consistently poor quality of 1830 + 063 observations and the sparsity of uranus observations , we have used 1771 + 096 as both the phase and primary flux calibrator . the flux of 1771 + 096 was determined from each of the six datasets , using the miriad task bootflux with uranus as the primary calibrator . the average final flux for 1771 + 096 is 2.3 jy ; we estimate the absolute flux calibration error to be 10% . the pointing and phase center used for mapping the region around g34.26 + 0.15 is @xmath12 = 18@xmath1353@xmath141855 @xmath15 = 114582 and the @xmath16 = 58 . the data were reduced using the miriad @xcite software package . continuum maps were constructed by averaging over spectral windows that did not contain line emission . in order to achieve better phases the continuum maps were iteratively self calibrated . the spectral line maps within a particular spectral window were made by first fitting a low - order polynomial to the visibilities in the channels not contaminated by any line emission and then subtracting the fit to the continuum from the individual channels . the observed spectral lines were identified using the spectral line catalog by @xcite . we have detected multiple transitions of methanol ( ) , two transitions of ethyl cyanide ( ) and single transitions of methyl formate ( ) , dimethyl ether ( ) , formamide ( ) , , ocs ( and its @xmath17s , @xmath18c isotopomers ) , so , so@xmath1 and the hydrogen recombination line h53@xmath2 . table [ speclines ] presents spectroscopic details of the molecular lines that were detected in the present data set . table [ basic_results ] summarizes the basic results of the bima observations : synthesized beam sizes , the peak intensity , the peak brightness temperature , the central velocity ( @xmath19 ) , the velocity width ( @xmath20 ) and the achieved rms per channel of the interferometric maps , for each of the observed spectral lines . the @xmath19 and @xmath20 were determined from gaussian fits to the line profiles . in addition to the 10% error in absolute calibration we estimate the statistical errors in intensities to be @xmath21% . figure [ cont_cont ] shows an overlay of the @xmath222.8 mm continuum image with the @xmath222 cm continuum emission observed by @xcite . the 2.8 mm continuum is peaked at the nominal map center ( @xmath12 = 18@xmath1353@xmath141855 , @xmath15 = 114582 ) and coincides with the cometary component c of the g34.26 + 0.15 uc h ii region complex @xcite . as will be discussed in sec . [ intmap_peak ] the continuum peak detected at 2.8 mm does not coincide with the emission peaks of the lines originating in the hot molecular gas . although continuum at 2.8 mm shows a single peak , most of the radio continuum maps at cm wavelengths show the component c to consist of two emission peaks c1 and c2 , separated by @xmath23 . however @xcite suggest that c1 and c2 are not two separate continuum components , but they correspond to two regions of maximum emission in the nebula . recently @xcite mapped the g34.26 + 0.15 region at 43 ghz using vla and they find a single peak similar to the millimeter maps . based on the radio continuum spectrum the flux density due to the free - free emission from the components a & b is expected to be @xmath24 mjy at 100 ghz @xcite . this is consistent with our observations where we do detect some emission from the uc components a and b , though these components are not resolved in our maps . at 107 ghz ( @xmath3 mm ) we measure a peak flux density of 2.9 jy / beam and an integrated intensity of 6.7@xmath25 jy estimated by fitting a gaussian of 16@xmath2614 to the central source . in addition to the given statistical error for the continuum flux density , there is a 0.7 jy error corresponding to the uncertainty in the absolute flux calibration . the 2.8 mm continuum flux density measured here is consistent with previous observations at similar resolutions @xcite . @xcite measured an integrated radio continuum flux of 5 jy at 1.3 cm ( 24 ghz ) for the component c , which , when combined with the 2.8 mm flux indicates a spectral index of 0.2 . this spectral index is significantly flatter than the index of 0.45 derived by @xcite for the combined emission of all 3 components of the uc h ii region . at a resolution of 1 , the 2.8 mm emission is clearly dominated by the free - free emission from component c of the uc h ii region and does not arise from an embedded source that could be energizing the hmc internally . dust emission may be responsible for the eastern extension of the emission at a 10% level . figure [ allmap ] shows the integrated intensity maps of the different spectral line emission detected in the region around g34.26 + 0.15 overlaid with the continuum emission at @xmath3 mm and the positions of the h@xmath1o and oh masers detected in the region @xcite . the emission from none of the molecular species coincides with the positions of the h@xmath1o and oh masers . at a resolution of 1 the integrated intensity distribution for the different chemical species primarily show three kinds of morphological structures : _ single - peaked , double - peaked _ and _ irregular_. figure [ peaks ] shows the positions of the absolute maxima of the integrated intensities for the different species overlaid with the @xmath3 mm continuum map of g34.26 + 0.15 . the absolute positional accuracy is estimated to be 01 . we note that it is possible to identify primarily two regions where the maxima due to the transitions of the different species are localized : the first lying slightly to the south - east of the continuum peak and the second lies to the north - east of the 2.8 mm continuum peak . _ ( a ) single - peaked : _ the spectral lines of ( with the exception of the @xmath2715@xmath28 e2 transition ) , , nh@xmath1cho , so and ocs , as well as the h53@xmath2 recombination line show well - defined single peaked structures . also shows a single - peaked structure though it is somewhat poorly defined as compared to the rest . of these spectral lines only the peaks of the h53@xmath2 and the emission coincide with the peak of the continuum at @xmath3 mm , while the peaks due to and ocs occur to the north - east of the continuum peak . the peak due to so lies 1 to the south - east of the 2.8 mm continuum peak . the peaks of and nh@xmath1cho lie close enough ( @xmath5 04 south - east ) to but not exactly at the position of the 2.8 mm continuum peak . at a first glance , it appears that the more abundant species with higher optical depths tend to show a single - peaked structure . we revisit the question of optical thickness of the lines in sec . [ kintemp ] . _ ( b ) double - peaked : _ the spectral lines of ch@xmath29ch@xmath1cn and show rather well - defined double - peaked structure . figure [ peaks ] suggests that the peaks corresponding to the two transitions of occur at rather different positions . in fact , they correspond to the two peaks that appear in both transitions , with essentially reversed ratios of relative intensities . the double - peaks of are also not aligned with those of , as emission as a whole is somewhat shifted to the north of the continuum peak and extends more along the east - west direction in contrast to the emission . _ ( c ) irregular shaped : _ the integrated intensity distribution of o@xmath18cs , oc@xmath17s and so@xmath1 appear to be irregularly shaped and do not have enough signal - to - noise in order to show well - defined peaks . the positions shown in figure [ peaks ] correspond to the absolute maxima within the emitting region and represent the approximate location of the enhancement in emission . these should not be taken too literally as peaks having significance comparable to the peaks for the single - peaked species . figure [ peaks ] shows that almost all peak positions to the north - east of the continuum peak correspond to single - peaked emission structure . among the tracers peaking to the south - east , only and so are single - peaked and is double - peaked . the remaining tracers have irregular intensity distribution with absolute maxima at the indicated positions . these two primary locations of the peaks of the spectral lines , ( fig . [ peaks ] ) may either correspond to different emitting clumps or they could be manifestation of the differences in the temperature and density that leads to the variation of the chemical abundances within the same clump . we revisit both these concepts in sec . [ discussion ] . in sec . [ chemabund ] we derive the chemical abundances of the different chemical species at the two nominal peak positions identified by the @xmath30@xmath31a@xmath11 peak to the north - east and the 1211 peak to the south - east of the continuum peak . figure [ velmap ] shows the centroid velocity distribution of some of the selected species for which the signal to noise was sufficient over large areas to generate the maps . with the exception of the hydrogen recombination line and the limited sensitivity line , the center velocity of all molecular lines changes from 56 in the south - west to 62 in the north - east over a spatial scale of 0.08 pc in a direction approximately with a position angle of 40@xmath32 . the direction and magnitude of the velocity gradient match well with other measurements of molecular gas velocity gradients towards g34.26 + 0.15 . using observations at angular resolutions similar to the present dataset @xcite derived a velocity gradient of @xmath33 pc@xmath34 , while @xcite and @xcite observed a velocity gradient of 6@xmath35 on 0.3 to 0.75 pc scales . @xcite derived a velocity gradient of 15@xmath36 on 0.2 to 0.3 pc scales . @xcite observed similar velocity gradients as we find , using only . sio emission primarily tracing outflow activity was not detected at the position of the hot core , it rather appears to surround the hot core @xcite . this implies that shocks do not play an important role in regulating the hot core chemistry in g34.26 + 0.15 . the ionized and molecular gas in g34.26 + 0.15 show rather different velocity gradients , both in direction and magnitude . the hot molecular gas consistently shows this sw - ne velocity gradient , while the ionized gas has a total gradient of @xmath37 in a direction perpendicular to the symmetry axis of the cometary uc h iiregion @xcite . a smaller velocity gradient is detected in the ionized gas parallel to the symmetry axis , _ i.e. _ along the east - west direction . in addition , the ionized gas shows typical velocity widths of @xmath38 @xcite . the hydrogen recombination line , h53@xmath2 shows an emission peak exactly coincident with the continuum peak at @xmath3 mm and a velocity gradient similar to the minor component of the velocity gradient as identified in the ionized gas . similar to all the other recombination lines detected in the region , the h53@xmath2 line has a linewidth much larger than all the other molecular species . these large widths of the recombination lines show conclusively that the ionized gas is driven primarily by the h ii region dynamics . comparison of figs . [ allmap ] and [ velmap ] show that the observed velocity gradient in the hot molecular gas is along the minor axis of the source emission for all species . this is also consistent with the results of @xcite . for a core with significant rotation , the velocity gradient is expected to be along the major axis of the source . this argues strongly against the velocity gradients arising due to the gravitationally bound rotation of a circumstellar slab or disk as was proposed by @xcite . several authors have interpreted the observed velocity distribution of the uc h iiregion g34.26 + 0.2c in terms of mainly two models : the moving star bow shock model and the champagne outflow model . however , @xcite point out that none of these models satisfactorily explain the observed velocities in the region . these authors proposed a model in which the velocities in the region are governed by the stellar winds from the components a and b. further , the ionized gas which is photoevaporated directly from the hot , dense uc molecular core by the exciting star of the cometary h ii region g34.26 + 0.2c , flows from the molecular core toward a region of lower density to the west . in this model , it is further proposed that the molecular core and the components a and b are at a slightly larger distance from the earth than g34.3 + 0.2c . the velocity pattern derived from the various species observed here is consistent with the wind and photoevaporation " model proposed by @xcite , and this further strengthens the conclusions of @xcite , which were derived based only on the velocity patterns of . hot cores are typically proposed to be precursors of high mass stars ( e.g. * ? ? ? the center of the hot core is identified with a collapsing , and rapidly accreting high mass protostar . high angular resolution mid - infrared ( mir ) observations have successfully detected the energizing sources for several hot cores like g11.94 - 0.62 , g45.07 + 0.13 , g29.96 - 0.2 etc . @xcite . using sub - arcsecond resolution centimeter and millimetre molecular line and continuum observations the existence of multiple deeply embedded uc and hypercompact h iiregions contributing to the formation of the hmcs g10.47 + 0.03 and g31.41 + 0.31 have been substantiated @xcite . most of the hmcs thus represent a stage in the evolutionary sequence of massive protostars . however , in the case of the hot molecular gas in g34.26 + 0.15 the existing continuum observations in mid - infrared , far - infrared , sub - millimeter and millimeter have consistently shown the peak of the dust emission to be coincident with the uc h ii region component c and not at the position of the hot core @xcite . in particular , using mir observations with 12 resolution @xcite have reported non - detection of any mir source associated with the hmc in g34.26 + 0.15 . thus g34.26 + 0.15 similar to the orion compact ridge and w3(oh ) @xcite represents the alternative scenario for hot cores , which are externally heated and viewed upon as manifestation of gas shocked and heated by the expanding ionization front and the stellar winds arising from the h ii regions . medium angular resolution mir and fir observations further suggest that the hmc is isolated from the component c of the h ii region and is heated by stellar photons from a and b components and not shocks @xcite . @xcite compared the and c@xmath39o peaks with the peaks seen in arcsecond resolution observations by @xcite and found that none of the peaks coincide with either component c of the h ii region or with each other . these suggest that the hot molecular gas detected in g34.26 + 0.15 traces a layer of a core that is being externally heated by shocks and stellar photons . however , non detection of sio emission from the position of the hot core rules out any significant role played by shocks in determining the hot core chemistry @xcite . the arcsecond resolution continuum and molecular line observations presented in this paper provide further evidence to this proposed model of the hot core in g34.26 + 0.15 being externally heated . the continuum map at 2.8 mm shows a single peak coincident with the radio continuum peak , and it is offset from the peaks of all molecular lines . the molecular line maps extending over a region of 9@xmath2612 ( @xmath40 ) show that ( i ) the emission from different species are not centrally peaked , ( ii ) the molecular peaks are not co - spatial , ( iii ) none of the molecular line emission peaks at the continuum peak and ( iv ) none of the molecular line peaks coincide with the h@xmath1o and oh masers , tracing locations of shocks due to the propagation of the ionization front from the uc h iiregions , detected in the region @xcite . the relative offsets between the peaks in the different molecular line emission further brings out the possibility of there being multiple hmcs within the beam . we note that the ne and the se peaks detected in the g34.26 + 0.15 region are separated along the north - south direction by 08 , which at a distance of 3.7 kpc translates to 0.014 pc . for comparison , the hot core and the compact ridge regions in the orion - kl cloud at a distance of 450 pc are separated by 0.018 pc . we consider the angular resolution of the present observations to be insufficient to either clearly resolve or rule out the possibility of the two peaks being two different cores . these results differ substantially from the single - dish observations which have so far suggested that the emission from the different molecular species originate from within a single bound core . the observations of g34.26 + 0.15 presented here , with the highest angular resolution to date , suggest multiple externally heated hmcs , a hypothesis which is a matter of further study . in the absence of observations of multiple transitions of molecular species at 1resolution , it is not possible to derive the density and temperature distributions appropriate to decipher factors contributing to the relative offsets of the different emission peaks . availability of such temperature and density profiles is also crucial for deriving accurate estimates for the abundance profiles for the detected complex molecular species . existing and recent high angular resolution observations of the hmc associated with g34.26 + 0.15 strongly favour that the source is externally heated . however there are no observations available that constrain ( i ) the geometry of the hot core material with respect to the external heating source ; ( ii ) the density structure of the core material , _ e.g. _ whether it shows a simple gradient in a particular direction or it is clumpy or it is a uniform density swept up region ; or ( iii ) the geometry and multiplicity of the core along the line of sight . we propose the present dataset as a building block for future observations which would help clarify some of the outstanding issues . the rather unconventional " and complicated geometry of the hmc associated with g34.26 + 0.15 poses further challenge to deriving a consistent chemical model for the source . all existing chemical models for hmcs , including those constructed specifically for g34.26 + 0.15 ( e.g. * ? ? ? * ; * ? ? ? * ) explicitly consider a centrally energized spherical cloud forming a massive protostar in two stages involving collapse to high densities ( @xmath41 ) followed by a warm - up phase resulting in the rich chemistry of the hot cores . the models for the hmc in g34.26 + 0.15 are primarily based on single dish observations and assume that the source consists of a hot ultracompact core ( ucc ) with a radius less than 0.025 pc and @xmath42 , a compact core ( cc ) with a radius of about 0.1 pc and a density of 10@xmath8 , surrounded by a massive halo extending out to 3.5 pc with a h@xmath1 density that falls off as r@xmath43 ( e.g. * ? ? ? * and the references therein ) . the latest chemical model for g34.26 + 0.15 by @xcite is more sophisticated in terms of using the density and temperature profiles derived from the dust continuum observations . these density and temperature profiles are subsequently used as inputs for the chemical model . however , the density and temperature distributions are primarily constrained by single - dish continuum observations and the model considers a centrally peaked spherically symmetric structure that is isothermal up to a radius of 0.05 pc ( @xmath44 ) . further it assumes the core of the region to be collapsing to form a massive protostar . most of these assumptions do not conform to the results of high angular resolution molecular line observations presented in this paper and also by @xcite and @xcite . similar arguments may be invoked to show the inappropriateness of the radiative transfer models developed by @xcite for the hmc in g34.26 + 0.15 . in the absence of high resolution multi - line spectroscopic observations suitable to derive the temperature and density distributions , we have adopted a simplistic approach to derive an estimate of the kinetic temperature from the observed brightness temperatures . using the brightness temperatures of oc@xmath17s 98 ( 50 k ) and ocs 98 ( 96 k ) at the position of the intensity peak of oc@xmath17cs ( table [ basic_results ] ) , and the relative abundance of the isotopomers @xmath45/@xmath46 @xmath47 @xcite we derive @xmath48 and @xmath49 . hence ocs 98 is optically thick and its peak brightness temperature ( 167 k ) is a reasonably good estimate of the gas kinetic temperature . we further note that the peak brightness temperature of 3@xmath504@xmath51 a@xmath11 , the brightest and the most optically thick of the methanol lines observed here is 187 k : the peak brightness temperatures of so and are 147 k and 143 k , respectively . thus beyond the calculated high opacity values of ocs 98 , the rather high peak brightness temperatures of 3@xmath504@xmath51 a@xmath11 , so and suggest that all these lines are optically thick and their peak brightness temperatures provide a firm lower limit to the true kinetic temperature of the ucc of g34.26 + 0.15 . based on these , we estimate the kinetic temperature to be 160@xmath52 k and use 160 k for the subsequent analysis . most of the available estimates of kinetic temperature of the hot core region g34.26 + 0.15 are rotation temperatures derived using a variety of molecules . based on an lvg analysis of ch@xmath29cn @xcite derived a gas temperature between 80175 k , while using the k - ladder of the same molecule @xcite estimated a kinetic temperature of @xmath53 k. @xcite derived a rotation temperature of 125 k from c@xmath1h@xmath54oh observations , while @xcite derived a rotation temperature of @xmath55 k from ammonia observations . the errorbars in the kinetic temperatures available from literature are rather large and they are all consistent with our derived value of 160@xmath52 k. the higher angular resolution of our observations better probe the small clumps in contrast to the other observations . we note that the molecules of which rotational transitions were used to derive the kinetic temperatures have similar dipole moments . however , comparison with temperatures derived from single dish observations with the present results is uncertain due to multiple issues : beam dilution , contamination from large scale emission , etc . we have derived the column densities of the different observed species assuming local thermal equilibrium ( lte ) , the kinetic temperature to be 160 k , and the observed transitions ( with the exception of ocs 98 ) to be optically thin . the assumption of a single temperature instead of a temperature profile leads to inaccuracies in our estimates of abundances . however , given the non - centrally peaked geometry of the hot molecular gas as detected in our observations and the lack of dust continuum data , as well as observations of multiple spectral lines constraining the density and temperature profiles , resolving these uncertainties is beyond the scope of this paper . we propose the use of these abundances in combination with upcoming high angular resolution complementary observations in order to provide constraints for future chemical models . based on the peak brightness temperatures presented in table [ basic_results ] we conclude that in all likelihood , so , and @xmath56@xmath57 transitions are also optically thick . in the absence of observations of rarer species , we use the estimated column densities as lower limits and/or guidelines . for both and we adopt the largest among the column densities derived from the different observed transitions . the @xmath30@xmath31 a@xmath58 transition gives the highest column density and it has a peak brightness temperature of 82 k , which suggests that it is still optically thin . the column density of ocs 98 is derived by using the oc@xmath17s column densities and assuming the relative galactic abundance of @xmath45/@xmath46 @xmath47 . table [ colden ] presents the observed column densities and abundances of the different species at the positions of the north - eastern ( ne ) and south - eastern ( se ) peaks , abundances of nh@xmath29 @xcite , @xcite , along with abundances of different species in other prototypical " hot cores . the present dataset does not provide an independent estimate of the total molecular hydrogen column density , , in the ultracompact core ( ucc ; * ? ? ? * ) of g34.26 + 0.15 . estimates of in the core , mostly based on a number of single - dish and interferometric observations of different chemical species , vary by almost an order of magnitude . using the peak nh@xmath29 abundance of 2@xmath2610@xmath59 @xcite and the observed nh@xmath29 column density of 7@xmath2610@xmath39 @xcite , @xcite estimated to be 4@xmath60 . in contrast based on nh@xmath29 @xcite and ch@xmath29cn @xcite observations , the is estimated to be 3@xmath61 . the lower h@xmath1 column densities deduced by @xcite are however consistent with the non - detection of c@xmath39o at the center of the core . following @xcite here we adopt the of the ucc within g34.26 + 0.15 to be 4@xmath2610@xmath62 in order to calculate the observed abundances of the observed chemical species relative to molecular hydrogen ( table [ colden ] ) . the and observations that have angular resolution comparable to our observations justify the adopted value of . the derived column densities of the complex molecules like , , and agree reasonably well with the previous observations ( at 13@xmath268resolution ) by @xcite . following table [ colden ] we find that at the two rather clearly segregated peak positions within the inner regions of the hmc in g34.26 + 0.15 , the abundances of , , , , and so@xmath1 differ by factors @xmath63 . we have detected several of the n- , o- and s - bearing parent and daughter molecules in the g34.26 + 0.15 hot core region . here we discuss the relative abundances of the different species in the light of the abundances observed in a few well - known hot core regions ( table [ colden ] ) . we also compare the abundances relative to a few well - known chemical models for hot cores , these include : chemical models for g34.26 + 0.15 by @xcite and @xcite , both at @xmath64 = 10@xmath65 yr after the grain mantle evaporation ; and the model for the orion compact ridge at @xmath64 = 3.3@xmath66 yr by @xcite . the model by @xcite though not fine - tuned for g34.26 + 0.15 , is relevant because previous chemical study by @xcite suggested that the observed abundances can be explained reasonably well by this model at @xmath67 yr . the chemical model for g34.26 + 0.15 presented by @xcite is essentially an improvement of the model by @xcite in terms of using latest collision rates and chemistry . however for the model by @xcite column densities are available only for the entire source , rather than explicitly for the ucc or the inner regions of the hmc in g34.26 + 0.15 . since , the enhanced angular resolutions of the present bima observations enables us to probe the inner regions of the hmc directly , the column density calculations for the ucc by @xcite are still relevant for our work . further we note the column densities of the entire source ( core+halo ) as estimated by these two models agree reasonably well . figure [ modelcomp ] compares the abundances of the different species relative to the molecular hydrogen abundance as observed at the ne peak of g34.26 + 0.15 with abundances predicted by the different chemical models . in the following sections we discuss the abundances of the o- , n- and s - bearing molecular species separately and compare them with abundances seen in other hot cores . we have detected , and . figure [ modelcomp ] shows that there is a huge discrepancy between the abundances of as predicted by @xcite and @xcite , with the latter model reproducing the observed abundances in g34.26 + 0.15 exactly . all the chemical models predict the relative abundance / to be between 10@xmath68 and 10@xmath43 and do not reproduce the high abundance ratio of 0.1 as observed in most hot cores ( table [ colden ] ) . on the other hand a relative abundance of / @xmath69 is predicted by all the models , which is more or less consistent with the observed range of values in the other hot cores . most of the chemical models for hot cores , including those considered in this paper rely solely on gas - phase chemistry to account for the formation of the large o - bearing complex molecules . however , @xcite have shown that the gas - phase processes are highly inefficient in producing and can never reproduce the observed abundances . recently @xcite has proposed a slower warm - up phase of the hot core and a combination of grain - surface and gas - phase chemistry in the hot cores , which better reproduce the observed abundances of . table [ colden ] suggests that the abundances of the n - bearing species , , , , , and in g34.26 + 0.15 lie within the range of values seen in other hot cores . discovery of ice has lead to the conclusion that is indeed a parent species evaporated from the grain mantle . presence or absence of in the grain mantle plays an important role in determining the abundances of the different daughter species @xcite . figure [ modelcomp ] suggests that while @xcite reproduces the abundance seen in g34.26 + 0.15 ; both @xcite and @xcite fall short by a factor of @xmath21 or more , with the latter model being much worse . , a symmetric top molecule , is likely to form via radiative association of ch@xmath70 with hcn or cn with ch@xmath29 @xcite . the chemical models for the ucc of g34.26 + 0.15 @xcite and the orion compact ridge @xcite predict abundances 3050 times lower than the observed value in g34.26 , while the core+halo model of @xcite predicts an even lower abundance ( fig . [ modelcomp ] ) . the high abundances of nitrile molecules like as found in many hot cores can not be reproduced by gas phase reactions alone . hydrogenation of accreted and to produce that when evaporates , is proposed to be a more viable path for formation in hot cores @xcite . the chemical model for orion compact ridge by @xcite predicts an abundance that is 10 times lower than observed in g34.26 . no predictions for the abundance of are available from the other two chemical models in consideration here . is thought of as a cold - cloud tracer but should also be formed efficiently in the hot gas if c@xmath1h@xmath1 is evaporated from grain mantles @xcite . figure [ modelcomp ] shows that all the chemical models predict an abundance larger by more than a factor of 100 than the value observed in g34.26 . this is consistent with our estimate that is optically thick ( sec . [ kintemp ] ) in g34.26 + 0.15 . is most likely formed by atom addition to hco on grain mantles and then evaporate @xcite . @xcite found to be formed upon uv photolysis and warm - up of h@xmath1o::co:= 100:50:10:110 mixture . however , most of the chemical models so far do not consider either or as parent species and there are no predictions of abundances available so far . based on single - dish observations of g34.26 + 0.15 and subsequent chemical modeling , @xcite inferred that the ocs emission , which requires rather high excitation energy , can not be produced in the extended halo ; the rotation temperature derived from so@xmath1 warrants a core origin ; and the so lines are more likely to arise from the halo . we estimate both ocs and so to be optically thick and to have peak brightness temperatures characteristic of the inner regions . table [ colden ] shows that abundance of ocs as observed in g34.26 + 0.15 is larger than those observed in many hot cores . the abundance of so in g34.26 + 0.15 is similar to that in orion , while sgr b2(n ) and g327.60.3 show much lower abundances . note that the so abundance derived here is likely a lower limit due to the optically thick so line . the so@xmath1 abundance in g34.26 + 0.15 , is similar to the abundance seen in sgr b2(n ) and both are lower by a factor of 10 than the values seen in orion and g327.30.6 . it was initially pointed out by @xcite and @xcite that the relative abundance ratios of so , so@xmath1 and h@xmath1s could be used to estimate the age of the hot cores of the massive protostars . however , @xcite has re - considered this issue and conclude that none of these ratios can be used by itself to estimate the age , since the ratios depend at least as strongly on the physical conditions and on the adopted grain mantle composition as on time . figure [ modelcomp ] shows that all the chemical models under consideration here predict very similar ocs abundances , although somewhat less than that observed in g34.26 + 0.15 . the model by @xcite reproduces the observed so abundance in g34.26 + 0.15 reasonably well , while the other two models predict abundances lower by factors of 510 . since we estimate so in g34.26 + 0.15 to be optically thick , it is hard to reconcile to the higher abundance derived observationally . the model by @xcite reproduces the observed abundances of so@xmath1 for g34.26 + 0.15 , while the models by @xcite and @xcite respectively overestimates and underestimates the observed values . none of the chemical models considered here reproduce the abundances of all the observed species satisfactorily . these models like all the other chemical and radiative transfer models for hmcs assume the energizing source to be at the center , with centrally peaked temperature and density profiles . in addition , there are reasonable arguments in favor , as well as , against the complex molecule formation mechanisms ( grain - mantle evaporation , grain - surface chemistry , gas - phase chemistry ) . thus , from the physical point of view all these models are not directly applicable to g34.26 + 0.15 , which as our observations show is externally heated and additionally might be harboring multiple unresolved hmcs . there are two main uncertainties in our estimates of the column densities of the different species : _ ( i ) _ we have assumed a single kinetic temperature , rather than a temperature profile , the value of which may also be off by 50 k for some of the species and _ ( ii ) _ the estimate is also not accurate . both of these uncertainties can only be addressed with complementary observations at comparable angular resolutions . the errorbars drawn at the 50% level in fig . [ modelcomp ] attempt to quantify these uncertainties in addition to the statistical errors and the errors in absolute calibration . since the difference in abundances between the ne and se peaks of g34.26 + 0.15 never exceeds a factor of 3 and the discrepancy between the observed and predicted abundances are of the order of factors of 10 or more , our conclusions would not have been significantly different had we compared the observed abundances at the se peak with the model predictions . further , we have shown that the abundances observed in g34.26 + 0.15 are not atypical compared to other galactic hot cores . thus , discrepancies by factors of 10 to 100 between the abundances observed in g34.26 + 0.15 and the predictions of the chemical models can not be explained by the estimated errors in abundance calculations . given the inappropriateness of the existing chemical models and uncertainties in the derived relative abundances we consider any age / timescale related discussion for the hmc in g34.26 + 0.15 beyond the scope of this paper . despite the present consensus on the contemporaneous existence of and bearing grain mantles , chemical differentiation in hot core regions is also well - established from observations . such chemical differentiation was prominently noticed in the two clumps in the orion - kl cloud core , very close to the luminous and massive young stellar object , irc2 : the hot core and the compact ridge . while the warmer and denser hot core shows unusually high abundances of h - rich complex n - bearing molecules like ch@xmath1chcn and , the compact ridge is characterized by high abundances of large oxygen - bearing molecules like , and . however , later observations of the orion - kl hot cores by @xcite suggest that the chemical differentiation is probably not as drastic as it was first thought to be . similar chemical differentiation is observed between the two hot cores w3(h@xmath1o ) and w3(oh ) in the w3 region @xcite . these sources are spatially offset by about 0.06 pc and yet they exhibit clear signs of n / o differentiation . both sources show emission from , h@xmath1co , , and , but only the maser source shows significant emission from , , and so2 @xcite , as well as from hcn @xcite and @xcite . several chemical models have partially explained the chemistry of both the hot core & the compact ridge @xcite in orion . @xcite have shown that the evaporation of rich grain mantles inhibit high abundances of o - bearing large molecules , though the already injected alcohols remain abundant for much longer . this leads to the formation of n - rich hot cores . @xcite further showed that in collapsing cores both n - rich and o - rich species tend to co - exist . however , it might be possible to derive the age of the cores based on the relative abundance of the two types of species because is found to be more abundant at earlier times whereas hcn and form later on @xcite . observational detection of large number of daughter species in hot cores suggest that the chemical timescale needs to be shorter than the dynamical timescale , _ i.e. , _ the cores need to be gravitationally supported up to 10@xmath65 yrs @xcite . within the inner regions of g34.26 + 0.15 we find that the n - bearing species tend to peak to the south - east , while the peaks of the o - bearing species are concentrated to the north - east of the continuum peak . the ne and the se peaks are primarily separated along the north - south direction by 08 , which at a distance of 3.7 kpc translates to 0.014 pc . for comparison , the hot core and the compact ridge regions in the orion - kl cloud at a distance of 450 pc are separated by 0.018 pc . thus , given the larger distance to the source and the angular resolution of the present observations , the possibility of the two emission peaks actually belonging to two different cores , one being n - rich and the other being o - rich , can not be ruled out . although the two peak positions in g34.26 do not show large difference in abundances , differences in the abundances of , , and show contrasts quite similar to what is seen in orion - kl @xcite . additional high angular resolution observations of molecular lines are required to derive a better estimate of the temperature distribution , in order to improve the derived abundances at the two peak positions seen within the inner regions of the hmc in g34.26 + 0.15 . we have presented high angular resolution mapping observations of the hmc in g34.26 + 0.15 tracing different chemical species characteristic of hot cores . the higher angular resolution enables us to probe the inner regions of the hot molecular gas . the observations presented here is only a first step and needs to be augmented by observations of standard temperature and density sensitive molecular transitions to constrain the physical attributes of the region which can then be used as constraints for future chemical models . we do not detect any evidence for an energizing source at the center of the hot core , the continuum peak at 2.8 mm is consistent with the free - free radio continuum emission from component c of the uc h iiregion . the intensity distribution of the various molecular tracers do not peak at the same position , the peaks of none of the distributions is located either at the position of the continuum peak or the positions of the h@xmath1o and oh masers . the temperature and density distributions of the inner regions of the hot cores can not be determined from the present observations . based on intensity and velocity distribution , only component c , the most evolved of the h iiregions associated with g34.26 + 0.15 appears to influence the energetics of the hot molecular gas . the kinematics of the hot core as derived from the velocity distribution of the molecular tracers is not strongly influenced by the h ii region dynamics . within the inner regions of the hot core in g34.26 + 0.15 we find that the nitrogen and oxygen - bearing species tend to show a dichotomy and peak at different positions , separated by 08 , the spatial separation being similar to the two hot / warm cores identified in the orion . we propose that as in orion and in w3(oh ) , ( i ) these peak positions may indeed be separate regions of chemical enrichment , resolution of which would require even higher angular resolution observation and ( ii ) may arise due to the external influence of the neighboring h ii regions . we have estimated the abundances of the observed molecular species at the two peak positions assuming a single kinetic temperature of 160 k under conditions of lte . the high angular resolution observations presented here , provides overwhelming evidence in favour of the hot molecular gas at spatial scales of 0.018 pc being externally heated . this together with the clumpiness of the region and possible existence of multiple hmcs in the region , implies geometries more complicated than those considered by the state - of - art chemical models . for the sake of completeness , we follow @xcite to derive a crude upper limit for the luminosity of the internal energizing source if any , using the kinetic temperature derived in this paper and stefan - boltzmann law . the luminosity of the hypothetical and so far undetected internal source is estimated to be 8.9@xmath2610@xmath65 @xmath71 and this corresponds to a o7.5 zams star @xcite . comparison of the abundances observed in g34.26 + 0.15 with currently available somewhat inappropriate chemical models for hot cores by @xcite , @xcite , and @xcite suggest that the nitrogen richness of the region can only be explained by the evaporation of ammonia - rich ice - mantles . however , this implies that the abundances of the oxygenated daughter species be suppressed in the presence of ammonia , which is not the case in g34.26 + 0.15 . this can mean any of the following ( i ) the oxygenated species in question are also formed on grain mantles , ( ii ) the two regions are actually separate from each other or ( iii ) there may be additional gas - phase processes which can progress efficiently even in the presence of ammonia to create high abundances of oxygenated daughter species , which the chemical models do not yet consider . comparison of abundances observed in a few hot cores ( in addition to g34.26 + 0.15 ) and the chemical models show that the chemical models do not yet consistently explain the abundances of all species in any of the hot cores . the complicated geometry of g34.26 + 0.15 makes direct comparison of the observed abundances with the chemical models even more difficult . one of the major caveats of the estimates of the abundances presented here in the hot molecular gas is the assumption of a single kinetic temperature characterizing the emission of the different species . this is a consequence of the lack of complementary observations of molecular species at high angular resolutions to study the temperature and physical structure of the hot molecular gas . in order to understand the true nature of the different peak positions in g34.26 + 0.15 as identified in the bima data and to provide proper physical constraints for the chemical models for the region , sub - arcsecond resolution observations are necessary . this material is based upon work supported by the national science foundation under grant no . this research has made use of nasa s astrophysics data system . lwl acknowledges support from the laboratory for astronomical imaging at the university of illinois and nsf under grant ast-0228953 . akeson , r. l. , & carlstrom , j. e. 1996 , , 470 , 528 avalos , m. , lizano , s. , rodrguez , l. f. , franco - hernndez , r. , & moran , j. m. 2006 , , 641 , 406 beltrn , m. t. , cesaroni , r. , neri , r. , codella , c. , furuya , r. s. , testi , l. , & olmi , l. 2005 , , 435 , 901 beltrn , m. t. , cesaroni , r. , neri , r. , codella , c. , furuya , r. s. , testi , l. , & olmi , l. 2004 , , 601 , l187 bernstein , m. p. , sandford , s. a. , allamandola , l. j. , chang , s. , & scharberg , m. a. 1995 , , 454 , 327 campbell , m. f. , garland , c. a. , deutsch , l. k. , hora , j. l. , fazio , g. g. , dayal , a. , & hoffmann , w. f. 2000 , , 536 , 816 campbell , m. f. , harvey , p. m. , lester , d. f. , & clark , d. m. 2004 , , 600 , 254 carral , p. , & welch , w. j. 1992 , , 385 , 244 caselli , p. , hasegawa , t. i. , & herbst , e. 1993 , , 408 , 548 cesaroni , r. , hofner , p. , walmsley , c. m. , & churchwell , e. 1998 , , 331 , 709 cesaroni , r. 2005 , iau symposium , 227 , 59 charnley , s. b. , tielens , a. g. g. m. , & millar , t. j. 1992 , , 399 , l71 charnley , s. b. 1997 , , 481 , 396 forster , j. r. , & caswell , j. l. 1989 , , 213 , 339 de buizer , j. m. , watson , a. m. , radomski , j. t. , pia , r. k. , & telesco , c. m. 2002 , , 564 , l101 de buizer , j. m. , radomski , j. t. , telesco , c. m. , & pia , r. k. 2003 , , 598 , 1127 doty , s. d. , van dishoeck , e. f. , & tan , j. c. 2006 , , 454 , l5 garay , g. , reid , m. j. , & moran , j. m. 1985 , , 289 , 681 garay , g. , rodriguez , l. f. , & van gorkom , j. h. 1986 , , 309 , 553 garrod , r. t. , & herbst , e. 2006 , , 457 , 927 gaume , r. a. , fey , a. l. , & claussen , m. j. 1994 , , 432 , 648 gibb , e. , nummelin , a. , irvine , w. m. , whittet , d. c. b. , & bergman , p. 2000 , , 545 , 309 hatchell , j. , thompson , m. a. , millar , t. j. , & macdonald , g. h. 1998 , , 338 , 713 hatchell , j. , fuller , g. a. , & millar , t. j. 2001 , , 372 , 281 hatchell , j. , & van der tak , f. f. s. 2003 , , 409 , 589 heaton , b. d. , little , l. t. , & bishop , i. s. 1989 , , 213 , 148 heaton , b. d. , little , l. t. , yamashita , t. , davies , s. r. , cunningham , c. t. , & monteiro , t. s. 1993 , , 278 , 238 henkel , c. , wilson , t. l. , & mauersberger , r. 1987 , , 182 , 137 horn , a. , mllendal , h. , sekiguchi , o. , uggerud , e. , roberts , h. , herbst , e. , viggiano , a. a. , & fridgen , t. d. 2004 , , 611 , 605 hunter , t. r. , neugebauer , g. , benford , d. j. , matthews , k. , lis , d. c. , serabyn , e. , & phillips , t. g. 1998 , , 493 , l97 kim , h .- d . , cho , s .- h . , chung , h .- s . , et al . , 2000 , , 131 , 483 kuchar , t. a. , & bania , t. m. 1994 , , 436 , 117 kurtz , s. , cesaroni , r. , churchwell , e. , hofner , p. , & walmsley , c. m. 2000 , protostars and planets iv , 299 liu , s .- y . , & snyder , l. e. 1999 , , 523 , 683 lovas , f. j. , johnson , d. r. , & snyder , l. e. 1979 , , 41 , 451 macdonald , g. h. , habing , r. j. , & millar , t. j. 1995 , , 224 , 177 macdonald , g. h. , gibb , a. g. , habing , r. j. , & millar , t. j. 1996 , , 119 , 333 mehringer , d. m. , & snyder , l. e. 1996 , , 471 , 897 millar , t. j. 1993 , dust and chemistry in astronomy , 249 millar , t. j. , macdonald , g. h. , & habing , r. j. 1995 , , 273 , 25 millar , t. j. , macdonald , g. h. , & gibb , a. g. 1997 , , 325 , 1163 nomura , h. , & millar , t. j. 2004 , , 414 , 409 nummelin , a. , bergman , p. , hjalmarson , . , friberg , p. , irvine , w. m. , millar , t. j. , ohishi , m. , & saito , s. 2000 , , 128 , 213 panagia , n. 1973 , , 78 , 929 reid , m. j. , & ho , p. t. p. 1985 , , 288 , l17 rodgers , s. d. , & charnley , s. b. 2001 , , 546 , 324 rodgers , s. d. , & charnley , s. b. 2003 , , 585 , 355 sault , r. j. , teuben , p. j. , & wright , m. c. h. 1995 , asp conf . ser . 77 : astronomical data analysis software and systems iv , 77 , 433 sewilo , m. , churchwell , e. , kurtz , s. , goss , w. m. , & hofner , p. 2004 , , 605 , 285 sutton , e. c. , peng , r. , danchi , w. c. , jaminet , p. a. , sandell , g. , & russell , a. p. g. 1995 , , 97 , 455 tielens , a. g. g. m. , & charnley , s. b. 1997 , origins of life and evolution of the biosphere , 27 , 23 turner , b. e. , balick , b. , cudaback , d. d. , heiles , c. , & boyle , r. j. 1974 , , 194 , 279 turner , j. l. , & welch , w. j. 1984 , , 287 , l81 van buren , d. , mac low , m .- m . , wood , d. o. s. , & churchwell , e. 1990 , , 353 , 570 van der tak , f. f. s. 2005 . the chemistry of high - mass star formation . iau symposium 227 , 70 - 79 . van dishoeck , e. f. , & blake , g. a. 1998 , , 36 , 317 wakelam , v. , caselli , p. , ceccarelli , c. , herbst , e. , & castets , a. 2004 , , 422 , 159 watt , s. , & mundy , l. g. 1999 , , 125 , 143 wilson , t. l. , gaume , r. a. , & johnston , k. j. 1993 , , 402 , 230 wilson , t. l. , & rood , r. 1994 , , 32 , 191 wood , d. o. s. , & churchwell , e. 1989 , , 69 , 831 wyrowski , f. , schilke , p. , walmsley , c. m. , & menten , k. m. 1999 , , 514 , l43 wyrowski , f. , hofner , p. , schilke , p. , walmsley , c. m. , wilner , d. j. , & wink , j. e. 1997 , , 320 , l17 cccc frequency & configuration & date & t@xmath72 + ghz & & & k + 87 & a & 23 dec 1999 & 180 + & b & 26 feb 2000 & 450 + 107 & a & 29 jan 2000 & 200 + & b & 10 mar 2000 & 350 + 109 & a & 6 feb 2000 & 250 + & b & 13 mar 2000 & 300 + llllll species & transition & frequency & e@xmath73 t & @xmath74 + & & ghz & k & & + @xmath75 & @xmath30@xmath31 a@xmath76 & 86.6155080 & 102.7 & 0.644 + & @xmath30@xmath31 a@xmath11 & 86.9030180 & 102.7 & 0.644 + & @xmath56@xmath57 a@xmath11 & 107.0138500 & 28.36 & 1.43 + & @xmath27@xmath77 e2 & 107.15992 & 305 & 1.24 + @xmath78 & @xmath79@xmath80 e & 103.46659 & 24.66 & 19.992 + @xmath81 & @xmath82@xmath83 ea+ae & 105.7683438 & 86.02 & 5.405 + @xmath84 & @xmath85@xmath86 & 105.972601 & 27.2 & 4.2 + @xmath87 & 98 & 106.7873889 & 25.7 & 9.0 + @xmath88 & 9 - 8 & 109.1108477 & 26.2 & 9.0 + ocs & 98 & 109.4630630 & 26.28 & 9.0 + @xmath89 & @xmath90@xmath91 & 107.0435270 & 37.91 & 11.662 + & @xmath92@xmath93 & 109.6502630 & 35.42 & 11.906 + @xmath94 & @xmath95@xmath96 & 107.0602085 & 369.5 & 3.102 + @xmath97 & 1211 & 109.1736340 & 34.07 & 12.0 + so & @xmath98@xmath99 & 109.2522200 & 21.05 & 1.510 + lccccccc line & beam & @xmath100 & @xmath101 & @xmath19 & @xmath20 & rms + & ( arcsec ) & ( jy beam@xmath34 ) & ( k ) & ( ) & ( ) & ( mjy beam@xmath34 ) + h53@xmath2 & 148@xmath26125 & 0.454 & 42 & @xmath102 & @xmath103 & 66.3 + @xmath75 @xmath30@xmath31 a@xmath76 & 148@xmath26125 & 0.934 & 82 & @xmath104 & @xmath105 & 69.1 + @xmath75 @xmath30@xmath31 a@xmath11 & 148@xmath26125 & 0.952 & 83 & @xmath106 & @xmath107 & 69.6 + @xmath78 @xmath79@xmath80 e & 102@xmath26085 & 0.240 & 31 & @xmath108 & @xmath109 & 35.6 + @xmath81 @xmath82@xmath83 & 102@xmath26085 & 0.218 & 26 & 49.3 & & 55.1 + @xmath84 @xmath85@xmath86 & 102@xmath26085 & 0.383 & 46 & @xmath110 & @xmath111 & 106 . + @xmath87 98 & 102@xmath26085 & 0.394 & 50 & @xmath112 & @xmath113 & 86.1 + @xmath75 @xmath56@xmath57 a@xmath11 & 102@xmath26085 & 1.48 & 187 & @xmath114 & @xmath115 & 44.9 + @xmath89 @xmath90@xmath91 & 102@xmath26085 & 0.365 & 46 & @xmath116 & @xmath117 & 45.3 + @xmath94 @xmath95@xmath96 & 102@xmath26085 & 0.198 & 25 & @xmath118 & @xmath119 & 45.3 + @xmath75 @xmath27@xmath77 e2 & 102@xmath26085 & 0.643 & 81 & @xmath120 & @xmath121 & 94.9 + @xmath88 9 - 8 & 099@xmath26089 & 0.165 & 20 & 55.6 & & 67.5 + @xmath97 1211 & 099@xmath26089 & 1.23 & 143 & @xmath122 & @xmath123 & 64.3 + so @xmath98@xmath99 & 099@xmath26089 & 1.27 & 147 & @xmath124 & @xmath125 & 123.1 + ocs 98 & 099@xmath26089 & 1.44 & 167 & @xmath114 & @xmath126 & 127.0 + @xmath89 @xmath92@xmath93 & 099@xmath26089 & 0.360 & 42 & @xmath127 & @xmath128 & 65.2 + lllllllll & 3.4(17 ) & 8.5(-7 ) & 2.6(17 ) & 6.4(-7 ) & 1.4(-7 ) & 4.0(-7 ) & 2.0(-7 ) & 2.0(-5 ) + & 2.9(16 ) & 7.3(-8 ) & 2.7(16 ) & 6.8(-8 ) & 1.4(-8 ) & 3.0(-8 ) & 1.0(-9 ) & 2.0(-6 ) + & 5.7(16 ) & 1.4(-7 ) & 3.4(16 ) & 8.5(-8 ) & 8.0(-9 ) & 1.9(-8 ) & 3.0(-9 ) & 3.4(-7 ) + & 7.0(18 ) & 1.8(-5 ) & & & 6.7(-8 ) & 5.7(-7 ) & & + & 1.3(16 ) & 3.3(-8 ) & & & 4.0(-9 ) & 5.0(-9 ) & 3.0(-8 ) & 7.0(-7 ) + & 1.3(15 ) & 3.2(-9 ) & 2.7(15 ) & 6.8(-9 ) & 3.0(-9 ) & 5.0(-9 ) & 6.0(-10 ) & 4.0(-7 ) + & 5.1(13 ) & 1.3(-10 ) & 1.3(14 ) & 3.3(-10 ) & 1.8(-9 ) & 6.0(-9 ) & 5.0(-9 ) & 3.0(-11 ) + nh@xmath1cho & & & 2.5(15 ) & 6.2(-9 ) & 1.4(-10 ) & 3.0(-10 ) & 2.0(-10 ) & 2.0(-8 ) + so@xmath1 & 5.8(15 ) & 1.4(-8 ) & 1.5(16 ) & 3.7(-8 ) & 1.2(-7 ) & 1.6(-7 ) & 3.0(-8 ) & 2.0(-7 ) + oc@xmath17s & 1.8(15 ) & 4.4(-9 ) & 2.2(15 ) & 5.6(-9 ) & & & 3.0(-10 ) & 1.0(-9 ) + o@xmath18cs & 2.0(15 ) & 5.1(-9 ) & 1.4(15 ) & 3.5(-9 ) & & & 7.0(-10 ) & 7.0(-10 ) + ocs & 3.9(16 ) & 9.7(-8 ) & 4.9(16 ) & 1.2(-7 ) & 1.1(-8 ) & 3.0(-8 ) & @xmath1292.0(-9 ) & @xmath1292(-9 ) + so & 4.4(16 ) & 1.1(-7 ) & 5.3(16 ) & 1.3(-7 ) & 1.9(-7 ) & 3.0(-7 ) & 2.0(-8 ) & @xmath1293(-9 ) + mm continuum image the contour levels are -3 , 3 , 10 , 20 , 30 , 40 , 50 , 70 , 90 , 100 , 120 , 140 of the noise level of 20 mjy beam@xmath34 . for the @xmath130 cm continuum image the contour levels are at -3 , 3 , 5 , 10 , 20 , 30 , 40 , 80 , 120 , 180 , 250 , 350 , 450 , 600 , 800 , 850 times the noise level of 0.3 mjy beam@xmath34 . the synthesized beams are shown on the right - hand corner of the plot . the synthesized beam at @xmath3 mm is 100@xmath26085 with p.a . = 35 and the beam at @xmath130 cm is 048@xmath26041 with p.a.=21 . [ cont_cont],width=283 ] . the noise levels ( in jy beam@xmath34 ) are : _ ( a ) _ 0.58 , _ ( b ) _ 0.40 , _ 0.26 , _ ( d ) _ 0.45 , _ ( e ) _ 0.61 , _ ( f ) _ 0.58 , _ ( g ) _ 0.30 , _ ( h ) _ 0.48 , _ ( i ) _ 30 , _ ( j ) _ 0.24 , _ ( k ) _ 0.50 , _ ( l ) _ 0.35 , _ ( m ) _ 0.58 , _ ( n ) _ 0.22 , _ ( o ) _ 0.46 and _ ( p ) _ 0.32 . the contour levels in panels _ ( a ) , ( f ) , ( g ) , ( h ) , ( i ) , ( j ) , ( m ) , ( n ) _ and _ ( o ) _ are in steps of the noise level , starting from twice the noise level . the contour levels in the remaining panels are in steps of twice the noise level , starting from twice the noise level . the negative contours are not shown to simplify the image . the synthesized beams are shown on the left hand corner of each panel . the filled @xmath131 mark the positions the components a , b , and c @xcite of the uc h ii region g34.3 + 0.2 . the squares and diamonds signify positions of h@xmath1o and oh masers respectively @xcite . [ allmap],width=566 ] mm . the symbols with additional circles around them denote that these correspond to single - peaked intensity distributions . the errorbars on the left hand bottom corner show the positional uncertainty . [ peaks],width=377 ] . the errorbars correspond to 50% errors in abundance estimates . @xcite corresponds to the chemical model for the orion compact ridge at @xmath132 yr . @xcite corresponds to the chemical model for the ultracompact core , for which = 2.7@xmath133 at @xmath67 yr . @xcite corresponds to a chemical model for the entire source g34.26 + 0.15 , for a line of sight through the center of the core at @xmath67 yr , with = 3.6 10@xmath134 . [ modelcomp],width=377 ]	we present high angular resolution ( @xmath0 ) multi - tracer spectral line observations toward the hot core associated with g34.26 + 0.15 between 87109 ghz . we have mapped emission from ( i ) complex nitrogen- and oxygen - rich molecules like , , , , , ; ( ii ) sulfur - bearing molecules like ocs , so and so@xmath1 ; and ( iii ) the recombination line h53@xmath2 . the high angular resolution enables us to directly probe the hot molecular core associated with g34.26 + 0.15 at spatial scales of 0.018 pc . at this resolution we find no evidence for the hot core being internally heated . the continuum peak detected at @xmath3 mm is consistent with the free - free emission from component c of the ultracompact h ii region . velocity structure and morphology outlined by the different tracers suggest that the hot core is primarily energized by component c. emission from the n- and o - bearing molecules peak at different positions within the innermost regions of the core ; none are coincident with the continuum peak . lack of high resolution complementary datasets makes it difficult to understand whether the different peaks correspond to separate hot cores , which are not resolved by the present data , or manifestations of the temperature and density structure within a single core . based on the brightness temperatures of optically thick lines in our sample , we estimate the kinetic temperature of the inner regions of the hmc to be @xmath4 k. comparison of the observed abundances of the different species in g34.26 + 0.15 with existing models does not produce a consistent picture . there are uncertainties due to : ( i ) the unavailability of temperature and density distribution of the mapped region within the hot core , ( ii ) typical assumption of centrally peaked temperature distribution for a hot core with an accreting protostar at the center , by the chemical models , an aspect not applicable to externally heated hot cores like g34.26 + 0.15 and ( iii ) inadequate knowledge about the formation mechanism of many of the complex molecules .
cake cutting is one the most fundamental topics in fair division ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? it concerns the setting in which a cake is represented by an interval @xmath0 $ ] and each of the @xmath1 agents has a value function over the cake . the main aim is to divide the cake fairly . the framework is general enough to encapsulate the important problem of allocating a heterogeneous divisible good among multiple agents with different preferences . the cake cutting problem applies to many settings including the division of rent among housemates , disputed land between land - owners , and work among co - workers . it is especially useful in scheduling the use of a valuable divisible resource such as server time . within the cake cutting literature , the most important criteria of a fair allocation are _ envy - freeness _ and _ proportionality_. in an envy - free allocation , each agent considers his allocation at least as good as any other agent s allocation . an envy - free allocation is guaranteed to exist ( see * ? ? ? * ; * ? ? ? . in a proportional allocation , each agent gets at least @xmath2 of the value he assigns to the cake . a desirable aspect of envy - freeness is that it implies proportionality . computation of a fair allocation of cake is one of the fundamental problems in algorithmic economics . @xcite designed an envy - free cake cutting algorithm for an arbitrary number of players . although their algorithm is guaranteed to eventually terminate , its running time is unbounded . moreover , the algorithm can divide the cake into infinitely small segments . since the result of @xcite , researchers have examined restricted value density functions and proposed envy - free algorithms for them . in order to ascertain the running time of a cake cutting algorithm , it is important to know the computational model and input to the problem . in most of the literature ( see * ? ? ? * ) , it is assumed that the value an agent ascribes to any segment of the cake can be queried or evaluated via an oracle . while the classical literature uses this query model , computer scientists recently looked at the problem from the point of view of full report , as is common in mechanism design . throughout the paper we focus on _ piecewise constant value density functions _ and _ piecewise uniform value density functions_. piecewise constant value density functions are one of the most fundamental class of value functions . piecewise uniform valuations are a restricted class of piecewise constant valuations . @xcite presented a discrete , strategyproof , polynomial - time , envy - free and pareto optimal algorithm for piecewise uniform valuations . they stated that generalizing their results for piecewise constant valuations is an open problem . they also presented an envy - free and proportional algorithm that satisfies strategyproofness by resorting to randomization . in this paper , we consider three of the most enduring goals in mechanism design and fair division : fairness , pareto optimality and strategyproofness . since many fair division algorithms need to be deployed on a large scale , we will also aim for algorithms that are _ computationally _ efficient . our main research question in this paper is as follows : _ among the different levels of fairness , pareto optimality , strategyproofness , and efficient computability , what are the maximal set of properties that can be satisfied simultaneously for piecewise constant and piecewise uniform valuations ? _ our main contribution is a detailed study of this question including the formulation of a number of desirable cake cutting algorithms satisfying many of the properties . in the case where pareto optimality can not be satisfied , we also consider a weaker notion of efficient called _ non - wastefulness_. non - wastefulness dictates that every portion of the cake that is desired by at least one agent is allocated to some agent who desires it . for fairness , we not only consider the standard notions envy - freeness and proportionality but we also propose the concept of _ robust fairness _ in particular _ robust envy - freeness _ and _ robust proportionality_. the main idea of an allocation being robust envy - free is that even if an agent readjusts his value density function , as long as the ordinal information of the function is unchanged , then the allocation remains envy - free . the main advantages of robust envy - freeness are less information requirements and envy - freeness under uncertainty . for strategic properties , we consider three notions of truthfulness namely strategyproofness , weak group - strategyproofness and group - strategyproofness . in most of the cake - cutting literature , an algorithm is considered ` strategyproof ' if truth - telling is a maximin strategy @xcite and it need not be dominant strategy incentive compatible . when we refer to strategyproofness , we will mean the reporting the truthful valuations is a dominant strategy . this stronger notion of strategyproofness has largely been ignored in cake - cutting barring a few recent exceptions @xcite . we present _ ccea ( controlled cake eating algorithm ) _ for piecewise constant valuations and show that it is polynomial - time and robust envy - free and robust proportional . ccea depends on a reduction to the generalizations @xcite of the _ ps ( probabilistic serial ) _ algorithm introduced by @xcite in the context of random assignments . the algorithm relies on solving the parametric network flows ( see * ? ? ? we show that the algorithm can handle variable claims and private endowments for piecewise constant valuations and also satisfies group - strategyproofness under piecewise uniform valuations . if we insist on pareto optimality , then we show that there exists an algorithm which we refer to as the _ mea ( market equilibrium algorithm ) _ that is discrete , polynomial - time pareto optimal , envy - free , and proportional for piecewise constant valuations . the algorithm relies on the walras equilibrium formulation of @xcite for finding an @xmath3-envy - free for general cake cutting valuations and the result of @xcite that market equilibrium for fischer markets with linear utilities can be computed in polynomial time . both ccea and mea not only coincide on piecewise uniform valuations but are also group - strategyproof . although ccea and mea are desirable algorithms , they are not strategyproof for piecewise constant valuations . we present another algorithm called _ csd ( constrained serial dictatorship ) _ which is strategyproof in expectation , robust proportional , and satisfies unanimity . for the important case of two agents , it is polynomial - time , and robust envy - free . to the best of our knowledge , it is the first cake cutting algorithm for piecewise constant valuations that satisfies strategyproofness , proportionality , and unanimity at the same time . csd requires some randomization to achieve strategyproof in expectation . however , csd is discrete in the sense that it gives the same utility guarantee ( with respect to the reported valuation functions ) over all realizations of the random allocation . although csd uses some essential ideas of the well - known _ serial dictatorship _ rule for discrete allocation , it is significantly more involved . our main technical results are as follows . [ th : maintheorem1 ] for cake cutting with piecewise constant valuations , there exists an algorithm ( ccea ) that is discrete , polynomial - time , robust envy - free , and non - wasteful . [ th : maintheorem3 ] for cake cutting with with piecewise constant valuations , there exists an algorithm ( mea ) that is discrete , polynomial - time , pareto optimal , and envy - free . [ th : maintheorem2 ] for cake cutting with piecewise uniform valuations , there exists algorithms ( ccea and mea ) that are discrete , group strategyproof , polynomial - time , robust envy - free and pareto optimal . [ th : maintheorem1b ] for cake cutting with piecewise constant valuations , there exists a randomized implemention of an algorithm ( csd ) that is ( ex post ) robust proportional , ( ex post ) symmetric , and ( ex post ) unanimous and strategyproof in expectation . for two agents , it is polynomial - time , robust proportional and robust envy - free . our positive results are complemented by the following impossibility results . these impossibility results suggest the properties satisfied by ccea and mea are maximal subsets of properties that can be satisfied by any algorithm for piecewise constant valuation profiles . [ pe fair impossibility ] for piecewise constant valuation profiles with at least two agents , there exists no algorithm that is both pareto optimal and robust proportional . [ pe sp prop impossibility ] for piecewise constant valuation profiles with at least two agents , there exists no algorithm that is strategyproof , pareto optimal , and proportional . [ sp rob prop non w impossibility ] for piecewise constant valuation profiles with at least two agents , there exists no algorithm that is strategyproof , robust proportional , and non - wasteful . as a consequence of ccea and mea , we generalize the positive results regarding piecewise uniform valuations in @xcite and piecewise constant valuations in @xcite in a number of ways such as handling richer cake cutting settings , handling more general valuations functions , achieving a stronger envy - free concept , or a stronger strategyproofness notion . moreover , we prove that ccea and mea two different algorithms are equivalent in the domain of piecewise constant valuations . furthermore , we show that both ccea and mea are generalizations of the main mechanism in @xcite . we also show which combinations of properties are impossible to satisfy simultaneously . some of our main results are summarized in table [ table : summary : cake ] . after presenting our main results , we enrich the cake cutting domain in two ways . we allow agents to have initial endowments of the cake . moreover we consider the more general setting in which agents may have different claims or entitlements to the cake . we show that many of our results carry over to these more general settings . we consider a cake which is represented by the interval @xmath0 $ ] . a _ piece of cake _ is a finite union of disjoint subintervals of @xmath0 $ ] . the length of an interval @xmath4 $ ] is @xmath5 . as usual , the set of agents is @xmath6 . each agent has a piecewise continuous _ value density function _ @xmath7\rightarrow [ 0,\infty]$ ] . the value of a piece of cake @xmath8 to agent @xmath9 is @xmath10 . as generally assumed , valuations are non - atomic ( @xmath11)=0 $ ] ) and additive : @xmath12 where @xmath8 and @xmath13 are disjoint . the basic cake cutting setting can be represented by the set of agents and their valuations functions , which we will denote as _ a profile of valuations_. in this paper we will assume that each agent s valuation function is private information for the agent that is not known to the algorithm designer . each agent reports his valuation function to the designer and the designer then decides how to make the allocations based on the reported valuations . later on we will also consider two important extensions of cake cutting : claims and private endowments . we will assume that agents have the following _ claims _ on the cake respectively : @xmath14 . in the original cake cutting problem agents have equal claims . each agent @xmath15 has a _ private endowment _ @xmath16 which is a segment of the cake privately owned by @xmath9 . the cake is assembled by joining the pieces @xmath17 . therefore the cake cutting setting in its full generality can be represented as a quadruple @xmath18 . an allocation is a partitioning of the cake into @xmath1 pieces of cake @xmath19 such that the pieces are disjoint ( aside from the interval boundaries ) and @xmath20 is allocated to agent @xmath15 . a cake cutting algorithm takes as input @xmath18 and returns an allocation . in this paper we will only consider _ piecewise uniform _ and _ piecewise constant _ valuations functions . a function is _ piecewise uniform _ if the cake can be partitioned into a finite number of intervals such that for some constant @xmath21 , either @xmath22 or @xmath23 over each interval . a function is _ piecewise constant _ if the cake can be partitioned into a finite number of intervals such that @xmath24 is constant over each interval . in order to report his valuation function to the algorithm designer , each agent will specify a set of points @xmath25 that represents the consecutive points of discontinuity of the agent s valuation function as well as the constant value of the valuation function between every pair of consecutive @xmath26 s . for a function @xmath24 , we will refer by @xmath27\}$ ] as the set of density functions _ ordinally equivalent _ to @xmath24 . note that if an algorithm takes as input @xmath24 and returns the same output for any @xmath28 , then it is oblivious to the exact cardinal information of @xmath24 . an _ allocation _ is a partition of the interval @xmath0 $ ] into a set @xmath29 , where @xmath1 is the number of agents and @xmath20 is a piece of cake that is allocated to agent @xmath9 . and @xmath30 is the piece of the cake that is not allocated . all of the fairness and efficiency notations that we will discuss next are with respect to the reported valuation functions . in an _ envy - free allocation _ , we have @xmath31 for each pair of agent @xmath32 , . an allocation is _ individually rational _ if @xmath33 . in a _ proportional _ allocation , each agent gets at least @xmath2 of the value he has for the entire cake . an allocation satisfies _ symmetry or equal treatment of equals _ if any two agents with identical valuations get same utility . clearly , envy - free implies proportionality and also symmetry . an allocation @xmath8 is _ pareto optimal _ if no agent can get a higher value without some other agent getting less value . formally , @xmath8 is pareto optimal if there does not exists another allocation @xmath13 such that @xmath34 for all @xmath15 and @xmath35 for some @xmath15 . for any @xmath36 $ ] , define @xmath37 . an allocation @xmath8 is _ non - wasteful _ if for all @xmath36 $ ] , @xmath38 . in other words , an allocation is non - wasteful if every portion of the cake desired by at least one agent is allocated to some agent who desires it . we now define robust analogues of the fairness concepts defined above . an allocation satisfies _ robust proportionality _ if for all @xmath32 and for all @xmath28 , @xmath39 . an allocation satisfies _ robust envy - freeness _ if for all @xmath32 and for all @xmath28 , @xmath40 . notice that both robust envy - freeness and robust proportionality would require each agents to get a piece of cake of the same length if every agent desires the entire cake . we give an example of piecewise constant value density function and demonstrate how the standard concept of envy - freeness is not robust under uncertainty . [ example : piecewise ] consider the cake cutting problem in figure [ figure : piecewise ] . an allocation in which both agents get regions in which their value density function is the highest is envy - free . agent 1 gets utility one for his allocation and has the same utility for the allocation of agent 2 . however if its probability density function is slightly lower in region @xmath41 $ ] , then agent 1 will be envious of agent 2 . @xmath42 \centering \draw[- ] ( 0,-.5 ) -- ( 0,15 ) ; \draw[- ] ( -.5,0 ) -- ( 20,0 ) ; \draw[- ] ( 20,15 ) -- ( 20,0 ) ; \fill[pattern = vertical lines , ] ( 0,0 ) rectangle ( 2,15 ) ; \fill[pattern = vertical lines ] ( 10,0 ) rectangle ( 20,5 ) ; \fill[pattern = horizontal lines ] ( 6,0 ) rectangle ( 20,7.5 ) ; \draw ( 2,-.6 ) node(c){\small $ 0.1 $ } ; \draw ( 6,-.6 ) node(c){\small $ 0.3 $ } ; \draw ( 20/2,-.6 ) node(c){\small $ 0.5 $ } ; \draw ( 20,-.6 ) node(c){\small$1 $ } ; \draw(-.7,15 ) node(z){\small $ 10 $ } ; \draw(-.7,7.5 ) node(z){\small $ 3 $ } ; \draw(-.7,5 ) node(z){\small $ 2 $ } ; \end{tikzpicture}\ ] ] @xmath43 \centering \draw[- ] ( 0,-.5 ) -- ( 0,15 ) ; \draw[- ] ( -.5,0 ) -- ( 20,0 ) ; \draw[- ] ( 20,15 ) -- ( 20,0 ) ; \fill[pattern = vertical lines , ] ( 0,0 ) rectangle ( 2,10 ) ; \fill[pattern = vertical lines ] ( 10,0 ) rectangle ( 20,5 ) ; \fill[pattern = horizontal lines ] ( 6,0 ) rectangle ( 20,10 ) ; \draw ( 2,-.6 ) node(c){\small $ 0.1 $ } ; \draw ( 6,-.6 ) node(c){\small $ 0.3 $ } ; \draw ( 20/2,-.6 ) node(c){\small $ 0.5 $ } ; \draw ( 20,-.6 ) node(c){\small$1 $ } ; \draw(-.7,15 ) node(z){\small $ 10 $ } ; \draw(-.7,7.5 ) node(z){\small $ 3 $ } ; \draw(-.7,5 ) node(z){\small $ 2 $ } ; \end{tikzpicture}\ ] ] a _ deterministic cake cutting algorithm _ is a mapping from the set of valuation profiles to the set of allocations . a _ randomized cake cutting algorithm _ is a mapping from the set of valuation profiles to a space of distributions over the set of allocations . the output of the algorithm in this case for a specific valuation profile is a random sample of the distributional function over the set of allocation for that profile . an algorithm ( either deterministic or randomized ) satisfies property @xmath44 if it always returns ( with probability 1 ) an allocation that satisfy property @xmath44 . a deterministic algorithm is _ strategyproof _ if no agent ever has an incentive to misreport in order to get a better allocation . the notion of strategyproofness is the one well - established in social choice ( see * ? ? ? * ) and much stronger than the notion of ` strategyproofness ' used in some of the cake - cutting literature ( see * ? ? ? * ) . by strategyproof , we mean truthful as has been used in @xcite . similarly , a deterministic algorithm is _ group - strategyproof _ if there exists no coalition @xmath45 such that members of @xmath46 can misreport their preferences so that each agent in @xmath46 gets as preferred an allocation and at least one agent gets a strictly better payoff . a deterministic algorithm is _ weak group - strategyproof _ if there exists no coalition @xmath45 such that members of @xmath46 can misreport their preferences so that each agent in @xmath46 gets a strictly more preferred allocation . a randomized algorithm is _ strategyproof in expectation _ if the expected utility from the random allocation that every agent receives in expectation under a profile where he reported truthfully is at least as large as the expected that he receives under a profile where he misreports while fixing the other agents reports . we say that a cake cutting algorithm satisfies _ unanimity _ , if when each agent s most preferred @xmath2 length of the cake is disjoint from another agent , @xmath2 length of the cake , then each agent gets their most preferred piece of cake of length @xmath2 . in this subsection , we recap the main properties of cake cutting algorithms : proportionality , robust proportionality , envy - freeness , robust envy - freeness , symmetry , non - wastefulness , pareto optimality , unanimity strategyproofness , weak group strategyproofness , group strategyproofness , polynomial - time . for cake cutting , envy - freeness and non - wastefulness @xmath47 proportionality ; robust proportionality @xmath47 proportionality ; robust envy - freeness @xmath47 envy - freeness ; robust envy - freeness and non - wastefulness @xmath47 robust proportionality ; group strategyproofness @xmath47 weak group strategyproofness @xmath47 strategyproofness ; pareto optimality @xmath47 non - wastefulness ; pareto optimality @xmath47 unanimity ; two agents , proportionality @xmath47 envy - freeness ; two agents , robust proportionality @xmath47 robust envy - freeness . we may assume without lost of generality that every part of the cake is desired by at least one agent . if that is not the case , then we can discard the parts that are desired by no one and rescale what is left so that we get a @xmath0 $ ] interval representation of the cake . notice that this procedure preserves the aforementioned properties of fairness , efficiency and truthfulness . the free disposal assumption that we are making is necessary to ensure strategyproofness for piecewise uniform valuation functions . see @xcite for a discussion on the necessity of this assumption . + before we present our algorithms , we will first take a detour to the literature on random assignments , as some of the algorithms in the random assignment literature are closely related to our algorithms . an assignment problem is a triple @xmath48 such that @xmath6 is a set of agents , @xmath49 is a set of houses and @xmath50 is the preference profile in which @xmath51 denotes the preferences of agent @xmath9 over houses @xmath52 . a _ deterministic assignment _ is a one - to - one mapping from @xmath53 to @xmath52 . a _ random allocation _ is a probability distribution over @xmath52 . a random assignment @xmath54 gives a random allocation to each agent . it can be represented by a bistochastic matrix @xmath54 in which the @xmath9th row is denoted by @xmath55 and all @xmath56 , and @xmath57 , @xmath58 , @xmath59 . holds when there are the same number of agents as there are objects , which can be assumed without lost of generality by adding dummy agents or objects ] the term @xmath60 denotes the probability with which agent @xmath9 gets house @xmath61 . an assignment problem has commonalities with cake cutting with piecewise constant valuations . they also have some fundamental differences . for example , in cake cutting , the agents do not have continuous constant valuations over pre - determined segments of the cake . given two random assignments @xmath54 and @xmath62 , @xmath63 i.e. , a player @xmath9 weakly sd prefers @xmath55 to @xmath64 if for all @xmath61 , @xmath65 another way to see the sd relation is as follows . a player @xmath9 weakly sd prefers allocation @xmath55 to @xmath64 if for all vnm utilities consistent with his ordinal preferences , @xmath9 gets at least as much expected utility under @xmath55 as he does under @xmath64 . furthermore @xmath66 , i.e. , @xmath54 _ stochastically dominates _ @xmath62 if @xmath63 for all @xmath15 and @xmath67 . an algorithm satisfies _ sd - efficiency _ if each returned assignment is pareto optimal with respect to the sd - relation ( see * ? ? ? an algorithm satisfies _ sd envy - freeness _ if each agent ( weakly ) sd prefers his allocation to that of any other agent . sd envy - freeness is a very demanding notion of fairness . the reader may be able to notice that our notion of robust envy - freeness in cake cutting is based on a similar idea as sd envy - freeness . we will consider random allocations as fractional allocations and random assignments as fractional assignments . viewing the probability of getting a house simply as getting a fraction of the house is especially useful when some houses are not complete but only partial . in this vein , the definition of sd dominance should also be considered from the perspective of fractional allocations rather than probability distributions . the most basic assignment problem concerns @xmath1 agents having strict preferences over @xmath1 objects . for this basic setting a simple yet ingenious _ ps ( probabilistic serial ) _ algorithm introduced by @xcite and @xcite which uses the _ simultaneous eating algorithm ( sea)_. each house is considered to have a divisible probability weight of one , and agents simultaneously and with the same eating rate consume the probability weight of their most preferred house until no house is left . the random allocation allocated to an agent by ps is then given by the amount of each object he has eaten until the algorithm terminates . the main result of @xcite was that the fractional assignment returned by the ps algorithm is sd envy - free and sd - efficient . the ps algorithm has been extended in various ways . the eps ( extended ps algorithm ) of @xcite generalized ps to the case for indifferences using parametric network flows . eps also generalized the _ egalitarian rule _ of @xcite for dichotomous preferences . @xcite and @xcite extended the work of @xcite to propose ps generalization which also takes care of private endowments where @xmath68 indicates the endowment of agent @xmath9 . for the case of endowments , @xcite introduced the idea of justified envy - freeness . an assignment @xmath54 satisfies justified envy - freeness if for all @xmath32 , @xmath69 or @xmath70 the algorithms in @xcite satisfy justified envy - freeness in the presence of private endowments . in our algorithm ccea , we rely on the full power of the controlled - consuming ( cc ) algorithm of @xcite which combines almost all the desirable features of other extensions of ps . in particular , we use the following fact . _ there exists an extension of the ps algorithm which can simultaneously handle indifferences in preferences , unacceptable objects in preferences , allocation of multiple objects to agents , agent owning fractions of houses , partial houses being available , and still returns an assignments which satisfies sd justified envy - freeness and sd - efficiency . _ in addition , if there are no private endowments , then the extension can also handle variable eating rates . the controlled - consuming ( cc ) algorithm of @xcite can handle the case where each agent owns fractions of the complete houses . we also require that for some houses , only an arbitrary fraction of the house is available to the market . this can be handled by a modification to cc ( page 30 , * ? ? ? finally we require the agents to want to be allocated as many houses as possible . this does not require any modification to cc . in the absence of endowments but presence of variable eating rates , cc is equivalent to the eps algorithm that can also cater for variable eating rates ( section 6.4 , * ? ? ? * ) . ccea is based on cc ( controlled consuming ) algorithm of @xcite . since the original ps algorithm utilized the simultaneous eating algorithm , hence the name _ controlled cake eating algorithm . _ ccea first divides the cake up into disjoint intervals each of whose endpoints are consecutive points of discontinuity of the agents valuation functions . we will refer to these intervals as _ intervals induced by the discontinuity points_. the idea is to form a one - to - one correspondence of the set of cake intervals with a set of houses of an assignment problem . since intervals may have different lengths , we consider the house corresponding to the interval with the maximum length as a complete house where as intervals corresponding to other houses are partial houses . the preferences of agents over the houses are naturally induced by the relative height of the piecewise constant function lines in the respective intervals . if an agent @xmath9 owns a sub - interval @xmath71 , then in the housing setting , @xmath72 is set to @xmath73 and not to one . the reason is that an agent can only own as much of the house as exists . the technical heart of the algorithm is in cc ( controlled consuming ) algorithm of @xcite . we recommend the reader to section 3.2 of @xcite in which an illustrative example on cc is presented . once cc has been used to compute a fractional assignment @xmath54 , it is straightforward to compute a corresponding cake allocation . if an agent @xmath9 gets a fraction of house @xmath74 , then in the cake allocation agent @xmath9 gets the same fraction of subinterval @xmath71 . piecewise constant value functions . a robust envy - free allocation . divide the regions according to agent value functions . let @xmath75 be the set of subintervals of @xmath0 $ ] formed by consecutive marks . consider @xmath76 where * @xmath77 where @xmath78 for all @xmath79 * @xmath80 is defined as follows : @xmath81 if and only if @xmath82 for @xmath83 and @xmath84 ; + @xmath85 is defined in the way that each house in the market is unacceptable to agent @xmath86 . * @xmath87 for @xmath88 ; + @xmath89 + for all @xmath90 ; @xmath91 for interval @xmath71 , agent @xmath9 is an allocated subinterval of @xmath71 , denoted by @xmath92 , which is of length @xmath93 . for example , if @xmath94 $ ] , then one possibility of @xmath95 can be @xmath96 $ ] . + @xmath97 for all @xmath15 @xmath98 [ algo : f ] in example [ example : transform ] , we show how ccea transforms a cake cutting problem with piecewise constant valuations into a random assignment problem . [ example : transform ] we examine how ccea runs on the cake cutting problem in figure [ figure : piecewise ] . firstly , let @xmath99 be the set of subintervals of @xmath0 $ ] formed by consecutive points of discontinuity are identified : @xmath100 , j_2=[0.1,0.3 ] , j_3=[0.3,0.5]$ ] , and @xmath101 $ ] . @xmath102 is discarded because it is desired by no agent . in set @xmath103 , each house @xmath74 corresponds to subinterval @xmath71 . the preferences of the agents over @xmath52 are inferred from their valuation function height in the subintervals so that @xmath104 and @xmath105 we also set the number of units of each house that is available . since @xmath106 is the biggest interval , we consider @xmath107 as complete house . so , @xmath108 , @xmath109 , and @xmath110 . if we run cc over the housing market instance with the specified set of agents , houses , fraction of houses available to the market , and agent preferences , then the assignment returned by cc is as follows : @xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , and @xmath116 . the house assignment @xmath54 can be used to divide the subintervals among the agents : @xmath117 , [ 0.7,1]\}$ ] and @xmath118 , [ 0.5,0.7]\}$ ] . ccea satisfies the strong fairness property of robust envy - freeness . [ prop : ccea is robust envy - free ] for piecewise constant valuations , ccea satisfies robust envy - freeness and non - wastefulness . let @xmath119 be the number of relevant subintervals in a cake cutting problem with piecewise constant valuations . [ prop : ccea time ] ccea runs in time @xmath120 , where @xmath1 is the number of agents and @xmath119 is the number of subintervals defined by the union of discontinuity points of the agents valuation functions . although ccea satisfies the demanding property of robust envy - freeness , it is not immune to manipulation . we show that ccea is not strategyproof even for two agents . in the next section , we will present a different algorithm that is both robust envy - free and strategyproof for two agents . [ prop : ccea is not sp ] for piecewise constant valuations , ccea is not strategyproof even for two agents . if we restricted the preferences to piecewise uniform with no private endowment or variable claims , then ccea is not only strategyproof but group - strategyproof . we first show that in this restricted setting , ccea is in fact equivalent to the algorithm of @xcite . [ prop : equivalence ] for piecewise uniform value functions with no private endowments and variable claims , ccea is equivalent to mechanism 1 of @xcite . since the set of valuations that can be reported is bigger in cake cutting than in the assignment problem , establishing group strategyproofness does not follow automatically from group - strategyproofness of cc for dichotomous preferences ( theorem 2 , * ? ? ? using similar arguments , we give a detailed proof that ccea and hence mechanism 1 of @xcite is group strategyproof for piecewise uniform valuations . in section [ sec : extensions ] , we extend the result to the case where agents may have variable claims . [ prop : gsp1 ] for cake cutting with piecewise uniform value functions , ccea is group strategyproof . for piecewise uniform valuations , ccea is also pareto optimal . the result follows directly from lemma [ prop : equivalence ] along with the fact that mechanism 1 of @xcite is pareto optimal . [ prop : pareto optimality ] for cake cutting with piecewise uniform value functions , ccea is pareto optimal . in the previous section we presented ccea which is not pareto optimal for piecewise constant valuations . it turns out that if we relax the robust notion of fairness to envy - freeness , then we can use fundamental results in general equilibrium theory and recent algorithmic advances @xcite to formulate a convex program that always returns an envy - free and pareto optimal allocation as its optimal solution . for each valuation profile , let @xmath121 be the intervals whose endpoints are consecutive points in the union of break points of the agents valuation functions . let @xmath122 be the length of any subinterval of @xmath123 that is allocated to agent @xmath124 . then we run a convex program to compute a pareto optimal and envy - free allocation . we will call the convex program outlined in algorithm [ algo : market ] as the _ market equilibrium algorithm ( mea)_. mea is based on computing the market equilibrium via a primal - dual algorithm for a convex program that was shown to be polynomial - time solvable by @xcite . notice that if we ignore strategyproofness , or in other words , if we assume that agents report truthfully , then agents are truly indifferent between which subinterval they receive since their valuation function is a constant over any @xmath123 . hence , one we determine the length of @xmath71 to be allocated to an agent , we can allocate any subinterval of that length to the agent . cake - cutting problem with piecewise constant valuations . a proportional , envy - free , and pareto optimal allocation . let @xmath121 be the intervals whose endpoints are consecutive points in the union of break points of the agents valuation functions . let @xmath122 be the length of any subinterval of @xmath123 that is allocated to agent @xmath124 . @xmath125 solve the following convex program . @xmath126 let @xmath127 , @xmath128 be an optimal solution to the convex program . partition every interval @xmath123 into @xmath1 subintervals where the @xmath124-th subinterval @xmath129 has length @xmath128 . @xmath130 be the allocation of each @xmath131 . @xmath132 . [ algo : market ] [ pe , ef , prop ] mea is polynomial - time , pareto efficient and envy free . we mention here that the connection between a fair and efficient algorithm for cake cutting and computing market equilibria was already pointed by @xcite . @xcite presented an algorithm to compute an approximately envy - free and pareto optimal allocation for cake cutting with general valuations . however their algorithm is not polynomial - time even for piecewise constant valuations @xcite . mea requires the machinery of convex programming . it remains open whether mea can be implemented via linear programming . @xcite presented an algorithm that uses a linear program to compute an optimal envy - free allocation . the allocation is pareto optimal among all envy - free allocations . however it need not be pareto optimal in general . although mea is not robust envy - free like ccea , it is pareto optimal because it maximizes the nash product . what is interesting is that under uniform valuations , both mea and ccea are equivalent . in the next result we demonstrate this equivalence ( proposition [ ccea equivalence ] ) . the proof requires a careful comparison of both ccea and mea under uniform valuations . [ ccea equivalence ] for piecewise uniform valuations , the allocation given by ccea is identical to that given by mea . for piecewise uniform valuations , mea is group - strategyproof . thus if we want to generalize mechanism 1 of @xcite to piecewise constant valuations and maintain robust envy - freeness then we should opt for ccea . on the other hand , if want to still achieve pareto optimality , then mea is the appropriate generalization . in both generalization , we lose strategyproofness . thus far , we presented two polynomial time algorithms , each of which satisfies a different set of properties . ccea is robust envy - free and non - wasteful , whereas mea is pareto optimal and envy - free . this naturally leads to the following question : does there exist an algorithm satisfies all of the properties that ccea and mea satisfy ? it turns out that the answer is no , as theorem [ pe fair impossibility ] shows that there is no algorithm that is both pareto efficient and robust proportional . similarly , theorem [ pe sp prop impossibility ] argues that there is no algorithm that satisfies the properties ccea satisfies along with strategyproofness . lastly , theorem [ sp rob prop non w impossibility ] argues that there is no algorithm that satisfies the properties ccea satisfies plus strategyproofness . the impossibility results are summarized in table [ table : summary : impossible ] . consequently , we may conclude that the properties satisfied by ccea and mea are respectively maximal subsets of properties that an algorithm can satisfy for piecewise constant valuations . in the previous sections , we saw that ccea and mea are only strategyproof for piecewise uniform valuations . in light of the impossibility results established in the preivous section , it is reasonable to ask what other property along with strategyproofness can be satisfied by some algorithm . it follows from ( * ? ? ? * theorem 3 , ) that the only type of strategyproof and pareto optimal mechanisms are dictatorships . @xcite raised the question whether there exists a strategyproof and proportional algorithm for piecewise constant valuations . the algorithm csd answers this question partially . before diving into the csd algorithm , it is worth noting that there is some fundamental difference between random assignment setting and the cake cutting setting . in the random assignment setting , the objects that we are allocating are well defined and known to the public . on the other hand , in the cake cutting setting , the discontinuity points of each agent s valuation function is private information for the agent . hence , any algorithm that uses the reported discontinuity points to artificially create the objects runs into the risk of having the objects created by the algorithm be manipulated by the reports of the agents . in order to illustrate this difficulty , consider the uniform allocation rule . the uniform allocation rule ( that assigns @xmath2 of each house ) @xcite is both strategyproof and proportional in the random assignment setting . however it can not be adapted for cake cutting with piecewise constant valuations since strategyproofness is no longer satisfied if allocating @xmath2 of each interval ( induced by the agent valuations ) is done deterministically . [ prop : uniform allocation is not sp ] the uniform allocation rule ( done deterministically ) is not strategyproof . now we are ready to present csd . in order to motivate csd , we will give a randomized algorithm that is strategyproof and robust proportional in expectation . the algorithm is a variant of random dictatorship : each agent has uniform probability of being chosen as a dictator . however , if the whole cake is acceptable to each agent , then each time a dictator is chosen , he will take the whole cake . this approach is not helpful since we return to square one of having to divide the whole cake . we add an additional requirement which is helpful . we require that each time a dictator is chosen , the piece he takes has to be of maximum value @xmath2 length of the total size of the cake . we will call this algorithm constrained random serial dictatorship ( crsd ) . formally speaking , crsd draws a random permutation of the agents . the algorithm then makes the allocation to agents in the order that the lottery is drawn . everytime when it is agent @xmath9 s turn to receive his allocation , crsd looks at the remaining portion of the cake and allocates a maximum value @xmath2 length piece of the cake to agent @xmath9 ( break ties arbitrarily ) . notice that crsd is strategyproof , as in every draw of lottery , it is in the best interest of the agents to report their valuation function truthfully in order to obtain a piece that maximizes his valuation function out of the remaining pieces of cake . later on we will see , through the proof of proposition [ prop : csd is robust prop ] , that crsd is robust proportional in expectation . + csd is an algorithm that derandomizes crsd by looking at its allocation for all @xmath133 different permutations and aggregate them in a suitable manner . the algorithm is formally presented as algorithm [ csd - algo ] . + cake - cutting problem with piecewise constant valuations . a robust proportional allocation . @xmath134 $ ] ( intervals left ) @xmath135 maximum preference cake piece of size @xmath2 from @xmath136 @xmath137 . @xmath138 . construct a disjoint and exhaustive interval set @xmath139 induced by the discontinuities in agent valuations and the cake cuts in the @xmath133 cake allocations . @xmath140 empty allocation for each @xmath15 . let @xmath141 where @xmath142 is the number of permutations in which @xmath9 gets @xmath71 . generate @xmath143 , which is of length @xmath144 according to some subroutine . @xmath145 @xmath132 [ algo : csd ] although csd does not necessarily require @xmath133 cuts of the cake , it may take exponential time if the number of agents is not a constant . in example [ example : csd ] , we illustrate how csd works . [ example : csd ] we implement csd on the cake cutting problem in figure [ figure : piecewise ] . for permutation @xmath146 , agent @xmath147 first chooses the cake piece @xmath148,[0.6,1]\}$ ] and agent @xmath149 then takes the remaining piece @xmath150\}$ ] . for permutation @xmath151 , agent @xmath149 first chooses the cake piece @xmath152\}$ ] and agent @xmath147 then takes the remaining piece @xmath153,[0.8,1]\}$ ] . the set of all relevant subintervals induced by the two permutations are @xmath148 , [ 0.1,0.3],[0.3,0.6],[0.6,0.8],[0.8,1]\}$ ] . when we we additionally consider the discontinuities in the players valuations , the set of relevant subintervals is @xmath154 , [ 0.1,0.3],[0.3,0.5],[0.5,0.6 ] , [ 0.6,0.8],[0.8,1].$ ] then @xmath155,\frac{1}{2}[0.1,0.3],\frac{1}{2}[0.6,0.8 ] , [ 0.8,1]\}\ ] ] and @xmath156,[0.3,0.5],[0.5,0.6 ] , \frac{1}{2}[0.6,0.8]\}.\ ] ] [ prop : csd - well - defined ] for piecewise constant valuations , csd is well - defined and returns a feasible cake allocation in which each agents gets a piece of size @xmath2 . [ prop : csd is robust prop ] for piecewise constant valuations , csd satisfies robust proportionality and symmetry . notice that unlike crsd , csd interprets the probability of allocating each interval to an agent as allocation a fractional portion of the interval to that agent . unless the actual way of allocating the fractions is specified , one can not discuss the notion of strategyproofness for csd because a deviating agent is unable to properly evaluate his allocation against his true valuation function in the reported profile . contrary to intuition , csd may or may not be strategproof depending on how the fractional parts of each interval are allocated . in fact , the following remark illustrates this issue . [ remark : csd - not - sp ] csd is not strategyproof if the fraction of each interval of @xmath139 is allocated deterministically ( please see the appendix for the proof ) . in light of difficulty , we will implement a method ( see algorithm [ csd : subroutine ] ) that randomly allocating the fractions of intervals to agents . with this implementation , csd is strategyproof in expectation . generate @xmath157 $ ] . for @xmath158 , let @xmath159 if @xmath160 and @xmath161 if @xmath162 . let @xmath163\ ] ] if @xmath164 and @xmath165 \cup [ \mod(u_j + \sum_{k = 1}^{i-1}p_{nj}(b_j - a_j ) ) , b_j]\ ] ] otherwise . [ csd : subroutine ] we will refer this randomized implemention of csd as _ constrained mixed serial dictatorship _ , or _ cmsd _ for short . [ csd : sp ] csd implemented with the aforementioned random allocation rule is strategyproof in expectation . although csd is strategyproof in expectation , it fails to satisfy truthfulness based on group - based deviations no matter how the fractional parts of each interval are allocated . [ csd : not gsp for pwc ] for cake cutting with piecewise constant valuations , csd is not weakly group - strategyproof even for two agents . moreover , for cake cutting with piecewise uniform valuations , csd is not weakly group - strategyproof for at least seven agents . the statement above follows from the fact that rsd is not weakly group - strategyproof for dichotomous preferences when there are at least seven agents @xcite . even though csd satisfies both proportionality and symmetry , it does not satisfy the stronger notion of envy - freeness . [ prop : csd is not envy - free ] csd is not necessarily envy - free for three agents even for piecewise uniform valuations . another drawback of csd is that it is not pareto optimal for piecewise constant valuations . the statement follows from the fact that rsd is not sd - efficient @xcite . however for the case of two agents , it is robust envy - free and polynomial - time . for two agents and piecewise constant valuations , csd is robust envy - free , and polynomial - time but not pareto optimal . for piecewise uniform valuations , csd can be modified to be made pareto optimal . the main change is that for each permutation csd , the resultant outcome needs to be made pareto optimal . this can be done by using the idea in @xcite . in this section , we show how some of our positive results concerning ccea extend to more general settings where agents may have variable claims or they may have initial endowments ( please see algorithm [ algo : ccea - extensions ] ) . piecewise constant value functions with priority claims @xmath166 or private endowments @xmath167 a robust envy - free and individually rational allocation . let @xmath168 where @xmath86 is the agent owning all public cake but with no interest in any of the cake . join the segments @xmath169 to assemble a cake . divide the regions according to agent value functions . let @xmath75 be the set of subintervals of @xmath0 $ ] formed by consecutive marks . consider @xmath170 where * @xmath77 where @xmath78 for all @xmath79 * @xmath80 is defined as follows : @xmath81 if and only if @xmath82 for @xmath83 and @xmath84 ; + @xmath85 is defined in the way that each house in the market is unacceptable to agent @xmath86 . * @xmath87 for @xmath88 ; + @xmath89 + for all @xmath90 ; * @xmath171 ; * @xmath172 if @xmath173 and zero otherwise . @xmath174 for interval @xmath71 , agent @xmath9 is allocated a subinterval of @xmath71 , denoted by @xmath92 , which is of length @xmath93 . + @xmath175 for all @xmath15 + @xmath98 [ algo : ccea - extensions ] we consider the concept of _ variable claims _ in cake cutting . it is generally assumed in the literature that each agent has equal claim to the cake . in this case we modify the definition of proportionality and envy - freeness accordingly . reasoning about claims has a long tradition in the fair division literature . however this strand of research assumes that agents have uniform and identical preferences over the whole of the divisible object and their only concern is the proportion of the object they get ( see * ? ? ? * ; * ? ? ? @xcite refer to variable claims as entitlements and touch upon entitlements in cake cutting at a few places in their book . they present a general idea of handling entitlements by cloning agents ( page 152 , * ? ? ? * ) but doing so can lead to an exponential blowup of time and space . @xcite also considered entitlements in pie - cutting but presented an impossibility result and also a positive result for two agents . on the other hand , one of our algorithms handles variable claims for arbitrary number of agents and does not require cloning . an allocation satisfies _ proportionality for variable claims _ if @xmath176)$ ] for all @xmath177 . an allocation satisfies _ robust proportionality for variable claims _ if for all @xmath177 and for all @xmath28 , @xmath178}v_i'(x)dx$ ] . an allocation satisfies _ envy - freeness for variable claims _ , if @xmath179 for all @xmath32 . an allocation satisfies _ robust envy - freeness for variable claims _ if for all @xmath180 and for all @xmath28 , @xmath181 . [ prop : ccea envy for claims ] for cake cutting with variable claims and piecewise constant valuations , ccea satisfies robust envy - freeness for variable claims . for uniform valuations , group - strategyproofness of ccea even holds when agents have variable claims . [ prop : sp - with - speed ] for cake cutting with piecewise uniform value functions and variable claims , ccea is group - strategyproof . on the other hand , it can be shown that the following natural modification of csd does not satisfy proportionality in the presence of variable claims : each agent @xmath9 gets @xmath182 length of the most preferred interval for each permutation . we note that mea can be easily adapted to cater for variable claims . each agent s budget is proportional to his claim . [ th : maintheorem3b ] for cake cutting with with piecewise constant valuations and variable claims , there exists an algorithm ( mea ) that is polynomial - time , pareto optimal , envy - free , and proportional . in the cake cutting literature , the cake is generally considered to be a public good which is divided among the agents . it could also be that each agent brings part of the cake and then the assembled cake needs to be reassigned among the agents fairly . we call the setting cake cutting with _ private endowments_. in this context , individual rationality is a minimal requirement and envy - freeness need to be redefined . classic cake cutting can be modelled by cake cutting with private endowments in the following manner : none of the agents have any endowments and one additional agents owns the whole cake but has no interest in the cake . an allocation is _ individually rational _ if @xmath33 . an allocation satisfies _ justified envy - freeness for private endowments _ if @xmath183 or @xmath184 for each pair of agent @xmath32 . either @xmath9 should not be envious of @xmath124 or if he is envious , then giving @xmath9 s allocation to @xmath124 violates individual rationality . it is not entirely clear how to obtain the canonical concept of justified envy - freeness which takes into account both private endowments and variable claims . an allocation satisfies _ robust justified envy - freeness for private endowments _ if for all @xmath32 either @xmath185 or @xmath186 . if @xmath9 does not get an unambiguously as preferred a piece as @xmath124 s , @xmath124 has a counter claim that @xmath124 s endowment is better than @xmath9 s piece for some ordinally equivalent valuation function of @xmath124 . [ prop : ccea endowments ] for cake cutting with private endowments and piecewise constant valuations , ccea satisfies robust justified envy - freeness for private endowments . generalizing csd to handle endowments without losing its properties seems to be a challenging task . we have not addressed it in this paper . a mathematical analysis of cake cutting started with the work of polish mathematicians steinhaus , knaster , and banach @xcite . as applications of fair division have been identified in various multiagent settings , a topic which was once considered a mathematical curiosity has developed into a full - fledged sub - field of mathematical social sciences ( see * ? ? ? * ) . in particular , in the last few decades , the literature of cake cutting has grown considerably ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the relation between the random assignment problem and cake cutting has been noticed before @xcite . however , in their discussion of related work , @xcite argue that techniques from the random assignment literature can not be directly applied even to piecewise uniform functions a subclasses of piecewise constant functions . stumbling blocks identified include the fact that in the random assignment problem , each agents gets one object . we observed that ps satisfies envy - freeness even when agents get multiple objects . another issue raised is that ` if two agents desire two subintervals , both agents would value the longer subinterval more than the shorter . ' this problem can be circumvented by using the algorithm to compute the greatest common denominator of the lengths of the intervals and then dividing the intervals in equally sizes subintervals . however the number of subintervals can be exponential in the input size . we use the idea of @xcite that some houses may only be partially available . @xcite stated that generalizing their strategyproof algorithm for piecewise uniform valuations to the case for piecewise constant valuations as an open problem . we presented two algorithms ccea and mea that generalize mechanism 1 of @xcite . although they both satisfy certain desirable properties , both natural generalizations are not strategyproof . a number of works in the cake - cutting literature reason about strategyproofness . however they refer to a very weak notion of strategyproofness which is equivalent to a maximin strategy . apart from the papers of @xcite , we are aware of no positive results regarding discrete , strategyproof , and fair algorithms even for the restricted domain of piecewise constant valuations . in this paper we present a proportional algorithm ( csd ) for piecewise constant valuations that although not formally strategyproof seems less like likely to manipulate in comparison to ccea . if we are allowed to use randomization , then we show that csd can be adapted to be strategyproof in expectation . notice that if we instead require our algorithm to be strategyproof always , and proportional in expectation , then crsd would satify these properties . moreover , we note that @xcite also presented an algorithm that is randomized , envy - free and proportional , and strategyproof in expectation . however , it does not satisfy unanimity and requires much more randomization in contrast to csd . it remains an open question whether there exists an algorithm that always returns a proportional allocation and is always immune to agent manipulation . a difficulty that arises in coming up with strategyproof and proportional algorithm lies in that there is no restriction on the distribution of the discontinuity points of the agents valuation functions . to illustrate this point , suppose the algorithm designer knows that the discontinuity points of the agents valuation functions come from a set @xmath187 , where @xmath188 . consequently , a mechanism that partitions @xmath0 $ ] into intervals of the form @xmath189 $ ] and allocates @xmath2 of each interval to each agent would be proportional , envy - free and strategyproof . even if the designer does not know such a @xmath46 , but instead we require the minimum distance between any two consecutive discontinuity points of the agent s valuation function to be at least some @xmath190 , then there exists a strategyproof and @xmath191-proportional algorithm for this setting . as for private endowments , we are only aware of @xcite where private endowment are considered in pie - cutting . however , in comparison to our positive results concerning envy - freeness , @xcite presents a negative result concerning the core . in this paper , we made a number of conceptual and technical contributions . it will be interesting to consider new results with respect to robust versions of fairness , endowments , and variable claims . cake cutting is an exciting sub - field of microeconomics with numerous applications to computer science . in order for theory to be more relevant to practice , we envision exciting work in much richer models of cake cutting . the authors acknowledge the helpful comments of jay sethuraman and simina brnzei . this material is based upon work supported by the australian government s department of broadband , communications and the digital economy , the australian research council , the asian office of aerospace research and development through grant aoard-12405 . 37 [ 1]#1 [ 1]`#1 ` urlstyle [ 1]doi : # 1 s. athanassoglou and j. sethuraman . house allocation with fractional endowments . _ international journal of game theory _ , 400 ( 3):0 481513 , 2011 . h. aziz , f. brandt , and p. harrenstein . pareto optimality in coalition formation . in g. persiano , editor , _ proceedings of the 4th international symposium on algorithmic game theory ( sagt ) _ , lecture notes in computer science ( lncs ) , pages 93104 . springer - verlag , 2011 . a. bogomolnaia and h. moulin . a new solution to the random assignment problem . _ journal of economic theory _ , 1000 ( 2):0 295328 , 2001 . a. bogomolnaia and h. moulin . random matching and assignment under dichotomous preferences . econ / papers/2001papers/03moulin.pdf , 2001 . a. bogomolnaia and h. moulin . random matching under dichotomous preferences . _ econometrica _ , 720 ( 1):0 257279 , 2004 . s. j. brams . _ mathematics and democracy : designing better voting and fair - division procedures_. princeton university press , 2008 . s. j. brams and a. d. taylor . an envy - free cake division protocol . _ the american mathematical monthly _ , 1020 ( 1):0 918 , 1995 . s. j. brams and a. d. taylor . _ fair division : from cake - cutting to dispute resolution_. cambridge university press , 1996 . s. j. brams , m. a. jones , and c. klamler . proportional pie - cutting . _ international journal of game theory _ , 36:0 353367 , 2008 . c. chambers . consistency in the probabilistic assignment model . _ journal of mathematical economics _ , 40:0 953962 , 2004 . y. chen , j. k. lai , d. c. parkes , and a. d. procaccia . truth , justice , and cake cutting . in _ _ proc . of 24th aaai conference _ _ , pages 756761 , 2010 . y. chen , j. k. lai , d. c. parkes , and a. d. procaccia . truth , justice , and cake cutting . _ games and economic behavior _ , 770 ( 1):0 284297 , 2013 . y. j. cohler , j. k. lai , d. c. parkes , and a. d. procaccia . optimal envy - free cake cutting . in _ proceedings of the 25th aaai conference on artificial intelligence ( aaai ) _ , pages 626631 , 2011 . h. crs and h. moulin . scheduling with opting out : improving upon random priority . _ operations research _ , 490 ( 4):0 565577 , 2001 . n. devanur , c. h. papadimitriou , a. saberi , and v. vazirani . market equilibrium via a primal dual algorithm for a convex program . _ journal of the acm _ , 550 ( 5 ) , 2008 . g. gallo , m. grigoriadis , and r. tarjan . a fast parametric maximum flow algorithm and applications , . _ siam journal on computing _ , 18:0 3055 , 1989 . a. gibbard . manipulation of voting schemes . _ econometrica _ , 41:0 587602 , 1973 . katta and j. sethuraman . a solution to the random assignment problem on the full preference domain . _ journal of economic theory _ , 1310 ( 1):0 231250 , 2006 . f. kojima . random assignment of multiple indivisible objects . _ mathematical social sciences _ , 570 ( 1):0 134142 , 2009 . a. maya and n. nisan . incentive compatible two player cake cutting . in _ proceedings of the 8th international workshop on internet and network economics ( wine ) _ , pages 170183 , 2012 . e. mossel and o. tamuz . truthful fair division . in s. kontogiannis , e. koutsoupias , and p. spirakis , editors , _ proceedings of the 3rd international symposium on algorithmic game theory ( sagt ) _ , volume 6386 of _ lecture notes in computer science ( lncs ) _ , pages 288199 . springer - verlag , 2010 . h. moulin . _ fair division and collective welfare_. the mit press , 2003 . a. d. procaccia . cake cutting : not just child s play . _ communications of the acm _ , 560 ( 7):0 7887 , 2013 . j. h. reijnierse and j. a. m. potters . on finding an envy - free pareto - optimal division . _ mathematical programming _ , 83:0 291311 , 1998 . j. m. robertson and w. a. webb . _ cake cutting algorithms : be fair if you can_. a. k. peters , 1998 . j. schummer . strategy - proofness versus efficiency on restricted domains of exchange economies . _ social choice and welfare _ , 14:0 4756 , 1997 . h. steinhaus . the problem of fair division . _ econometrica _ , 16:0 101104 , 1948 . w. stromquist . how to cut a cake fairly . _ american mathematical monthly _ , 870 ( 8):0 640644 , 1980 . f. e. su . rental harmony : sperner s lemma in fair division . _ american mathematical monthly _ , 10:0 930942 . , 1999 w. thomson . axiomatic and game - theoretic analysis of bankruptcy and taxation problems : a survey . _ mathematical social sciences _ , 45:0 249297 , 2003 . w. thomson . children crying at birthday parties . _ economic theory _ , 31:0 501521 , 2007 . v. v. vazirani . combinatorial algorithms for market equilibria . in n. nisan , t. roughgarden , . tardos , and v. vazirani , editors , _ algorithmic game theory _ , chapter 10 , pages 103134 . cambridge university press , 2007 . d. weller . fair division of a measurable space . _ journal of mathematical economics _ , 140 ( 1):0 517 , 1985 . o. yilmaz . random assignment under weak preferences . _ games and economic behavior _ , 660 ( 1):0 546558 , 2009 . h. p. young . _ equity : in theory and practice_. princeton university press , 1994 . l. zhou . on a conjecture by gale about one - sided matching problems . _ journal of economic theory _ , 520 ( 1):0 123135 , 1990 . r. zivan , m. dudk , s. okamoto , and k. sycara . reducing untruthful manipulation in envy - free pareto optimal resource allocation . in _ ieee / wic / acm international conference on web intelligence and intelligent agent technology _ , pages 391398 , 2010 . consider the following two - agent profile . agent 1 : @xmath192 , \ v_1(x ) = v^{1}_b \ \text{for } \ x \in ( 0.5,1].\ ] ] agent 2 : @xmath193 , \ v_2(x ) = v^{2}_b \ \text{for } \ x \in ( 0.5,1].\ ] ] choose @xmath194 in such a way that @xmath195 and @xmath196 and @xmath197 . by either robust envy - freeness or robust proportionality , the mechanism must make the following allocation @xmath198 where @xmath199 . on the other hand , in order for the mechanism to be pareto efficient , @xmath198 must be an element of @xmath200 hence , we have reached an impossibility . for cake cutting with piecewise constant valuations and @xmath201 , it follows from ( * ? ? ? * theorem 3 , ) that the only type of strategyproof and pareto optimal mechanisms are dictatorships . consequently , there exists no strategyproof and pareto optimal mechanism that is also proportional or symmetric . the following example shows that there exists no cake - cutting algorithm that is strategyproof , robust proportional , and non - wasteful . + profile 1 : @xmath202 , \ v_1(x ) = b \ \text{if } \ x \in [ 0.5,1]\ ] ] @xmath203 , \ v_2(x ) = b \ \text{if } \ x \in [ 0.5,1]\ ] ] for some @xmath204 . + since the algorithm is robust proportional , it must be the case that each agent receive @xmath205 of @xmath206 $ ] and @xmath205 of @xmath207 $ ] . this is because the fractional parts of @xmath206 $ ] and @xmath208 $ ] that each agent receives must stochastically dominate the uniform allocation , otherwise the allocation would not satisfy proportionality for some valuation function in the ordinal equivalence class . without lost of generality ( up to reordering and regrouping of the cake ) , we may assume that agent @xmath147 receives @xmath209 \cup [ 0.5 , 0.75]$ ] and agent @xmath149 receives @xmath210 \cup [ 0.75 , 1]$ ] . + now consider profile 2 : @xmath211 , \ v_1(x ) = b \ \text{if } \ x \in [ 0.5,0.75 ] , \ v_1(x ) = 0 \ \text{otherwise}\ ] ] @xmath212 , \ v_2(x ) = b \ \text{if } \ x \in [ 0.5,1]\ ] ] by sp , agent @xmath147 must again receive @xmath209 \cup [ 0.5 , 0.75]$ ] and agent 2 receives @xmath210 \cup [ 0.75 , 1]$ ] . if agent @xmath147 receives anything less in profile @xmath149 , then he would deviate from profile @xmath149 to profile @xmath147 . if agent @xmath147 receives anything more in profile @xmath149 , then he would deviate from profile @xmath147 to profile @xmath149 . + now consider profile 3 : @xmath213 , v_1(x ) = b \ \text{if } \ x \in [ 0.5,0.75 ] , \ v_1(x ) = 0 \ \text{otherwise}\ ] ] @xmath214 , v_2(x ) = a \ \text{if } \ x \in [ 0.25,0.5 ] , v_2 ( x ) = b \ \text{if } \ x \in [ 0.5,1]\ ] ] by robust proportionality , both agent @xmath147 and @xmath149 must receive @xmath205 of @xmath215 $ ] . by non - wastefulness , agent @xmath149 must receive @xmath210 $ ] and @xmath216 $ ] since agent @xmath147 has a utility of @xmath217 on these intervals . hence , agent @xmath149 in profile 2 would misreport so that he receives the allocation in profile @xmath218 . first of all , ccea is non - wasteful because an agent is never allowed to eat a piece of the cake that he has no desire for . on the other hand , the algorithm terminates only when every portion of the cake that is desired by at least one agent is completely consumed by some agent who desires it . next , we show that the algorithm is robust envy - free . consider a fractional assignment @xmath54 returned by the cc algorithm . without private endowments cc is equivalent to the eps algorithm of @xcite . assignment @xmath54 satisfies justified envy - freeness in presence of variable eating rates : @xmath219 for all utilities @xmath220 consistent with preferences of @xmath9 over the houses . the intuition is that at any point during the running of cc , an agent @xmath9 will be ` eating ' his most favoured object(s ) at the same rate as any other agent @xmath124 even if @xmath124 is also the eating the same object(s ) . hence , for all @xmath28 , it is the case that for @xmath221 , @xmath40 . + for a cake cutting instance @xmath222 , @xmath223 is the input size where @xmath1 is the number of agents and @xmath119 is the number of relevant subintervals . once the lengths of the subintervals in @xmath224 are computed , the size of each house can be computed in linear time . the number of houses in ccea is @xmath119 . we now analyse the running time of cc on @xmath1 agents and @xmath119 houses ( section 3.5 , * ? ? ? * ) . in the cc algorithm , the flow network consists of @xmath225 vertices and @xmath226 arcs where @xmath227 and @xmath228 . the number of parametric flow problems needed to be solved is @xmath229 . a parametric flow network problem can be solved in time @xmath230 due to @xcite . hence , the running time of ccea is @xmath231 . ccea is not strategyproof even if all the piecewise intervals are of equal length , and there are no private endowments or variable claims , and agents have strict preferences over the intervals . in this case ccea is equivalent to the classic ps algorithm . it is known that ps is not strategyproof even for strict preferences when there are more objects than agents @xcite . in the absence of private endowments and variable claims , ccea can be solved by invoking eps instead of cc but with the slight modification that in the corresponding flow network of eps , the capacity of each arc @xmath232 is set to @xmath233 in step 2 of eps ( algorithm 1 , * ? ? ? let us refer to this simplified ccea as simpleccea . when simpleccea is run , it invokes eps and solves repeated parametric network flow problems ( step 3 , algorithm 1 , * ? ? ? * ) . in the step , eps computes a bottleneck set of agents and houses at each break - point . simpleccea computes bottleneck sets in the same way as mechanism 1 of @xcite and then allocates the resources in the bottleneck sets to the agents in the bottleneck set . the flow networks of the slightly modified eps ( figure 2 , * ? ? ? * ) and that of mechanism 1 ( figure 2 , * ? ? ? * ) are identical with only two insignificant differences namely that in the flow network of mechanism 1 of @xcite i ) the source and target are swapped and all the arcs are inverted and ii ) the size of the houses / intervals is not normalized . however , the eventual allocations are same . in light of lemma [ prop : equivalence ] , it suffices to show that simpleccea is gsp . we begin with some notations . + let @xmath234 denote the length of @xmath235 $ ] . since the utility function is piecewise uniform , it suffices to consider the length of pieces of the cake that are desired by each agent . + let @xmath236 be a coalition of agents who misreport their true preference . + let @xmath222 denote the instance where every agent reports truthfully and @xmath237 denote the instance where agents in @xmath46 misreport . + let @xmath238 $ ] denote the pieces of cake that are truly desired by each agent . + let @xmath239 $ ] denote the desired pieces of cake that are reported by each agent . + let @xmath240 $ ] denote the allocation received by each agent under truthful reports . + let @xmath241 $ ] denote the allocation received by each agent when the agents in @xmath46 misreport . + let @xmath242 be the bottleneck sets of agents with respect to the true preferences of agents arranged in the order that they are being allocated by the algorithm in instance @xmath222 . + let @xmath243 be the bottleneck sets of agents with respect to the true preferences of agents arranged in the order that they are being allocated by the algorithm in instance @xmath237 . + let @xmath244 denote the length of the allocation each agent receives in the bottleneck set @xmath20 . let @xmath245 @xmath246 in other words @xmath247 is the subset of agents of @xmath248 who weakly gain in utility when the agents in @xmath46 misreport , and @xmath249 is the subset of agents of @xmath248 who weakly lose in utility when the agents in @xmath46 misreport . we will show that for all @xmath250 , @xmath251 . this would then directly imply that @xmath46 must be empty since no coalition @xmath46 can exist such that by misreporting , everyone in the coalition is weakly better off and at least one agent in the coalition is strictly better off . we will prove this result via induction on @xmath252 . in order to carry on with the induction , we will show that no agent in @xmath253 appears in the coalition @xmath46 . we begin with a lemma . it is the case that @xmath254 . in other words , no agent in @xmath253 is strictly worse off when some subset of agents misreport their preference . it is clear that @xmath255 . suppose the claim does not hold , then there exists some agent @xmath256 . in other words , agent @xmath9 is strictly worse off due to the misreports . since an agent would only misreport his preference if misreporting weakly improves his utility , we deduce that @xmath257 , which implies that @xmath258 . consequently , the following set of inequalities hold for agent @xmath9 : @xmath259 where the first equality follows from the fact that @xmath260 by the way the algorithm allocates the pieces of cake to agents . the second equality follows from @xmath261 . and the last inequality follows from @xmath256 . + we claim that since @xmath253 is the first bottleneck set with respect to the true preference , it can not be the case that @xmath262 , where @xmath263 is the bottleneck set that @xmath9 belongs to in the instance @xmath237 . suppose this is the case , then we have that @xmath264 . it is clear that @xmath265 can not contain an agent who misreports in @xmath237 , since a misreporting agent in @xmath265 only receives a piece of cake with length @xmath266 , which is strictly less than what he wouldve gotten had he reported truthfully . hence , every agent in @xmath265 must report his true preference in @xmath237 . since @xmath265 is the first bottleneck set in @xmath237 , we have that @xmath267)}{\|x'_1\| } & = \frac{len((\cup_{j \in x'_1 } d_j)\cap [ 0,1])}{\|x'_1\| } \\ \geq \frac{len((\cup_{j \in x_1 } d_j)\cap [ 0,1])}{\|x_1\| } & = len(x_1 ) , \end{split}\ ] ] where the first and third equalities follows from the way the algorithm makes an allocation , the second equality follows from the fact that the agents in @xmath265 have the same reports in @xmath237 as in @xmath222 , and the inequality follows from the fact that @xmath268))}{\|s\|}$ ] . this contradicts the fact that @xmath266 . it is the case that @xmath269 . in other words , no agent in @xmath253 is strictly better off when some subset of agents misreport their preference . since we have established that @xmath254 , and @xmath270 , it suffices to show that @xmath271 . suppose not , then for all @xmath272 , we have that @xmath273 and there exists some @xmath274 such that @xmath275 . summing over @xmath272 , we get that @xmath276 where the first two equalities follow from the fact that the @xmath277 s and @xmath278 s are disjoint subsets and the third equality follows from the way the algorithm allocates to the agents in the smallest bottleneck set . but this set of inequalities contradict the fact that @xmath279 , which implies that @xmath280 . hence , it must be the case that for every @xmath272 , we have that @xmath281 , which implies that @xmath282 . no agent in @xmath253 appears in the coalition @xmath46 and @xmath253 is also the first bottleneck set for @xmath237 . by the previous lemma , no agent in @xmath283 is strictly better off by misreporting his preference . thus , any agent in @xmath253 would potentially be in @xmath46 if by misreporting , he makes himself no worse off and simultaneously make some other agent in @xmath46 strictly better off . the only way that this can happen is by misreporting , the agents in @xmath253 make their collective claim over their desired pieces smaller , so that agents in later bottleneck set can claim some of their desired pieces . on the other hand , the following inequality @xmath284 implies that if any subset of agents of @xmath253 wants to misreport so they are not worse off , then collectively , they must over - report their preference to obtain allocations that together is weakly larger in total than the allocations they would get had they reported truthfully . thus , having a subset of agents in @xmath253 misreport will not benefit the other agents in the coalition @xmath46 . hence , we may conclude that no agent in @xmath253 appears in the coalition @xmath46 . provided that every agent in @xmath253 also reports truthfully in @xmath237 , there is no incentive for an agent that belongs to a subsequent bottleneck set in @xmath222 to misreport and prevent @xmath253 from being the first bottleneck set in @xmath237 since that would make the misreporting agent strictly worse off , as in doing so , he needs to create a bottleneck set @xmath285 such that @xmath286 and he would consequently receive an allocation of @xmath287 . + since no agent in @xmath253 appears in the coalition @xmath46 and @xmath253 is also the first bottleneck set for @xmath237 , we can remove @xmath253 from @xmath53 and @xmath288 from @xmath0 $ ] and do induction on the set of remaining agents @xmath289 and the remaining piece of cake @xmath0 \backslash \cup_{i \in x_1 } a_i$ ] to be allocated and the proof is complete by invoking the inductive hypothesis with @xmath290 being the first bottleneck set in the new instance . consider the following math program @xmath291 where @xmath292 , + notice that the feasible region of the math program contains all feasible allocations . an optimal solution given by the lp is not pareto dominated by any other feasible allocation because that would contradict the optimality of the solution . hence , it is pareto efficient . + to see that the optimal solution of the math program is also an envy free allocation , if we instead view @xmath122 as the fractional amount of @xmath123 that is allocated to agent @xmath124 , then scaling the @xmath293 s appropriately ( i.e. setting @xmath294 ) , then solving the math program stated in the proposition is equivalent to solving the following math program . @xmath295 which in turn equivalent to solving @xmath296 notice that the above math program is a convex program since we are maximizing a concave function ( or equivalently minimizing a convex function ) subject to linear constraints . + in ( pp 105 - 107 , * ? ? ? * ) , vazirani invites us to consider a market setting of buyers ( agents ) and divisible goods ( intervals ) . each good is assumed to be desired by at least one buyer ( i.e. for every good @xmath9 , @xmath297 for some buyer @xmath124 ) . there is a unit of each good and each buyer is given the same amount of money say @xmath147 dollar , for which he uses to purchases the good(s ) that maximizes his utility subject to a set of given prices . the task is to find a set of equilibrium prices such that the market clears ( meaning all the demands are met and no part of any item is leftover ) when the buyers seek purchase good(s ) to maximize their utility given the equilibrium prices . + using duality theory , one can interpret the dual variable @xmath55 associated with the constraints @xmath298 as the price of consuming a unit of good @xmath9 . by invoking the kkt conditions , @xcite shows the prices given by the optimal dual solution is a unique set of equilibrium prices . moreover , the primal optimal solution for each buyer @xmath124 is precisely the quantity of good(s ) that the buyer ends up purchasing that maximizes his utility given the equilibrium prices . + now we can argue as to why the optimal primal solution is an envy free allocation . because given the equilibrium prices , if a buyer desires another buyer s allocation , since he has the same purchasing power as any other buyer , he would instead use his money to obtain the allocation of the buyer that he envies . this would result in some surplus and deficit of goods , contradicting the fact that the given prices are equilibrium prices . for piecewise uniform valuations , it is known that ccea is equivalent to mechanism @xmath147 in @xcite , which we will refer to as _ simpleccea_. the remainder of the proof will focus on showing the equivalence between simpleccea and the convex program for piecewise uniform valuations . to do so , given an allocation of simpleccea , which is a feasible solution of the convex program , we will find a set of prices corresponding to the allocation and show that the prices are in fact the equilibrium prices defined by vazirani on pages 105 - 107 of @xcite . moreover , this allocation would be an allocation that maximizes the agents utility given the equilibrium prices . + using the same notations as those in @xcite , given a valuation profile , let @xmath299 be the set of buyers / agents and @xmath300 be the set of goods or intervals . let @xmath301 be the @xmath9-th bottleneck set computed by simpleccea , i.e. @xmath302 in the @xmath9-th iteration of the subroutine of simpleccea . let @xmath303 be the set of goods that are distributed amongst the buyers in @xmath301 . in the convex program , since each buyer is endowed with @xmath147 dollar and every buyer in @xmath301 receives @xmath304 units of good(s ) , it is natural to define the price of a unit of each good @xmath305 to be @xmath306 notice that the prices for each good is well defined ( i.e. each good has exactly one nonnegative price ) . this follows from the following observations : 1 . @xmath307 ( or every good has at least one price ) . this follows from the assumption that every good is desired by at least one agent , which means that simpleccea will allocate all of the goods . 2 . @xmath308 for all @xmath309 ( or every good has at most one price ) . this follows from the fact that no fractional parts of any good is allocated to agents from two or more bottleneck sets , which another algorithmic property of simpleccea . to show that the @xmath310 s form a set of equilibrium prices , we will show that given the @xmath310 s , the buyers in every @xmath301 will choose to purchase _ only _ goods from @xmath303 to maximize their utility function . we will do induction on the number of bottleneck sets . consider the first bottleneck set @xmath311 , we will show that @xmath312 are the only items that are desirable by buyers in @xmath311 . simpleccea finds @xmath311 by solving a parametrized max flow problem on a bipartite network . the network has a node for every buyer and a node for every good . there is a directed edge from a buyer @xmath9 to a good @xmath124 with infinite capacity if good @xmath124 is desired by buyer @xmath9 . in addition , there is a source @xmath313 and a sink @xmath314 . there is a directed edge from @xmath313 to each buyer @xmath9 with capacity @xmath315 and a directed edge from edge good @xmath124 to @xmath314 with capacity equaling the quantity of good @xmath124 . @xmath315 is set to @xmath217 initially so that the unique min cut in the network for @xmath316 is @xmath317 . @xmath315 is gradually raised until @xmath317 is no longer the unique min cut , at which point @xmath311 and @xmath312 are found by looking for another min cut of the form @xmath318 . since @xmath318 is a min cut , it must be the case that @xmath319 in the network , which proves the lemma . ( if @xmath320 , then there will be an infinite capacity edge crossing the cut . on the other hand , if @xmath321 , then replacing @xmath312 with @xmath322 on the @xmath313 side of the cut would give a cut with a smaller capacity . ) since the agents have piecewise uniform valuation , in the setting of the convex program , each buyer in @xmath311 has the same utility for the items that he desire in @xmath312 . moreover , given that the prices of goods are identical in @xmath312 , each buyer is indifferent between choosing among his desirable items for the best item in terms of bang per buck . hence , for all buyers in @xmath311 , the allocation given by simpleccea maximizes buyer s utility given the prices @xmath310 s . moreover , notice that the money of buyers in @xmath311 and goods in @xmath312 are exhausted by the allocation given by simpleccea . + after the goods in @xmath312 are allocated to the buyers in @xmath311 we repeat the same argument for the remaining buyers and goods and the inductive hypothesis allows to conclude that for every @xmath9 and for all buyers in @xmath301 , the allocation given by simpleccea maximizes buyer s utility given the prices @xmath310 s . moreover , the money of buyers in @xmath301 and goods in @xmath303 are exhausted by the allocation given by simpleccea . there is a slight difference between the inductive step and the base case , as it is possible that some buyer @xmath323 in @xmath324 also desires certain goods in @xmath325 for some @xmath326 . however , @xcite state that @xmath304 is a weakly increasing function of @xmath9 , which means that @xmath327 is a weakly decreasing function of @xmath9 . since we are dealing with piecewise uniform valuations , the utility of a buyer over his desirable goods are identical , this means that for any buyer in @xmath301 , the goods that maximize his bang for buck are in @xmath303 . + putting everything together and we have shown that @xmath310 s constitute the set of equilibrium prices and simpleccea gives an equilibrium allocation , which is an optimal solution to the convex program . consider the following profile of two agents . profile 1 : * @xmath328 , \ v_1(x ) = 0 \ \text{if } \ x \in ( 0.2 , 1].$ ] * @xmath329 , \ v_2(x ) = 1 \ \text{if } \ x \in ( 0.6 , 1].$ ] the uniform allocation rule gives us the allocation : * @xmath330 , \frac{1}{2}(0.6,1]\}$ ] . + * @xmath331 , \frac{1}{2}(0.6,1]\}$ ] . let @xmath332 $ ] be the allocation that agent @xmath149 receives in this case . + now consider profile 2 : * @xmath328 , \ v_1(x ) = 0 \ \text{if } \ x \in ( 0.2 , 1].$ ] * @xmath333 \backslash a , \ v_2(x ) = 1 \ \text{if } \ x \in a.$ ] the uniform allocation rule gives us the allocation : * @xmath330 , \frac{1}{2}a\}$ ] . + * @xmath331 , \frac{1}{2}a\}$ ] . hence , agent @xmath149 in profile @xmath149 would misreport so that the reported profile is profile @xmath147 . we first prove that csd is well - defined and results in a feasible allocation in which each agent gets @xmath2 size of the cake . let @xmath334 be a partitioning of the interval @xmath0 $ ] induced by the discontinuities in agent valuations and the cake cuts in the @xmath133 cake allocations . we make a couple of claims about @xmath139 that following from the way @xmath139 is constructed . an agent is completely indifferent over each subinterval in @xmath139 . let @xmath335 denote a maximum preference cake piece of size @xmath2 chosen by agent @xmath9 in the serial order @xmath336 . for each @xmath337 either @xmath335 contains @xmath338 completely or it does not contain any part of @xmath338 . now consider a matrix of dimension @xmath339 : @xmath340 such that @xmath341 if @xmath342 and @xmath343 if @xmath344 . since for each @xmath345 , each agent @xmath15 gets @xmath2 of the cake in @xmath346 , then it follows that @xmath347 . hence , @xmath348 also consider a matrix of dimension @xmath349 : @xmath350 such that @xmath351 denotes the fraction of @xmath71 that agent @xmath9 gets in @xmath352 . from the algorithm csd , we know that @xmath353 where @xmath142 is the number of permutations in which @xmath9 gets @xmath71 . it is immediately seen that each column sums up to @xmath147 . hence each @xmath71 is complete allocated to the agents . we now prove that each agent gets a total cake piece of size @xmath2 . we do so by showing that @xmath354 . @xmath355 hence @xmath132 the allocation returned by csd is a proper allocation of the cake in which each agent gets a total cake piece of size @xmath2 . we first argue for proportionality of csd . in the case where all agents have the same valuations as the valuation of @xmath9 , @xmath9 is guaranteed @xmath2 of the value of the whole cake because of anonymity of csd . first note that for each @xmath345 and preferences @xmath356 of all agents other than @xmath9 , @xmath357 the reason is that when valuations are not identical , predecessors of @xmath9 in @xmath336 leave weakly better cake for @xmath9 as when their valuations are same as agent @xmath9 . hence , @xmath358)/n . $ ] finally , note that when an agent selects his best possible cake piece in each permutation , the exact height of the valuation function is not relevant and only the relative height matters . hence , csd in fact satisfies robust proportionality . symmetry for csd follows directly from symmetry for rsd . csd would not be strategyproof if the fraction of each subinterval of @xmath139 is allocated deterministically and the allocation is made public information before the agents submit their valuation function . to see this , consider the following example of two agents . profile 1 : * @xmath359 , \ v_1(x ) = 0 \ \text{if } \ x \in ( 0.5 , 1].$ ] * @xmath360 , \ v_2(x ) = 0 \ \text{if } \ x \in ( 0.5 , 1].$ ] running csd gives us the allocation : * @xmath361 , \frac{1}{2}(0.5,1]\}$ ] . + * @xmath362 , \frac{1}{2}(0.5,1]\}$ ] . profile 2 : * @xmath359 , \ v_1(x ) = 0 \ \text{if } \ x \in ( 0.5 , 1].$ ] * @xmath363 , \ v_2(x ) = 1 \ \text{if } \ x \in ( 0.25 , 0.75 ] , v_2(x ) = 0 \ \text{if } \ x \in ( 0.75 , 1].$ ] running csd gives us the allocation : * @xmath364 , \frac{1}{2}(0.25,0.5 ] , \frac{1}{2}(0.75,1]\}$ ] . + * @xmath365 , ( 0.5,0.75 ] , \frac{1}{2}(0.75,1]\}$ ] . now it is possible that the csd mechanism decides to gives @xmath215 \cup ( 0.75,1]$ ] to agent 1 and @xmath366 $ ] to agent 2 in profile 1 . consequently , knowing this , agent 2 in profile 2 would misreport his valuation function to be @xmath360 , v_2(x ) = 0 \ \text{if } \ x \in ( 0.5 , 1]$ ] in order to receive the allocation given in profile 1 and gain utility in doing so . consider the profiles @xmath44 and @xmath367 , where @xmath44 is a profile where every agent reports truthfully and @xmath367 is a profile where agent @xmath9 misreports while fixing every other agent s report to be the same as that in @xmath44 . let @xmath368 denote a permutation of @xmath369 = \{1,\ldots , n\}$ ] and let @xmath46 denote the set of all permutations of @xmath369 $ ] . let @xmath370 denote the intervals whose fractional allocations are specified to each agent by csd in profile @xmath44 and @xmath371 denote the intervals whose fractional allocations are specified to each agent by csd in profile @xmath367 . let @xmath372 denote agent @xmath9 s total utility derived from receiving the interval @xmath338 and @xmath373 and @xmath374 denote the agent @xmath9 s total utility derived from his allocated pieces when the serial ordering of the agents is @xmath368 in profile @xmath44 and @xmath367 respectively . let @xmath375 denote the probability that interval @xmath71 is assigned to agent @xmath9 . since random serial dictatorship is strategyproof in expectation , we have that @xmath376 now csd views @xmath375 as allocating a @xmath375 fraction of interval @xmath71 to agent @xmath9 . in order for a deviating agent to properly evaluate the utility derived from his allocation in the deviating profile @xmath367 , we have to come up with an allocation rule that actually _ attains _ the utility @xmath377 for agent @xmath9 ( either deterministically or in expectation ) when the profile of reports is @xmath367 . in particular , say if we want to allocate a subinterval of @xmath123 with length @xmath375 times that of @xmath123 to agent @xmath9 at random , then this random allocation rule must satisfy the property that @xmath378 = p_{ij}v_i(j'_j)$ ] , where @xmath379 such that @xmath380 . in order to do so , we will show that the randomized allocation rule for csd satisfy the property that every agent @xmath9 and interval @xmath71 , we have that @xmath381 = p_{ij}v_i(j'_j)$ ] for all valuation functions @xmath382 , where @xmath379 such that @xmath380 . + notice that @xmath383 is uniformly distributed on @xmath384 $ ] and @xmath385 has length @xmath386 . the fact that @xmath381 = p_{ij}v_i(j'_j)$ ] follows from the following lemma . [ lemma : csd - random ] let @xmath387 be uniformly distributed on the interval @xmath388 $ ] and let @xmath389 . let @xmath390 $ ] if @xmath391 and @xmath392 \cup [ u , b]$ ] if @xmath393 , then we have that @xmath394 = \alpha v_i([a , b])$ ] , where @xmath395 for any integrable function @xmath24 . define @xmath396 for @xmath397 and @xmath398 for @xmath198 outside @xmath388 $ ] . since @xmath399 is periodic , then it suffices to show that @xmath400 = \int_{a}^{b } \int_{x}^{x+\alpha ( b - a ) } \bar{v_i}(y ) dy \frac{1}{b - a } dx = \alpha \int_{a}^{b } \bar{v_i}(x ) dx = \alpha v_i([a , b]).\ ] ] by drawing a picture of the region that we evaluate the integral over and due to the fact that @xmath399 is periodic , we have that @xmath401 which proves the lemma . let @xmath402 , b = ( 0.25 , 0.5 ] , c = ( 0.5 , 0.75 ] , d=(0.75,1]$ ] . consider the following two profiles of valuations . + profile 1 : * @xmath403 * @xmath404 running csd gives us the allocation : * @xmath405 . + * @xmath406 . profile 2 : * @xmath407 * @xmath408 running csd gives us the allocation : * @xmath409 . + * @xmath410 . hence , agents with true valuation in profile 1 would misreport together to profile 2 , which means that csd is not group strategyproof for 2 agents . there are three agents , each with piecewise uniform valuation function . @xmath411 for @xmath412 $ ] and @xmath217 otherwise . + @xmath413 for @xmath414 \cup ( 2/3,1]$ ] and @xmath217 otherwise . + @xmath415 for @xmath416 $ ] and @xmath217 otherwise . + let @xmath417 $ ] , @xmath418 $ ] , @xmath419 $ ] . we adopt the following implementation of csd : when it is agent i s turn to pick , out of the pieces of the remaining cake that he likes , he takes the _ left - most _ such piece with length 1/n , where n is the number of agents . if the priority ordering were @xmath420 , then a feasible assignment that respects the preferences is 1 @xmath421 a , 2 @xmath421 c , 3 @xmath421 b. if the priority ordering were 1,3,2 , then a feasible assignment that respects the preferences is @xmath422 @xmath423 , @xmath424 . if the priority ordering were @xmath425 , then a feasible assignment that respects the preferences is @xmath426 , @xmath427 , @xmath428 . if the priority ordering were @xmath429 , then a feasible assignment that respects the preferences is @xmath426 , @xmath423 , @xmath430 . if the priority ordering were @xmath431 , then a feasible assignment that respects the preferences is @xmath432 , @xmath433 , @xmath424 . if the priority ordering were @xmath434 , then a feasible assignment that respects the preferences is @xmath423 , @xmath426 , @xmath430 . then , the csd allocation is as follows . @xmath435 , \frac{1}{6}(1/3 , 2/3 ] , \frac{1}{3}(2/3 , 1]\}\\ y_2 = & \quad \{\frac{1}{2}[0,1/3 ] , \frac{1}{2}(2/3,1]\}\\ y_3 = & \quad \{\frac{5}{6}(1/3 , 2/3 ] , \frac{1}{6}(2/3,1]\ } \end{aligned}\ ] ] clearly , agent 1 envies agent 3 in this case . consider a fractional assignment @xmath54 returned by the cc algorithm . without private endowments cc is equivalent to the eps algorithm of @xcite . assignment @xmath54 satisfies justified envy - freeness in presence of variable eating rates : @xmath436 for all utilities @xmath220 consistent with preferences of @xmath9 over the houses . the informal intuition is that at any point during the running of cc , an agent @xmath9 with a higher eating rate than @xmath124 will be ` eating ' his most favoured object(s ) faster than @xmath124 even if @xmath124 is also has the eating the same object(s ) . hence , for all @xmath28 , it is the case that for @xmath221 , @xmath181 . given a variable claim instance @xmath222 of @xmath1 agents where agent @xmath9 has claim rate @xmath437 for @xmath438 . we may assume without lost of generality that the claim rates are integral . if they are not , then we can simply multiple each claim rate @xmath437 by a common denominator to make each @xmath437 integral . doing so will not change the allocation given by the algorithm since only relative claim rates matter to the algorithm . + now consider a cake cutting instance @xmath237 of @xmath439 agents , where the agents have piecewise uniform utility function and there are no private endowments or variable claims . moreover , for every @xmath440 , there are @xmath437 agents in @xmath237 each of whom has the same utility function as that of agent @xmath9 in @xmath222 . it is not difficult to see that if one aggregates the allocation that the @xmath437 agents in @xmath237 who share agent @xmath9 s valuation in @xmath222 , then one would get an equivalent allocation ( in terms of utility ) to agent @xmath9 s allocation . suppose for the sake of contradiction that ccea is not gsp for the case of variable claims , then in some instance @xmath222 , there exists some coalition @xmath46 of the agents that weakly gains in utility by misreporting their preference . now consider the equivalent instance @xmath237 with no variable claims under the aforementioned transformation , then there exists some coalition @xmath441 of the agents in @xmath237 that weakly gains in utility by misreporting their preference , which implies that ccea is not group - strategyproof for the no variable claims case , contradicting the result of proposition [ prop : gsp1 ] . the allocation can be obtained by solving the following convex program . @xmath442 the proof of the desired properties is similar to the case where @xmath443 . consider a fractional assignment @xmath54 returned by the cc algorithm.we know that @xmath54 satisfies justified envy - freeness for the random / fractional assignment problem ( prop . 4 , * ) . if @xmath69 , then @xmath185 . if @xmath444 , then @xmath186 . hence @xmath8 satisfies justified envy - freeness for private endowments .	cake cutting is one of the most fundamental settings in fair division and mechanism design without money . in this paper , we consider different levels of three fundamental goals in cake cutting : fairness , pareto optimality , and strategyproofness . in particular , we present robust versions of envy - freeness and proportionality that are not only stronger than their standard counter - parts but also have less information requirements . we then focus on cake cutting with piecewise constant valuations and present three desirable algorithms : _ ccea ( controlled cake eating algorithm ) _ , _ mea ( market equilibrium algorithm ) _ and _ csd ( constrained serial dictatorship)_. ccea is polynomial - time , robust envy - free , and non - wasteful . it relies on parametric network flows and recent generalizations of the probabilistic serial algorithm . for the subdomain of piecewise uniform valuations , we show that it is also group - strategyproof . then , we show that there exists an algorithm _ ( mea ) _ that is polynomial - time , envy - free , proportional , and pareto optimal . mea is based on computing a market - based equilibrium via a convex program and relies on the results of @xcite and @xcite . moreover , we show that mea and ccea are equivalent to mechanism 1 of chen et . al . for piecewise uniform valuations . we then present an algorithm _ csd _ and a way to implement it via randomization that satisfies strategyproofness in expectation , robust proportionality , and unanimity for piecewise constant valuations . for the case of two agents , it is robust envy - free , robust proportional , strategyproof , and polynomial - time . many of our results extend to more general settings in cake cutting that allow for variable claims and initial endowments . we also show a few impossibility results to complement our algorithms . the impossibilities show that the properties satisfied by ccea and mea are maximal subsets of properties that can be satisfied by any algorithm for piecewise constant valuation profiles .
one of the most important tools in transformation groups is p. a. smith theory @xcite , ( * ? ? ? iii ) , which gives constraints on the homology of the fixed point set for actions of finite @xmath0-groups . for topological actions the fixed point sets may not be manifolds , but generalized manifolds " with complicated local topology ( in the sense of wilder ( * ? ? ? * chap . this means that an appropriate homology theory must be used to capture the essential features . smith theory in the generalized manifold setting , as developed in the 1960 classic seminar on transformation groups " @xcite , was used recently by bridson and vogtmann @xcite to study the actions of @xmath1 , the automorphism group of a free group , on acyclic homology manifolds and generalized homology @xmath2-spheres . in this paper we provide a coarse homology " version of p. a. smith theory suitable for further applications in geometric group theory . we study discrete groups of isometries of metric spaces , from the perspective of large - scale " geometry introduced by m. gromov in @xcite . this subject is now known as coarse geometry . we introduce a coarse generalization of the usual fixed set , called the _ bounded fixed set _ ( see definition [ bdd ] ) . it is defined when the coarse type of a certain sequence of approximate fixed sets stabilizes , even when the actual fixed set is empty . a group action is called _ tame _ if the bounded fixed set exists with respect to any subgroup . we say that a metric space is a ( mod @xmath0 ) _ coarse homology @xmath2-sphere _ if it has the same ( mod @xmath0 ) coarse homology as the euclidean space @xmath3 . the main application is : let @xmath4 be a proper geodesic metric space with finite asymptotic dimension , which is a ( mod @xmath0 ) coarse homology @xmath2-sphere , for some prime @xmath0 . let @xmath5 be a finite @xmath0-group with a tame action on @xmath4 by isometries . then @xmath6 is a ( mod @xmath0 ) coarse homology @xmath7-sphere , for some @xmath8 . if @xmath0 is odd , then @xmath9 is even . the coarse geometry of group actions extends to _ quasi - actions _ on proper metric spaces ( see section [ sec : quasi ] ) . in particular , the bounded fixed set is a quasi - isometry invariant ( see proposition [ bdfix - same ] ) . the coarse analogues of the usual smith theory inequalities are established in theorem [ thm : pasmith ] , and used to derive theorem a in section [ proof of thma ] . another well - known application of the classical p. a. smith theory is that a rank two group @xmath10 , for @xmath0 a prime , can not act freely on a finitistic mod @xmath0 homology @xmath2-sphere ( see bredon ( * ? ? ? * iii.8.1 ) ) . in theorem [ thm : semifree ] , we give a coarse version of this result . the group @xmath10 , for @xmath0 a prime , can not act tamely and semifreely at the large scale on a ( mod @xmath0 ) coarse homology @xmath2-sphere @xmath4 , whenever @xmath4 is a proper geodesic metric space with finite asymptotic dimension , and @xmath6 is a ( mod @xmath0 ) coarse homology @xmath7-sphere , for some @xmath11 . we do not yet know complete necessary and sufficient conditions for tameness of actions on a given metric space . example [ theexample ] shows that the sequence of approximate fixed sets does not always stabilize . on the other hand , in section [ sec : hyperbolic ] we show that the action of any finite subgroup of isometries of hyperbolic @xmath12-space , or more generally any proper @xmath13 space , is tame . in theorem [ thm : existence ] , we show that a finite group action on a coarsely homogeneous metric space @xmath4 is tame ( e.g. if @xmath4 admits a compatible proper and cocompact discrete group of isometries ) . coarse geometry studies the properties of coarse spaces and coarse maps . we will consider only the metric examples of coarse spaces . for the general definition of a coarse space see roe @xcite or mitchener @xcite , @xcite . [ def : coarse map ] let @xmath14 and @xmath15 be metric spaces and @xmath16 a map , not necessarily continuous . * the map @xmath17 is ( metrically ) _ proper _ if the inverse image under @xmath17 of any bounded subset of @xmath18 is a bounded subset of @xmath4 . * the map @xmath17 is _ bornologous _ if for every @xmath19 there is @xmath20 such that @xmath21 for all @xmath22 . * the map @xmath17 is _ coarse _ if it is proper and bornologous . two maps @xmath17 , @xmath23 from a set @xmath4 to a metric space @xmath18 are said to be _ close _ if @xmath24 is bounded , uniformly in @xmath4 . [ coarse - equiv ] two metric spaces @xmath4 and @xmath18 are _ coarsely equivalent _ if there exist coarse maps @xmath16 and @xmath25 such that @xmath26 and @xmath27 are close to the identity maps on @xmath18 and on @xmath4 respectively . the maps @xmath17 and @xmath23 are called _ coarse equivalences_. we remark that if @xmath28 and @xmath29 are coarse equivalences , then the composite @xmath30 is also a coarse equivalence . [ eff_prop ] let @xmath4 , @xmath18 be metric spaces and @xmath16 . * @xmath17 is called _ eventually lipschitz _ ( or _ large - scale lipschitz _ ) if there are positive constants @xmath31 and @xmath32 such that @xmath33 for any @xmath34 . * @xmath17 is called _ effectively proper _ if for every @xmath19 , there is @xmath35 such that the inverse image under @xmath17 of each ball of radius @xmath36 in @xmath18 is contained in a ball of radius @xmath37 in @xmath4 . [ equiv - to - image ] let @xmath16 be a coarse equivalence and @xmath38 . then @xmath39 and @xmath40 ( with the induced metrics ) are coarsely equivalent . the restriction of @xmath17 to @xmath39 is a coarse map . for any @xmath41 we choose @xmath42 and define @xmath43 . we obtain a map @xmath44 such that @xmath45 and @xmath46 is close to @xmath47 . one can easily check that @xmath48 is a coarse map . let @xmath4 and @xmath18 be two metric spaces . a map @xmath49 is an @xmath50-_quasi - isometric embedding _ for the positive constants @xmath31 and @xmath51 , if for any @xmath52 we have : @xmath53 if @xmath54 is also _ coarsely dense _ in @xmath18 ( that is , if any point in @xmath18 lies in the @xmath51-neighbourhood @xmath55 of @xmath56 ) , then @xmath17 is called an @xmath50-_quasi - isometry _ , or just a _ quasi - isometry _ for short . if @xmath28 and @xmath29 are quasi - isometries , then the composite @xmath57 is also a quasi - isometry . any quasi - isometric embedding is a coarse map : the first part of the above inequality shows that @xmath17 is proper and the second part shows that @xmath17 is eventually lipschitz , thus bornologous . the next result shows that any quasi - isometry is a coarse equivalence . if @xmath49 is a quasi - isometry , then there exists a quasi - isometry @xmath58 such that @xmath26 is close to the identity map on @xmath18 and @xmath27 is close to the identity map on @xmath4 . let @xmath59 . if @xmath60 , then choose @xmath61 such that @xmath62 . if @xmath63 , choose @xmath64 such that @xmath65 . define @xmath58 by : @xmath66 for any @xmath67 we have that @xmath68 ( from [ quasi ] ) , so @xmath69 for any @xmath70 . also , from the definition of @xmath23 we get that @xmath71 for any @xmath72 . therefore @xmath26 and @xmath27 are close to the identity maps on @xmath18 and on @xmath4 respectively . the image @xmath73 is coarsely dense in @xmath4 : for any @xmath70 , let @xmath74 . then @xmath75 . using ( [ quasi ] ) and the triangle inequality , one can prove that @xmath23 is a quasi - isometric embedding . for certain metric spaces the converse also holds . if @xmath4 and @xmath18 are length spaces , then any coarse equivalence @xmath16 is a quasi - isometry . let @xmath76 be a metric space and @xmath5 be a discrete group . we say that @xmath5 _ acts coarsely _ on @xmath4 ( or @xmath4 is a _ coarse @xmath5-space _ ) if there are positive constants @xmath31 and @xmath51 , and a map @xmath77 such that 1 . for each @xmath78 , the map @xmath79 is an @xmath50-quasi - isometry of @xmath4 ( with @xmath80 ) . the identity @xmath81 acts as the identity on @xmath4 , so @xmath82 for all @xmath70 . 3 . for each @xmath61 and each @xmath83 , @xmath84 . sometimes the condition ( ii ) is omitted in the definition of coarse actions . notice that any coarse action on @xmath4 in this more general sense is coarsely @xmath5-equivalent ( via the identity map on @xmath4 ) to a coarse action on @xmath4 in which @xmath85 , for all @xmath61 . a coarse @xmath5-action is also called a _ quasi - action _ of @xmath5 on @xmath4 ( see @xcite ) . sometimes we say that @xmath5 has an @xmath50-quasi - action on @xmath4 to specify the constants . 1 . a coarse action of @xmath5 on @xmath4 is _ cobounded _ if there exists @xmath19 such that for each @xmath61 , we have @xmath86 . 2 . a coarse action is _ proper _ if for each @xmath19 , there exists @xmath87 such that for all @xmath88 , we have @xmath89 . let @xmath4 and @xmath18 be coarse @xmath5-spaces . a map @xmath16 is called _ coarsely @xmath5-equivariant _ if there is a constant @xmath90 such that @xmath91 for any @xmath78 and @xmath61 . we say that the actions are _ coarsely @xmath5-equivalent _ if there exists a quasi - isometry @xmath28 which is a coarsely @xmath5-equivariant map . coarsely @xmath5-equivalent @xmath5-actions are also called _ quasi - conjugate _ in the literature . we also remark that the properties _ cobounded _ or _ proper _ for coarse actions are preserved by coarse @xmath5-equivalence . [ milnor - svarc ] let @xmath76 be a proper geodesic metric space . if @xmath92 is a discrete group with a proper , cobounded coarse action on @xmath4 , then @xmath92 is finitely - generated , and @xmath4 is quasi - isometric to the group @xmath92 with word metric . see ghys and de la harpe ( * ? ? ? * proposition 10.9 ) . if @xmath16 is a coarse @xmath5-equivalence , then the inverse quasi - isometry @xmath25 ( as in definition [ coarse - equiv ] ) is also coarsely @xmath5-equivariant : for any @xmath59 , we have @xmath93 , for @xmath94 , by definition of @xmath23 . then if @xmath5 has an @xmath95-quasi - action on @xmath18 , we have @xmath96 here we have assumed that @xmath97 for all @xmath70 , and denoted the constants for @xmath23 from definition [ def : coarse map ] by @xmath98 , etc . let @xmath28 be a quasi - isometry , and suppose that @xmath18 has a coarse action of a discrete group @xmath5 . then @xmath4 admits a coarse action of @xmath5 so that @xmath17 is a coarse @xmath5-equivalence . the given coarse action of @xmath5 on @xmath99 induces a map @xmath100 , via the formula @xmath101 , where @xmath23 denotes a quasi - inverse for @xmath17 . since @xmath17 is a quasi - isometry , it follows that this formula defines a coarse action of @xmath5 on @xmath76 . the induced coarse action on @xmath4 is coarsely @xmath5-equivalent ( by @xmath17 ) to the original coarse action on @xmath18 . a recent result of kleiner and leeb @xcite shows that a coarse action is always coarsely equivalent to an isometric action . [ kleiner - leeb]if a discrete group @xmath5 has an @xmath50-quasi - action @xmath102 on a metric space @xmath103 , then @xmath104 is @xmath105-quasi - conjugate to a canonically defined isometric @xmath5-action on a metric space @xmath99 . the proof of kleiner and leeb @xcite shows that if @xmath5 is a finite group and @xmath76 is a proper metric space , then so is @xmath99 . let @xmath76 be a metric space with an @xmath50-quasi - action of a discrete group @xmath5 . for any @xmath106 , let @xmath107 the sets @xmath108 form an increasing family of subsets of @xmath4 and one can ask if their coarse geometry type stabilizes . the following definition was given in @xcite . [ bdd ] we say that the _ bounded fixed set of a coarse action @xmath109 exists _ , provided that there exists a subspace @xmath110 such that 1 . @xmath111 for some @xmath112 , and 2 . the inclusion map @xmath113 is a coarse equivalence for all @xmath114 . in this case we write @xmath115 . if the coarse type of the subspaces @xmath108 does not stabilize , we say that @xmath6 does not exist . in general , @xmath116 can be empty for all @xmath117 ( and in this case we have @xmath118 ) . for example , take @xmath119 acting on @xmath120 by translations . however , if @xmath5 is finite , then the sets @xmath116 are always nonempty for large @xmath117 . in fact , we have that @xmath121 the inclusion @xmath122 ( @xmath123 ) preserves the metric , so it is a quasi - isometric embedding . if @xmath116 is coarsely dense in @xmath124 , then @xmath125 is a quasi - isometry , therefore a coarse equivalence . note that the subspaces are coarsely @xmath5-invariant , in the sense that @xmath126 , where @xmath127 . the orbit space @xmath128 of an isometric action has a natural ( pseudo)-metric @xmath129 induced by the standard projection @xmath130 from the metric on @xmath4 : @xmath131 if @xmath109 is a quasi - action , we can define @xmath132 , where @xmath133 is any isometric action coarsely equivalent to @xmath134 , as provided by theorem [ kleiner - leeb ] . this construction and the induced metric @xmath129 are both well - defined up to quasi - isometry . [ def : coarsely ineffective ] a coarse @xmath5-action is called _ coarsely ineffective _ if the map @xmath135 is a coarse equivalence . if @xmath5 acts coarsely ineffectively on @xmath4 , there is a coarse equivalence @xmath136 . for any @xmath61 and @xmath78 , we have that @xmath137 , so @xmath138 . also , from the definition of a coarse equivalence , there is a constant @xmath32 so that @xmath139 for any @xmath61.then @xmath140 so @xmath6 exists and is coarsely equivalent to @xmath4 . the converse also holds : if @xmath6 exists and is coarsely equivalent to @xmath4 , then the action is coarsely ineffective . choose @xmath141 so that @xmath116 is coarsely dense in @xmath4 . for any @xmath61 , there is @xmath142 such that @xmath143 . then @xmath144 it follows that @xmath145 since given @xmath146 , we can pick @xmath78 such that @xmath147 . also , from the definition of @xmath129 , we have that @xmath148 . the map @xmath0 is , obviously , surjective . therefore , @xmath0 is a quasi - isometry , so it is a coarse equivalence . let @xmath76 be a metric space with a coarse @xmath5-action . the coarse action of @xmath5 is called _ tame _ if @xmath149 exists for all subgroups @xmath150 in @xmath5 . it would be interesting to find a geometrical condition on @xmath4 which would guarantee that the action of any finite subgroup of quasi - isometries of @xmath4 is tame . we first point out that _ tameness _ of the action is a coarse invariant . [ bdfix - same ] suppose that @xmath103 and @xmath99 are coarse @xmath5-spaces . if @xmath16 is a coarse @xmath5-equivalence and @xmath6 exists , then @xmath151 exists and it is coarsely equivalent to @xmath6 . the existence of @xmath6 implies that there is some @xmath152 such that the inclusion @xmath153 is a coarse equivalence , for any @xmath154 . for any @xmath155 we have @xmath156 thus @xmath157 for @xmath158 . similarly , @xmath159 , for some @xmath160 . then , we have @xmath161 for @xmath162 . @xmath116 is coarsely dense in @xmath124 , so for any @xmath163 , there is @xmath142 such that @xmath164 , for some constant @xmath32 . then @xmath165 is also coarsely dense in @xmath124 : for any @xmath163 we have : @xmath166 it follows that @xmath167 is also coarsely dense in @xmath124 , thus they are coarsely equivalent . lemma [ equiv - to - image ] implies that @xmath168 is coarsely equivalent to @xmath6 for any @xmath169 greater than some value @xmath170 . to finish the proof , we need to show that the inclusion @xmath171 is a coarse equivalence , for any @xmath172 . we have that @xmath173 for some @xmath160 and that @xmath174 is coarsely dense in @xmath124 . this means that @xmath174 is coarsely dense in @xmath175 . so , for any @xmath176 there is a @xmath177 such that @xmath178 . then @xmath179 which completes the proof after this result , and theorem [ kleiner - leeb ] , to study the tameness of a coarse @xmath5-action we can assume that @xmath5 is acting by isometries on @xmath76 . in the rest of the paper we will also assume that @xmath5 is a _ finite group _ , and that @xmath76 is a proper metric space . the next example shows that assuming finite asymptotic dimension ( see section [ dim - asym ] , or roe @xcite ) is not sufficient to ensure that @xmath6 exists . [ theexample ] let @xmath180 . we will construct a space @xmath181 , consisting of infinitely many `` goalposts '' , on which @xmath5 acts freely and the coarse type of @xmath116 does not stabilize . , height=188 ] start with two parallel lines , @xmath182 . let @xmath183 be the set shown in figure 1 below , where the distances @xmath184 , @xmath185 and @xmath186 are equal to @xmath187 and the vertical rays @xmath36 and @xmath188 are at distance @xmath12 from each other . for any @xmath189 , consider the points @xmath190 and @xmath191 in @xmath192 and identify these points to the points @xmath39 and @xmath193 of a copy of @xmath183 such that @xmath183 makes an angle of @xmath194 with the @xmath195-plane ( see figure 2 below ) . , height=188 ] we obtain a space @xmath196 on which @xmath197 acts freely , by interchanging the two branches . the asymptotic dimension of @xmath4 ( with the induced metric ) is at most 3 , because it is a subset of @xmath198 ( see @xcite ) . one can see that the coarse type of the sets @xmath116 does not stabilize : for any integer @xmath12 , we have that @xmath199 , but @xmath183 contains points arbitrarily far away from @xmath124 , for any @xmath200 . so the coarse type of @xmath116 changes when @xmath117 takes integer values , therefore it does not stabilize . here is a condition which ensures that that the bounded fixed set exists . [ thm : existence ] let @xmath76 be a proper geodesic metric space . if @xmath76 admits a proper and cobounded coarse action by a discrete group @xmath92 , then the bounded fixed set @xmath6 exists for any finite subgroup @xmath201 . by the milnor - svarc theorem [ milnor - svarc ] , we may assume that @xmath202 is just the group @xmath92 with the ( left invariant ) word metric . the identity element @xmath203 will be taken as a base - point , and the action of an element @xmath78 on @xmath4 will be denoted @xmath204 , for @xmath205 . let @xmath206 denote the centralizer subgroup for @xmath5 in @xmath92 , and let @xmath207 denote the subspace of @xmath4 consisting of the group elements in the centralizer . since @xmath208 , for all @xmath209 and all @xmath78 , we see that @xmath210 as soon as @xmath211 . we also observe that @xmath150 acts on @xmath116 , for any @xmath212 , defined by the formula @xmath213 , for @xmath155 and @xmath209 , and @xmath214 . now suppose that @xmath141 is large enough so that @xmath215 . let @xmath216 denote the distinct elements in the ball @xmath217 of radius @xmath117 around the identity element @xmath203 . since @xmath218 for any @xmath155 and any @xmath78 , we see that each such element @xmath219 must equal one of the @xmath220 . we fix an ordering of the elements of @xmath5 and obtain a map @xmath221 defined by @xmath222 from @xmath116 to the finite subsets @xmath223 of @xmath37 . since @xmath224 , for all @xmath209 and all @xmath155 it follows that @xmath116 is contained in the union of finitely many @xmath150-orbits . pick elements @xmath225 in @xmath116 representing the distinct @xmath150-orbits . then any @xmath226 can be expressed as @xmath227 , for some @xmath209 and @xmath228 . but then @xmath229 . therefore the inclusion @xmath210 is coarsely dense ( with maximum distance @xmath230 , @xmath228 ) . we have actually shown that @xmath6 is quasi - isometric to the subspace @xmath231 defined in the proof . here are some examples of tame actions : let @xmath232 with the euclidean metric and @xmath5 be a finite group which acts on @xmath233 by isometries . then @xmath234 is the linear subspace of @xmath233 fixed by @xmath5 . ( semi - direct products ) the example which inspired the definition of the bounded fixed set was a semi - direct product @xmath235 , given by an integral representation @xmath236 . in this case , @xmath237 is just the group @xmath92 equipped with the word metric , and @xmath5 acts by left multiplication . then @xmath6 is coarsely equivalent to the fixed sub - representation of @xmath233 induced by the conjugation action @xmath238 of @xmath5 on the normal subgroup @xmath239 . in this section we will show that the action of any finite subgroup of @xmath240 on @xmath241 is tame , where @xmath241 is the hyperbolic @xmath12-space and @xmath240 is its group of isometries . we will use the poincar model , in which the points of hyperbolic @xmath12-space are represented by the points of the open unit ball @xmath242 in @xmath233 . the geodesic lines are the intersection of @xmath242 with those euclidean lines and circles which are orthogonal to the boundary of @xmath242 . an advantage of this model is that the angle between two geodesics issuing from the same point is the euclidean angle between them . let @xmath4 be a metric space and consider two geodesic rays @xmath243 . we say that @xmath244 and @xmath245 are _ asymptotic _ if there is a constant @xmath246 such that @xmath247 for any @xmath248 . one can easily check that this is an equivalence relation . the set of equivalence classes is called the _ boundary _ of @xmath4 and is denoted by @xmath249 . the points of @xmath249 are called _ points at infinity_. the equivalence class of a geodesic ray @xmath244 will be denoted @xmath250 . notice that the images of two asymptotic geodesic rays under any isometry of @xmath4 are again asymptotic geodesic rays . so , if @xmath5 is a group which acts on @xmath4 by isometries , there is an induced @xmath5-action on @xmath249 . let @xmath251 be a geodesic ray issuing from @xmath252 ( i.e. @xmath253 ) . then , for any @xmath254 , there is a unique geodesic ray which issues from @xmath255 and is asymptotic to @xmath244 . see bridson - haefliger @xcite , chapter ii.8 , proposition 8.2 . let @xmath4 be a complete @xmath13 space ( see @xcite , chapter ii.1 , definition 1.1 ) . there is a topology on @xmath256 , called the _ cone topology _ , such that the subspace topology of @xmath4 is the original metric topology . a neighbourhood basis for the points at infinity has the following form : given a geodesic ray @xmath244 and positive constants @xmath257 and @xmath160 , then @xmath258 where @xmath259 is the projection of @xmath260 onto the closed ball @xmath261 defined by : @xmath262 where @xmath245 is the geodesic ray issuing from @xmath263 and passing through @xmath252 ( if @xmath61 ) or representing @xmath252 ( if @xmath264 ) . if @xmath4 is the poincar model of the hyperbolic @xmath12-space , it is known that @xmath249 is homeomorphic to the @xmath265-sphere in @xmath233 . let @xmath5 be a finite subgroup of @xmath240 . then the fixed point set @xmath266 is nonempty ( see @xcite , chapter ii.2 , corollary 2.8 ( 1 ) ) . one can show that if @xmath266 is bounded , then the action on the boundary is free : if some point at infinity is fixed by @xmath5 , let @xmath244 be the geodesic ray issuing from a fixed point @xmath267 , so that @xmath250 is fixed . then @xmath244 and @xmath268 are asymptotic geodesic rays issuing from the same point , thus @xmath269 , for all @xmath270 , therefore @xmath266 contains a geodesic ray , which contradicts our assumption . the converse also holds : if the action on @xmath271 is free , then @xmath266 is bounded . we will prove this claim by contradiction . suppose that @xmath266 is unbounded . then fix a point @xmath272 and , for any @xmath273 , there is @xmath274 so that @xmath275 . consider the geodesic rays @xmath276 issuing from @xmath277 and passing through @xmath278 . then @xmath279 is a sequence in @xmath271 , thus it has a convergent subsequence . since @xmath276 passes through the fixed points @xmath277 and @xmath278 , it follows that @xmath280 for all @xmath281 and @xmath2 . to simplify the notations , we assume that @xmath282 converges to a point @xmath283 . since @xmath5 acts freely on the boundary , @xmath284 if @xmath281 is not the identity in @xmath5 . choose @xmath244 the geodesic ray issuing from @xmath277 so that @xmath285 . then , for any @xmath257 , there is @xmath286 such that @xmath287 for all @xmath288 . choose @xmath289 , so @xmath290 for all @xmath291 bigger or equal to some @xmath292 . fix @xmath293 and choose @xmath2 so that @xmath294 . then @xmath295 which is the desired contradiction . one can slightly adjust this argument to prove the following : @xmath266 is coarsely dense in @xmath296 , for any @xmath141 . suppose that , for some @xmath117 , @xmath266 is not coarsely dense in @xmath296 . thus , for any @xmath297 , there is @xmath298 so that @xmath299 . fix @xmath272 and let @xmath276 be the geodesic ray issuing from @xmath277 and passing through @xmath278 . then @xmath279 is a sequence in @xmath271 , thus it has a convergent subsequence . without loss of generality , we can assume that @xmath282 converges to a point @xmath283 which is not fixed by @xmath5 ( if it were , it would be represented by a geodesic ray contained in @xmath266 , which is clearly false ) . let @xmath244 be the geodesic ray issuing from @xmath277 so that @xmath285 . then , for any @xmath257 , there is @xmath286 such that @xmath287 for all @xmath288 . choose @xmath300 , so @xmath301 for all @xmath291 bigger or equal to some @xmath292 . fix @xmath293 and choose @xmath2 so that @xmath302 . then @xmath303 and @xmath304 which is a contradiction . it follows that the @xmath5-action on @xmath241 is tame and the @xmath5-action on @xmath271 is free iff @xmath266 is bounded and a point at infinity is fixed iff it is a limit point of @xmath266 . the above argument works for any proper @xmath13 space @xmath4 because it has compact boundary ( see @xcite ) . we conclude that the action of any finite subgroup @xmath5 of @xmath305 is tame . the notion of asymptotic dimension was introduced by gromov in @xcite and is a coarse geometry analogue to the topological covering dimension of a compact metric space . this section describes the basic properties of spaces with finite asymptotic dimension ( for more information , see dranishnikov @xcite , roe ( * ? ? ? 3 ) , or bell - dranishnikov @xcite ) . we say that @xmath4 has _ asymptotic dimension _ if for each @xmath160 the space @xmath4 can be decomposed into a union of @xmath307 subsets @xmath308 where each @xmath309 is _ r - disconnected _ : each @xmath309 is a disjoint union of sets of uniformly bounded diameter , and these sets are at least @xmath7 apart from each other @xmath310where @xmath311 . if @xmath312 , then @xmath313 . using the definition , one can check that @xmath314 and @xmath315 . in fact , one can show that @xmath316 ( that means @xmath317 but @xmath318 ) . before stating the main result of this section , we need to review some facts about simplicial complexes ( see bredon @xcite , chapter iii , section 1 ) . an _ abstract simplicial complex _ is a set @xmath246 , whose elements are called vertices , together with a collection of finite nonempty subsets of @xmath246 , called simplices such that : * every vertex is contained in some simplex , * every nonempty subset of a simplex is a simplex . the _ dimension _ of a simplicial complex @xmath246 is @xmath12 if @xmath246 contains an @xmath12-simplex , but no @xmath319-simplices or @xmath320 if @xmath246 contains @xmath12-simplices for any @xmath321 . a _ simplicial map _ @xmath322 is a function from the vertices of @xmath323 to the vertices of @xmath324 such that the image of any simplex of @xmath323 is a simplex of @xmath324 . two simplicial maps @xmath325 are _ contiguous _ if , for any simplex @xmath326 , @xmath327 and @xmath328 belong to a common simplex of @xmath324 . two contiguous maps induce the same map in homology . let @xmath150 be the hilbert space @xmath329 . define a map @xmath330 by sending any vertex @xmath331 to the corresponding unit vector @xmath332 . for any simplex @xmath333 of @xmath246 , its _ geometric realization _ ( or the _ closed simplex _ ) @xmath334 is the subset of @xmath150 consisting of all convex combinations of @xmath335 ( all linear combinations @xmath336 with positive coefficients such that @xmath337 ) . the _ geometric realization _ ( or the _ polyhedron _ ) @xmath338 of @xmath246 is the union of the geometric realizations of its simplices . the induced metric @xmath339 on @xmath338 is called the metric of _ barycentric coordinates_. the _ length metric _ ( see @xcite ) associated to the metric of barycentric coordinates is called the _ intrinsic metric _ on @xmath246 . a simplicial complex is _ locally finite _ if any vertex belongs to finitely many simplices . the _ barycentric subdivision _ of a simplicial complex @xmath246 is the simplicial complex @xmath340 whose vertices are the simplices of @xmath246 and whose simplices are the sets @xmath341 of vertices of @xmath340 ( simplices of @xmath246 ) , such that , after reordering , @xmath342 ( each @xmath343 is a face of @xmath344 ) . there is a canonical homeomorphism @xmath345 ( see @xcite , chapter ii , lemma 6.2 ) . let @xmath5 be a finite group which acts on @xmath246 such that each transformation is a simplicial map . such an action is called a _ simplicial action _ and @xmath246 with a simplicial action is called a _ simplicial @xmath5-complex_. [ regular complex ] a simplicial @xmath5-complex @xmath246 is called _ regular _ if it satisfies the following condition : * for any subgroup @xmath346 , if @xmath347 are in @xmath150 and @xmath348 and @xmath349 are simplices of @xmath246 , then there is an element @xmath350 such that @xmath351 , for all @xmath125 . the condition ( a ) implies the following equivalent conditions : * for @xmath331 and @xmath78 , if @xmath352 and @xmath353 belong to the same simplex , then @xmath354 ; * for any simplex @xmath355 of @xmath246 and @xmath78 , @xmath281 fixes every vertex in @xmath356 . if @xmath246 is a simplicial @xmath5-complex , then the induced action on the barycentric subdivision @xmath340 satisfies . if the action on @xmath246 satisfies , then the action on @xmath340 is regular . see bredon @xcite , section iii , proposition 1.1 . let @xmath357 be a map from @xmath4 to the polyhedron of a simplicial complex @xmath246 . we say that @xmath17 is _ uniformly cobounded _ if there is a uniform finite bound for the diameter of the inverse image under @xmath17 of the star of any vertex of @xmath246 . the _ degree _ of a covering @xmath358 is the maximum number of members of @xmath358 with nonempty intersection . the following result gives different characterizations for spaces with finite asymptotic dimension . the definition of a coarsening system " is given in the next section ( see definition [ coarsening ] ) . [ finit - cech ] let @xmath4 be a proper metric space . then the following are equivalent : * @xmath4 has asymptotic dimension @xmath306 ; * @xmath4 admits a coarsening system consisting of coverings of degree @xmath359 ; * for any @xmath360 there is an @xmath361-lipschitz and uniformly cobounded map from @xmath4 to an @xmath169-dimensional polyhedron equipped with the metric of barycentric coordinates . if @xmath4 is a geodesic space , these conditions are also equivalent to * for any @xmath360 there is an @xmath361-lipschitz and effectively proper map from @xmath4 to an @xmath169-dimensional polyhedron equipped with the intrinsic metric . see roe ( * ? ? ? * theorem 9.9 ) . in this section we will recall the definition of coarse homology @xcite , @xcite , @xcite . all metric spaces considered are proper . [ def : uniform ] a covering @xmath358 of @xmath4 with the property that any @xmath362 is relatively compact and any bounded subset of @xmath4 intersects only finitely many sets in @xmath358 is called a _ uniform covering _ of @xmath4 . it follows that any uniform covering is locally finite . let @xmath358 be a uniform covering of @xmath4 and let @xmath363 be a partition of unity subordinate to @xmath358 ( which exists because , by stone s theorem , any metric space is paracompact ) . let @xmath364 be the nerve of @xmath358 and regard its geometrical realization @xmath365 as a subspace of the hilbert space @xmath366 . define a map @xmath367 by @xmath368 where @xmath369 is the unit vector of @xmath150 associated to @xmath370 . the maps @xmath371 are continuous , so the map @xmath372 is also continuous : if @xmath373 in @xmath4 , then @xmath374 , for any @xmath375 , thus @xmath376 . [ coarsening ] a _ coarsening system _ of @xmath4 ( or an _ anti - ech system _ of @xmath4 ) is a sequence of uniform coverings @xmath377 of @xmath4 for which there exists an increasing sequence of real numbers @xmath378 such that for all @xmath12 , * each set @xmath379 has diameter less than or equal to @xmath380 , * the covering @xmath381 has a lebesgue number greater than or equal to @xmath380 . any proper metric space admits a coarsening system @xmath377 ( see roe ( * ? ? ? * lemma 3.15 ) ) . it follows that the covering @xmath382 is a refinement of @xmath381 , so there exist refinement projections @xmath383 , for any @xmath12 . these maps are called _ coarsening maps_. from now on , we will include a choice of such maps as part of a coarsening system . if @xmath358 is a covering of @xmath4 , one can define a simplicial complex called the nerve of @xmath358 and denoted by @xmath364 . the vertices are the sets @xmath375 of the covering and the simplices are finite non - empty subsets @xmath384 of @xmath358 with non - empty intersection . if the nerve of a covering is locally finite , then the covering is locally finite , but the converse is not true . however , if the covering is uniform , then its nerve is a locally finite simplicial complex . the coarsening maps @xmath385 induce proper simplicial maps @xmath386 . a different choice of the coarsening maps induces contiguous maps . the _ locally finite _ homology of a simplicial complex @xmath246 is defined using chains that are infinite , locally finite formal linear combinations of oriented simplices of @xmath246 . for example , consider the 1-dimensional simplicial complex @xmath246 whose set of vertices is @xmath387 and the simplices are of the form @xmath388 for any @xmath389 ( it follows that @xmath390 ) . the 0-dimensional locally finite homology group is trivial ( each vertex is the boundary of an infinite 1-chain consisting of all the simplices to the left of it ) , whereas the 1-dimensional locally finite homology group is non - trivial ( the sum of all the 1-simplices is a generator ) . in fact , for any field @xmath391 , it is true that @xmath392 let @xmath377 be a coarsening system of @xmath4 and let @xmath393 be the locally finite homology theory . then the _ coarse homology _ of @xmath4 is given by : @xmath394 let @xmath395 and denote by @xmath396 the subcomplex of @xmath397 consisting of those simplices @xmath398 such that @xmath399 one can check that the complex @xmath396 is isomorphic to the nerve of @xmath400 . the coverings @xmath401 form a coarsening system of @xmath39 ( they have increasing lebesgue number and sets with uniformly bounded diameter ) , thus @xmath402 if @xmath395 and @xmath377 is a coarsening system of @xmath4 , then the _ relative coarse homology _ is defined by @xmath403 two different coarsening systems of @xmath4 will give rise to canonically isomorphic coarse homology groups , therefore the coarse homology is independent of the coarsening system ( see @xcite ) . two close maps @xmath404 induce the same map on @xmath405 . see higson - roe ( * ? ? ? * proposition 2.2 ) . an easy consequence of this proposition is that each coarse equivalence induces an isomorphism in the coarse homology . in this section , we will define the appropriate equivariant version of a coarsening system , for @xmath76 a coarse @xmath5-space ( see section [ sec : quasi ] ) and @xmath5 a finite group . a _ regular @xmath5-coarsening system _ is the anti - ech analogue of the systems of regular @xmath5-coverings constructed in bredon ( * ? ? ? * iii.6 ) . theorem [ regular coarsening system ] shows the existence of such systems of coverings under certain assumptions . [ def : regular coarsening system ] let @xmath4 be a coarse @xmath5-space , for @xmath5 a finite group . we say that a covering @xmath358 of @xmath4 is is _ @xmath5-invariant _ or that @xmath358 is a @xmath5-_covering _ , provided that 1 . @xmath406 , for all @xmath78 . 2 . @xmath407 , for all @xmath408 and all @xmath409 . if @xmath5 acts by isometries on @xmath4 , then @xmath410 is also a covering of @xmath4 and the second condition is automatic . in that case , @xmath358 is a @xmath5-covering if and only if @xmath411 for all @xmath78 . if @xmath4 is a coarse @xmath5-space , there is a natural induced @xmath5-action on the nerve of a @xmath5-invariant covering , @xmath412 making @xmath364 a simplicial @xmath5-complex . we will always be working with uniform coverings ( see definition [ def : uniform ] ) , so that the nerves @xmath364 will also be locally - finite . let @xmath358 be a uniform @xmath5-invariant covering of @xmath4 . then @xmath358 is a _ regular _ @xmath5-covering if its nerve @xmath364 is a regular @xmath5-complex . a _ regular @xmath5-coarsening system _ for @xmath4 is coarsening system @xmath377 , such that @xmath382 is a regular @xmath5-covering , for any @xmath413 . we observe that the existence of a regular @xmath5-coarsening system is a coarse invariant . [ lem : coarsening ] suppose that @xmath103 and @xmath99 are coarse @xmath5-spaces . if @xmath16 is a coarse @xmath5-equivalence and @xmath18 admits a regular @xmath5-coarsening system , then so does @xmath4 . if @xmath377 is a regular @xmath5-coarsening system for @xmath18 , then for each of the coverings @xmath382 we let @xmath414 . from the metric properties of @xmath17 it is clear that each covering @xmath415 is a uniform @xmath5-covering , and that its nerve @xmath416 is a regular @xmath5-complex . by passing to a subsequence of the coverings @xmath417 , if necessary , we can obtain an anti - ech system , and therefore a regular @xmath5-coarsening system for @xmath4 . we will now show that any proper geodesic metric space @xmath4 with finite asymptotic dimension admits a regular coarsening system , whose @xmath5-invariant coverings have nerves with uniformly bounded dimension . in the rest of this section , we assume that @xmath5 acts by isometries on @xmath4 . let @xmath358 be a uniform covering of @xmath4 , and @xmath5 a finite group of isomtries of @xmath4 . then the collection @xmath418 is a @xmath5-invariant uniform covering of @xmath4 . it is clear that the sets in @xmath419 form a @xmath5-invariant covering by construction . since @xmath5 acts by isometries , @xmath420 . thus the sets in @xmath419 are bounded subsets of a proper metric space , and therefore relatively compact . let @xmath395 be a bounded set . suppose that @xmath39 intersects infinitely many sets from @xmath419 . then , since @xmath5 is finite , there exists @xmath78 such that @xmath421 for infinitely many sets @xmath422 . therefore the bounded set @xmath423 intersects infinitely many sets from @xmath358 , which contradicts that @xmath358 is uniform . if @xmath377 is a coarsening system of @xmath4 , then @xmath424 is a sequence of uniform @xmath5-coverings of @xmath4 . we will show that the coverings @xmath424 also form a coarsening system for @xmath4 . if @xmath425 , we may assume that @xmath426 for any @xmath12 ( see theorem [ finit - cech ] ) . [ inv.coars.syst ] if @xmath377 is a coarsening system of @xmath4 , the sequence @xmath424 is a coarsening system for @xmath4 . moreover , if @xmath427 , then the covering @xmath424 has degree @xmath428 , where @xmath429 . we check the properties from the definition of a coarsening system . for any @xmath413 , @xmath78 and @xmath430 , we have that @xmath431 , so the sets in @xmath432 have uniformly bounded diameter . the covering @xmath382 is a refinement of @xmath432 , so the lebesgue number of @xmath432 is at least @xmath433 . therefore @xmath424 is a coarsening system . assume that @xmath434 , and suppose that there are at least @xmath435 sets from @xmath432 with nonempty intersection . then , for some @xmath78 , there are at least @xmath436 sets @xmath437 of @xmath382 such that @xmath438 . therefore @xmath439 , so the nerve @xmath397 has dimension at least @xmath307 , which is impossible . so far , we have constructed a coarsening system consisting of @xmath5-invariant coverings whose nerves have uniformly bounded dimension . to obtain a regular @xmath5-coarsening system , we need to introduce some new notions ( see @xcite ) . let @xmath358 be a locally finite @xmath5-invariant covering of @xmath4 and let @xmath440 be a partition of unity subordinate to @xmath358 . then @xmath441 is called a _ g - partition of unity _ if @xmath442 , for all @xmath281 , @xmath252 and @xmath375 . if @xmath443 is any partition of unity subordinate to the @xmath5-invariant covering @xmath358 , we define a @xmath5-partition @xmath441 by putting @xmath444 if each @xmath445 is @xmath361-lipschitz , then @xmath446 is also @xmath361-lipschitz : @xmath447 recall that the map @xmath448 defined by : @xmath449 is continuous . if @xmath358 is a @xmath5-invariant open covering of @xmath4 and @xmath450 is a @xmath5-partition of unity subordinate to @xmath358 , then @xmath372 is also equivariant : @xmath451 the following observation will be used in the proof of the main result of this section . if @xmath16 is a @xmath361-lipschitz map between two metric spaces @xmath4 and @xmath18 and if @xmath452 is an open covering of @xmath18 with lebesgue number @xmath7 , then @xmath453 is an open covering of @xmath4 with lebesgue number at least @xmath454 . [ regular coarsening system ] for any proper geodesic metric space @xmath4 with asymptotic dimension @xmath455 , and for any finite group @xmath5 which acts on @xmath4 by isometries , there is a regular @xmath5-coarsening system of @xmath4 consisting of @xmath5-invariant coverings of degree at most @xmath456 , where @xmath429 . let @xmath424 be a @xmath5-invariant coarsening system of @xmath4 , given by lemma [ inv.coars.syst ] . for any @xmath360 there exists @xmath457 such that @xmath458 is @xmath361-lipschitz and effectively proper map , where @xmath459 is equipped with the intrinsic metric ( from theorem [ finit - cech ] ) . moreover , we can assume that @xmath372 is @xmath5-equivariant ( see equation [ epsilon - lip ] ) . let @xmath460 be the second barycentric subdivision of @xmath461 and regard the polyhedra @xmath462 as equal . then @xmath460 is a locally finite regular @xmath5-complex . denote by @xmath463 the covering of @xmath4 by inverse images of open vertex stars of @xmath464 . we will show that a subsequence of @xmath465 form a coarsening system with the desired properties . since @xmath372 is continuous , @xmath463 is an open covering whose sets are uniformly bounded ( because @xmath372 is effectively proper ) . its nerve is isomorphic to @xmath460 ( since @xmath372 is @xmath5-equivariant ) which is a regular @xmath5-complex , thus @xmath463 is a regular @xmath5-covering . let @xmath39 be a bounded subset of @xmath4 . then @xmath466 is a bounded subset of the proper metric space @xmath459 . then the set @xmath467 is also bounded . for any vertex @xmath468 , its open star @xmath469 intersects @xmath466 if and only if @xmath470 . since @xmath471 is compact , @xmath472 contains finitely many vertices , so @xmath466 intersects only finitely many open vertex stars of @xmath460 . thus , @xmath463 is a uniform covering of @xmath4 . the covering of @xmath464 by open vertex stars has a positive lebesgue number @xmath7 ( depending only on the dimension of @xmath460 ) , then @xmath463 has a lebesgue number at least @xmath454 . since we can choose @xmath361 arbitrarily small , we can construct regular , uniform coverings of @xmath4 with arbitrarily large lebesgue number . we assume that @xmath76 is a proper metric space , equipped with a coarse @xmath5-action of a finite group @xmath5 . in this section , we show how a regular @xmath5-coarsening system for @xmath4 ( see definition [ def : regular coarsening system ] ) can be used to compute the coarse homology of the bounded fixed set , whenever the bounded fixed set exists . the answer is given in corollary [ rhomap ] . let @xmath377 be a regular @xmath5-coarsening system for @xmath4 . by lemma [ lem : coarsening ] , the existence of a regular @xmath5-coarsening system is a coarse invariant . hence by theorem [ kleiner - leeb ] , we will assume that @xmath5 acts by isometries on @xmath4 . recall that @xmath473 and suppose that @xmath6 exists ( see definition [ bdd ] ) . consider the nerve @xmath397 and recall that @xmath474 denotes the subcomplex of @xmath397 consisting of those simplices @xmath398 such that @xmath475 @xmath474 is isomorphic to the nerve of @xmath476 . these coverings form a coarsening system of @xmath124 , so we have that @xmath477 we _ define _ the coarse homology of @xmath6 to be @xmath478 we would like to compare @xmath479 to @xmath480 , where @xmath481 denotes the subcomplex of @xmath397 spanned by those vertices @xmath482 invariant under the action of @xmath5 . in order to achieve this , we will use their intersection @xmath483 . explicitly : @xmath484 we will need the following algebraic lemma from bredon @xcite . [ direct limit diagram ] let @xmath485 be a directed set and let @xmath486 and @xmath487 be direct systems of abelian groups based on @xmath485 . let @xmath488 be a homomorphism of directed systems . assume that for each index @xmath125 there is an index @xmath489 and a homomorphism @xmath490 such that the diagram @xmath491^{{\theta}_i } \ar[d]_{f_{ij } } & { b_i } \ar[d]^{g_{ij } } \ar[dl]\|-{h_{ij } } \\ { a_j } \ar[r]_{{\theta}_j } & { b_j } } \ ] ] commutes . then the induced map @xmath492 is an isomorphism . see bredon @xcite , chapter iii , lemma 6.4 . with the above notations , we have that @xmath493 take @xmath494 and pick @xmath495 . then diam@xmath496 for any @xmath125 from @xmath497 to @xmath498 . choose @xmath499 such that @xmath500 , then @xmath501 , for some @xmath502 . since @xmath503 and @xmath504 is @xmath5-regular , it follows that @xmath505 is @xmath5-invariant , which means that @xmath506 . in this way we defined a map @xmath507 which makes the following diagram commute : @xmath508 \ar[d ] & { k({\mathscr u}_n\| x^g_r ) } \ar@{-->}[dl ] \ar[d ] \\ { k({\mathscr u}_m)^g_r } \ar[r ] & { k({\mathscr u}_m\| x^g_r ) } } \ ] ] where the horizontal maps are inclusions and the vertical ones are the coarsening maps associated to a coarsening system . the proof follows from lemma [ direct limit diagram ] . [ rem : fixed covering ] the proof just given shows that for any @xmath509 , and any @xmath510 , there exists an @xmath511 so that @xmath512 is a covering of @xmath124 . to see this , note that @xmath513 is a covering and each @xmath514 with @xmath515 is shown to be contained in @xmath512 , where @xmath511 is chosen so that @xmath516 . @xmath517 we have that for any @xmath160 , @xmath518 is a subcomplex of @xmath481 and , for any @xmath519 , @xmath518 is a subcomplex of @xmath520 , so we have a commutative diagram : @xmath521 \ar[d ] & { \lim\limits_{n\to\infty } { h^{l{\hskip-1pt}f}}_\bigl(k({\mathscr u}_n)^g\bigr ) } \\ { \lim\limits_{n\to\infty } { h^{l{\hskip-1pt}f}}_\bigl(k({\mathscr u}_n)^g_{r'}\bigr ) } \ar[ur ] & } \ ] ] from the definition of the direct limit we get a map @xmath522 the map @xmath523 is an isomorphism . we will show first that @xmath523 is surjective . let @xmath524 be an element in @xmath525 , then @xmath526 is a cycle in @xmath527 of the form : @xmath528 where @xmath529 is a simplex in @xmath530 . then @xmath531 is an invariant set with diameter less than @xmath532 , so the diameter of the orbit of any point in this set is bounded by @xmath532 . thus @xmath533 for any @xmath534 . since @xmath7 is independent of @xmath535 , the same one will work for all @xmath535 in @xmath526 which means that @xmath526 is a cycle in @xmath536 which defines an element in the group @xmath537 . by construction , @xmath523 maps this element to @xmath538 , so @xmath523 is surjective . to show that @xmath523 is injective , let @xmath539 there exists some @xmath7 such that @xmath540 is mapped to the zero element in @xmath525 . let @xmath541 and suppose that @xmath542 is a boundary in @xmath527 , so @xmath543 where @xmath526 is a chain in @xmath544 . an argument similar to the one used above shows that @xmath545 for some @xmath546 . since the map @xmath547 is induced by the inclusion @xmath548 , it follows that @xmath549 in @xmath550 , which implies that @xmath551 in @xmath537 . therefore , we obtain a characterization of the coarse homology of the bounded fixed set . [ rhomap ] let @xmath4 be a coarse @xmath5-space , for @xmath5 a finite group . assume that @xmath4 admits a regular @xmath5-coarsening system @xmath377 . then @xmath552 provided that the bounded fixed set exists . let ( @xmath553 ) be a proper metric space and let @xmath5 a finite group which acts on @xmath4 by isometries . define a map @xmath554 by @xmath555 with @xmath556 and @xmath557 , where @xmath130 is the canonical projection . this is independent of the choice of @xmath558 and @xmath559 because each map @xmath281 is an isometry . the map @xmath560 is a metric and ( @xmath561 ) is a proper metric space . it is obvious that @xmath560 is symmetric and takes nonnegative values . suppose that @xmath562 . then @xmath563 , for some @xmath78 and we have that @xmath564 . let @xmath565 . then @xmath566 for any @xmath567 . since @xmath5 is finite , @xmath568 for some @xmath569 . it follows that @xmath570 , for any @xmath78 . if we choose @xmath281 so that @xmath571 , we obtain the triangle inequality : @xmath572 for the second part , notice that @xmath0 is continuous with respect to the metric topologies on @xmath4 and @xmath128 ( if @xmath573 , then @xmath574 as well , since @xmath575 , for any @xmath576 ) . we will denote by @xmath577 an open @xmath7-ball in either metric space and by @xmath578 its closure . we want to show that @xmath579 for any @xmath160 . for later reference , we will show first that @xmath580 , which implies that @xmath0 is an open map : if @xmath581 is open , then @xmath582 so @xmath583 is open in @xmath128 . choose @xmath584 , so there is @xmath585 with @xmath557 . since @xmath586 , then @xmath587 , where @xmath588 , so @xmath589 . start with @xmath589 , so @xmath587 , which means that , for some @xmath78 , @xmath590 or @xmath591 , thus @xmath592 . now let @xmath593 , so there exists @xmath594 with @xmath557 . then , for any @xmath257 , we have that @xmath595 . this implies that @xmath596 , thus @xmath597 . if @xmath597 , then for any @xmath257 , we have @xmath598 . for any @xmath599 , there is @xmath600 , so there exist @xmath601 such that @xmath602 and @xmath603 . since @xmath5 is finite , there is some @xmath604 such that @xmath605 for infinitely many @xmath12 . for any @xmath606 , there is such an index @xmath12 with @xmath607 , therefore @xmath608 which implies that @xmath609 , thus @xmath610 . let @xmath611 be closed and bounded , so @xmath612 , for some @xmath613 and some finite @xmath160 . if @xmath614 ( for some @xmath558 ) , then @xmath39 is closed ( since @xmath0 is continuous ) and bounded in @xmath4 . thus @xmath39 is compact , so @xmath615 is also compact , which completes the proof . in general , @xmath616 , for example , if @xmath617 is the discrete metric . let @xmath358 be a regular @xmath5-covering of @xmath4 , and for each @xmath409 let @xmath618 denote the union of the orbit of @xmath375 in @xmath4 . let @xmath619 denote the covering of @xmath128 by the sets @xmath620 and indexed by @xmath621 . if @xmath622 but @xmath623 for any @xmath281 , then @xmath624 and @xmath625 are regarded as different elements of the covering . if @xmath358 is a regular @xmath5-covering of @xmath4 , then the assignment @xmath626 gives an isomorphism of the simplicial complexes @xmath627 see bredon @xcite , chapter iii , proposition 6.2 . suppose that @xmath358 and @xmath628 are regular @xmath5-coverings of @xmath4 such that @xmath628 is a refinement of @xmath358 . then there is a refinement projection @xmath629 that is _ equivariant _ ; that is , @xmath630 and @xmath631 . when @xmath632 is equivariant , the simplicial map @xmath633 is also equivariant . as an immediate consequence , given a regular @xmath5-coarsening system @xmath377 , we can assume that all the coarsening maps @xmath383 are equivariant . let @xmath377 be a regular @xmath5-coarsening system for @xmath4 . then @xmath634 is a coarsening system of @xmath128 . let @xmath130 be the canonical projection and notice that @xmath635 . it follows that any @xmath636 is open ( since @xmath0 is an open map ) and @xmath637 ( see the definition of the metric @xmath560 on @xmath128 at the beginning of this section ) . let @xmath638 with diam @xmath639 . there exists some point @xmath613 such that @xmath640 , the open ball of radius @xmath117 around @xmath641 . let @xmath61 such that @xmath556 . then @xmath642 , so there exists @xmath643 with @xmath644 . it follows that @xmath645 can not intersect infinitely many sets from @xmath646 because that would imply that one of the translates of @xmath39 intersects infinitely many sets from @xmath382 ( since @xmath5 is finite and @xmath382 is regular ) , which contradicts that @xmath382 is uniform . thus @xmath646 is a uniform covering . one can also see that the lebesgue number of @xmath646 is at least half of the lebesgue number of @xmath382 : if @xmath638 with diam @xmath647 , then @xmath648 , for some @xmath613 . as above , there is a set @xmath649 with @xmath644 and @xmath556 . then diam @xmath650 , so @xmath651 for some @xmath482 , thus @xmath652 . from the last two results we have the formula : [ orbit space ] @xmath653 [ rem : orbit ] if the bounded fixed set @xmath6 exists , then there exists @xmath654 such that its quasi - isometry type is represented by any subspace @xmath116 , for @xmath655 . in particular , the @xmath5-action on @xmath116 is coarsely ineffective ( see section [ sec : four ] ) , hence the projection map @xmath656 induces a coarse equivalence @xmath657 . we will use this observation to consider @xmath6 also as a subspace of @xmath128 . for completeness , we will describe the transfer map for coarse homology . let @xmath377 be a regular @xmath5-coarsening system of @xmath4 and @xmath397 the nerve of @xmath382 . then we have the transfer map for locally finite homology ( compare ( * ? ? ? * iii.2 ) ) : @xmath658 and , for @xmath659 we have the induced commutative diagram @xmath660 ^ -{\mu _ * } \ar[d ] & { { h^{l{\hskip-1pt}f}}\bigl(k({\mathscr u}_n)\bigr ) } \ar[d ] \\ { { h^{l{\hskip-1pt}f}}\bigl(k({\mathscr u}_m)/g\bigr ) } \ar[r]^-{\mu _ * } & { { h^{l{\hskip-1pt}f}}\bigl(k({\mathscr u}_m)\bigr ) } } \ ] ] by naturality of the transfer . passing to the direct limit , and applying corollary [ orbit space ] , we obtain the transfer for coarse homology @xmath661 which satisfies @xmath662 where @xmath663 is the canonical projection . we can define coarse homology for other coefficient groups @xmath664 , just by tensoring the simplicial chain complexes of the @xmath397 with @xmath664 ( over @xmath387 ) , and passing to the direct limit of the locally - finite homology . the formulas above continue to hold , and we have the usual consequence : if @xmath664 is a field of characteristic 0 or prime to @xmath665 , then @xmath666 is an isomorphism , as is @xmath667 in the statement , the coefficients @xmath664 have trivial @xmath5-action , and @xmath668 denotes the fixed set of the induced @xmath5-action on the coarse homology ( see bredon ( * ? ? ? * iii.2 ) for the usual transfer and its properties ) . the formulas can be used to prove a coarse version of a result of floyd ( * ? ? ? * iii.5.4 ) . the goal of p. a. smith theory @xcite is to relate the mod @xmath0 homology of a regular simplicial @xmath5-complex @xmath246 and its fixed set @xmath669 , in the case when @xmath5 is a finite @xmath0-group and @xmath0 is a prime ( see the borel seminar @xcite ) . we will follow the procedure from bredon ( * ? ? ? * chap . iii ) to pass from a simplicial complex to a coarse space , and establish the analogue of the p. a. smith theory for coarse homology . we first review the classical p. a. smith theory . let @xmath0 be a prime , and let @xmath670 , denote the cyclic group of order @xmath0 . fix a generator @xmath671 , and define elements @xmath672 in the group ring @xmath673 . if @xmath674 , for @xmath675 , then we define @xmath676 . since @xmath677 , we have @xmath678 and @xmath679 . the elements @xmath680 and @xmath681 , as well as @xmath682 and @xmath683 , are considered as operators on the mod @xmath0 homology @xmath684 of a regular simplicial @xmath5-complex . in the rest of the section , all homology groups will be understood with @xmath685-coefficients . let @xmath246 denote a regular simplicial @xmath5-complex , and @xmath686 a @xmath5-invariant subcomplex . the simplicial chain complex of the pair @xmath687 is denoted @xmath688 . smith special homology groups _ @xmath689 are defined for each @xmath674 , @xmath675 , as the mod @xmath0 homology of the chain complex @xmath690 . the main results of the classical p. a. smith theory are based on the exact sequences of chain complexes : @xmath691 valid for @xmath692 and @xmath693 . to generalize smith theory to our setting , we first apply locally finite homology to these chain complexes . let @xmath246 denote a regular simplicial @xmath5-complex , and @xmath694 a @xmath5-invariant subcomplex . then the _ locally finite _ smith special homology groups @xmath695 are defined , for @xmath692 and @xmath693 . the smith special homology group for @xmath696 is related to the orbit spaces ( denoted @xmath697 and @xmath698 ) , via the formula @xmath699 we have the following exact triangle ( locally finite homology with @xmath685 coefficients ) . [ exact triangle ] for @xmath674 , @xmath675 , there is an exact triangle @xmath700_{\varrho _ * } & \\ { { h^{l{\hskip-1pt}f , \varrho}}_(k , l ) } \ar[rr]^(0.4){\delta _ } & & { { h^{l{\hskip-1pt}f , \bar{\varrho}}}_(k , l)\oplus { h^{l{\hskip-1pt}f}}_(k^g , l^g ) } \ar[ul]_{i _ * } } \ ] ] where the horizontal map @xmath701 has degree @xmath702 and the other maps @xmath703 , @xmath704 have degree @xmath497 . see bredon ( * ? ? ? * iii.3.3 ) . the map @xmath703 is induced from the direct sum of the inclusions of the subcomplexes @xmath705 and @xmath706 in @xmath707 in . the first application of this exact triangle is to establish thesmith inequalities " for locally finite homology . [ ineg ] let @xmath246 be a finite - dimensional regular @xmath5-complex , and @xmath686 a @xmath5-invariant subcomplex . then , for any @xmath321 and for any @xmath708 with @xmath709 , @xmath710 if the right - hand side is finite , then the left - hand side is also finite , and in particular @xmath711 is finite , for all @xmath712 . see bredon ( * ? ? ? * iii.4.1 ) or floyd @xcite . by rank " we mean the rank ( or dimension ) of the indicated homology groups as @xmath685-vector spaces . the last part follows from . we now pass from locally finite homology to coarse homology when @xmath713 acts by isometries on a proper metric space . our results will apply equally to coarse @xmath5-spaces : we again use theorem [ kleiner - leeb ] to justify replacing a coarse @xmath5-action by a coarsely equivalent isometric @xmath5-action . let @xmath4 be a proper metric space with a coarse action of a finite group @xmath5 . we say that @xmath4 is @xmath5-_finitistic _ if the action admits a regular @xmath5-coarsening system @xmath377 whose nerves @xmath397 have uniformly bounded dimension , for all @xmath12 . for the remainder of this section , @xmath4 is a proper metric space and @xmath670 acts by isometries on @xmath4 . we will also assume that @xmath4 is @xmath5-finitistic and that the bounded fixed set @xmath6 exists , or the @xmath5-action is _ tame _ ( see definition [ bdd ] ) . for any regular @xmath5-coarsening system @xmath377 associated to the @xmath5-action on @xmath4 , we have seen that any two equivariant coarsening maps induce contiguous equivariant simplicial maps and the induced chain maps are equivariantly chain homotopic . thus we can define the _ smith special homology groups _ for coarse homology @xmath714 for @xmath674 and @xmath693 . the smith homology groups are natural with respect to equivariant coarse maps . passing to the limit in the smith triangle ( see theorem [ exact triangle ] ) associated to @xmath715 , and by using corollary [ rhomap ] , we obtain the smith exact triangle for coarse homology : @xmath716_{\varrho _ * } & \\ { { hc}^\varrho_(x ) } \ar[rr]^(0.4){\delta _ } & & { { hc}^{\bar{\varrho}}_(x)\oplus{hc}_(x^g_{bd } ) } \ar[ul]_{i _ * } } } \ ] ] where the horizontal map has degree @xmath702 and the other maps have degree @xmath497 . this triangle is exact because the direct limit functor is exact . note that the isomorphism from @xmath717 is natural , and thus @xmath718 where @xmath6 is considered via the projection map as a subspace of @xmath128 ( see remark [ rem : orbit ] ) . we now establish the coarse version of the smith inequalities . [ thm : pasmith ] let @xmath4 be a proper metric space and @xmath5 a cyclic group of prime order @xmath0 , with a coarse action on @xmath4 . assume that @xmath4 is @xmath5-finitistic and that @xmath6 exists then , for any @xmath321 and for any @xmath708 with @xmath709 , @xmath719 if the right - hand side is finite , then the left - hand side is also finite , and in particular @xmath720 is finite , for all @xmath721 . since @xmath4 is @xmath5-finitistic , there is a regular @xmath5-coarsening system @xmath377 for @xmath4 and an integer @xmath498 , such that @xmath722 for all @xmath12 . thus @xmath723 for @xmath724 and , similarly @xmath725 since the nerves @xmath397 do not contain any @xmath726 simplices , for any @xmath12 . let @xmath727 from theorem [ ineg ] it follows that , for any @xmath12 , @xmath728 the above discussion shows that @xmath729 for @xmath730 and for any @xmath12 . passing to the limit over @xmath12 we obtain @xmath719 the last part follows from equation . in the next statement , the notation @xmath731 , @xmath732 , and @xmath733 means the euler characteristic with respect to their coarse homology , and @xmath734 denotes its total rank " ( the sum of the ranks over @xmath685 of all the coarse homology groups of @xmath4 ) . [ thm : coarse euler ] let @xmath4 and @xmath5 be as before with @xmath735 . then @xmath736 therefore @xmath737 . compare bredon ( * ? ? ? * iii.4.3 ) or floyd @xcite . the exact sequence of the inclusion @xmath738 and theorem [ thm : pasmith ] shows that @xmath739 , and hence all three euler characteristics are defined . we have the relation @xmath740 by considering the long exact sequence of the pair @xmath741 are a chain complex with zero homology . let @xmath742 , and note that @xmath743 from the smith exact triangle for @xmath744 . now from the second exact sequence of smith chain complexes in one obtains the equations @xmath745 , for @xmath746 . by adding all these equations , and using @xmath747 , we get @xmath748 but @xmath749 by , hence @xmath750 which gives the required formula . the most often used result of the classical p. a. smith theory is the application to actions on mod @xmath0 homology spheres . for our coarse version of this result , we observe that euclidean space @xmath3 is a coarse @xmath2-sphere . we use this terminology because the coarse homology of @xmath3 is equal to the reduced ordinary homology of an @xmath2-sphere , by . more generally , a ( mod @xmath0 ) _ coarse @xmath2-sphere _ is a metric space with the same ( mod @xmath0 ) coarse homology as @xmath3 . we have the following application of coarse p. a. smith theory . [ final ] let @xmath4 be a proper metric space , which is a ( mod @xmath0 ) coarse homology @xmath2-sphere , for some prime @xmath0 . let @xmath5 be a finite @xmath0-group with a coarse action on @xmath4 . assume that @xmath4 is @xmath5-finitistic and that the @xmath5-action is tame . then @xmath6 is a ( mod @xmath0 ) coarse homology @xmath7-sphere , for some @xmath8 . if @xmath0 is odd , then @xmath9 is even . by theorem [ kleiner - leeb ] we may assume that @xmath5 acts by isometries on @xmath4 . since @xmath4 is @xmath5-finitistic , there is a regular @xmath5-coarsening system @xmath377 for @xmath4 and an integer @xmath498 , such that @xmath722 for all @xmath12 . let @xmath751 be a normal subgroup of index @xmath0 in @xmath5 , so that @xmath752 . by induction on the order of @xmath5 , we may assume that @xmath753 is a ( mod @xmath0 ) coarse homology @xmath291-sphere , for some @xmath754 , and @xmath755 if @xmath0 is odd . since the @xmath5-action on @xmath4 is tame , there exists an @xmath756 such that @xmath116 is coarsely equivalent to @xmath6 , and @xmath757 is coarsely equivalent to @xmath753 , for any @xmath758 ( see definition [ bdd ] ) . fix @xmath758 , and let @xmath759 . since @xmath760 is a @xmath5-invariant subspace , and the @xmath761-action on @xmath762 is coarsely ineffective , this subspace @xmath762 inherits a coarse @xmath763-action from the @xmath5-action on @xmath4 . the inductive step will be completed by applying theorem [ thm : pasmith ] , to @xmath762 , after we check that the hypotheses are satisfied . \(i ) _ the space @xmath764 is @xmath763-finitistic_. in remark [ rem : fixed covering ] , we noted that there is a cofinal subsequence @xmath765 of the regular @xmath5-coarsening system such that @xmath766 is a regular @xmath5-coarsening system for the subspace @xmath757 . since @xmath761 acts trivially on the nerves @xmath767 , we have a regular @xmath763-coarsening system for @xmath757 . \(ii ) _ the coarse @xmath763-action on @xmath764 is tame_. since @xmath768 for any @xmath769 , and the inclusion @xmath770 is coarsely dense , the inclusion @xmath771 is also coarsely dense . hence the coarse @xmath763-action on @xmath762 is tame , and @xmath772 . we now apply theorem [ thm : pasmith ] to the ( mod @xmath0 ) coarse homology @xmath291-sphere @xmath764 , with respect to the induced coarse action of @xmath752 . we have @xmath773 , which implies @xmath774 . but @xmath775 is not possible , since @xmath776 , by theorem [ thm : coarse euler ] . therefore @xmath777 so that @xmath778 is a coarse homology @xmath7-sphere , for some @xmath779 . if @xmath0 is odd , then @xmath780 implies that @xmath291 and @xmath7 are either both odd or both even , thus @xmath781 is even . since @xmath782 for @xmath758 , and @xmath783 with @xmath784 if @xmath0 odd , we are done . here @xmath4 is a proper geodesic metric space , with finite asymptotic dimension , and @xmath5 is a finite @xmath0-group acting by isometries . we can apply theorem [ regular coarsening system ] to conclude that @xmath4 is @xmath5-finitistic . now theorem a follows from theorem [ final ] . we conclude by giving a coarse version of another well - known application of classical smith theory : the group @xmath10 , for @xmath0 a prime , can not act freely on a finitistic mod @xmath0 homology @xmath2-sphere ( see bredon ( * ? ? ? * iii.8.1 ) ) . a tame @xmath5-action on a proper metric space @xmath4 is called _ semifree at the large scale _ if @xmath785 for all non - trivial subgroups @xmath786 . a coarsely effective action is called _ free at the large scale _ if there exists a compact subset @xmath787 , such that @xmath788 , for all non - trivial subgroups @xmath786 . let @xmath789 with @xmath670 , for @xmath0 prime , acting by rotations . we call this a free " action in the coarse sense because the fixed set , although non - empty , is compact . the action is free at the large scale and @xmath790 . [ thm : semifree ] the group @xmath10 , for @xmath0 a prime , can not act tamely and semifreely at the large scale on a ( mod @xmath0 ) coarse homology @xmath2-sphere @xmath4 , whenever @xmath4 is @xmath5-finitistic , and @xmath6 is a ( mod @xmath0 ) coarse homology @xmath7-sphere , for some @xmath11 . we write @xmath791 , where @xmath792 and @xmath793 are cyclic subgroups of order @xmath0 . suppose , if possible , that @xmath5 acts semifreely on @xmath4 under the assumptions given in the statement . then @xmath794 is a ( mod @xmath0 ) coarse homology @xmath7-sphere , for some @xmath11 . it will be convenient to replace @xmath4 by a coarsely equivalent space @xmath795 , where @xmath796 for some @xmath797 , @xmath117 odd . in this space , we have a product action of @xmath5 , where @xmath798 acts freely on @xmath796 via free isometric actions of @xmath792 , @xmath793 on each sphere factor . the projection map @xmath799 is a @xmath5-equivariant coarse equivalence , so @xmath800 . we first observe that @xmath801 . to compute this coarse homology group we use a regular @xmath5-coarsening system @xmath377 for @xmath4 , and form the coarsening system @xmath417 for @xmath802 , with @xmath803 for each @xmath12 . then @xmath804 to compute the right - hand side , we use the direct limit of the spectral sequences of the fibrations @xmath805 in locally finite homology with @xmath685-coefficients , with @xmath806 . we obtain a spectral sequence with @xmath807 converging to @xmath808 . since @xmath809 , the only possible differential @xmath810 can not be injective , since for example , @xmath811 , and @xmath812 . we can consider @xmath813 as the quotient of @xmath814 by the remaining @xmath793-action . by theorem [ thm : pasmith ] and formula applied to @xmath802 with the @xmath792-action , we have @xmath815 . by the same results applied to the @xmath793-action on @xmath814 , we have the inequality @xmath816 which contradicts the calculation above . the first author would like to thank erik kjr pedersen for helpful conversations about coarse geometry which led to the notion of the bounded fixed set , and to the second author s ph.d thesis topic @xcite on which this paper is based . a. borel , _ seminar on transformation groups _ , with contributions by g. bredon , e. e. floyd , d. montgomery , r. palais . annals of mathematics studies , no . 46 , princeton university press , princeton , n.j . , 1960 . m. r. bridson and a. haefliger , _ metric spaces of non - positive curvature _ , grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences ] , vol . 319 , springer - verlag , berlin , 1999 . . ghys and p. de la harpe , _ infinite groups as geometric objects ( after gromov ) _ , ergodic theory , symbolic dynamics , and hyperbolic spaces ( trieste , 1989 ) , oxford sci . publ . , oxford univ . press , new york , 1991 , pp . 299314 . n. higson and j. roe , _ on the coarse baum - connes conjecture _ , novikov conjectures , index theorems and rigidity , vol . 2 ( oberwolfach , 1993 ) , london math . lecture note ser . 227 , cambridge univ . press , cambridge , 1995 , pp . 227254 .	we define a generalization of the fixed point set , called the bounded fixed set , for a group acting by isometries on a metric space . an analogue of the p. a. smith theorem is proved for metric spaces of finite asymptotic dimension , which relates the coarse homology of the bounded fixed set to the coarse homology of the total space .
stellar mass compact binaries consisting of double neutron stars ( ns ) , double black holes ( bh ) or a mixed binary consisting of a neutron star and a black hole , are the primary targets for a direct first detection of gravitational waves ( gw ) by interferometric detectors , ligo @xcite , virgo @xcite , geo600 @xcite , and tama @xcite . under radiation reaction the orbit of a binary slowly decays , emitting a signal whose amplitude and frequency increases with time and is termed a `` chirp '' signal . while it is believed that there is a greater population of ns - ns binaries @xcite , it is the bh - bh binaries that are the strongest candidates for detection since they can be seen from a greater volume , about two orders - of - magnitude greater than ns - ns binaries @xcite . in order to detect such sources one employs the method of matched filtering @xcite . briefly , the method works as follows : firstly , one creates a set of waveforms , or templates as they are called , that depend on a number of parameters of the source and its location and orientation relative to the detector . these templates are then cross - correlated with the detector output weighted by the inverse of the noise spectral density . if a signal , whose parameters are close to one of the template waveforms , is actually present in the detector output then the cross - correlation builds up , with the dominant contribution coming from frequencies where the noise spectral density is low . thus , in the presence of a sufficiently strong signal the correlation will be much larger than the rms correlation in the absence of any signal . how large should it be before we can be confident about the presence of a signal depends on the combination of the rate of inspiral events and the false alarm probability ( see e.g. ref . @xcite for a simple estimation ) . the effectiveness of matched filtering depends on how well the phase evolution of the waveform is known . even tiny instantaneous differences , as low as one part in @xmath10 in the phase of the true signal that might be present in the detector output and the template that is used to dig it out could lead to a cumulative difference of several radians since one integrates over several hundreds to several thousands of cycles . in view of improving the signal - to - noise ratio for inspiral events there has been a world - wide effort in accurately computing the dynamics of a compact binary and the waveform it emits or to use phenomenologically defined detection template families @xcite . there have been parallel efforts on using two different approximation schemes : on the one hand the post - newtonian ( pn ) expansion of einstein s equations has been used to treat the dynamics of two bodies of comparable masses with and without spin , in orbit around each other . this approximation is applicable when the velocities involved in the system are small but there is no restriction on the ratio of the masses @xcite . on the other hand , black hole perturbation theory has been used to compute the dynamics of a test particle in orbit around a spin - less or spinning black hole . black hole perturbation theory does not make any assumptions on the velocity of the components , but is valid only in the limit when the mass of one of the bodies is much less than the other @xcite . the post - newtonian approximation is a perturbative method which expands the equations of motion , binding energy and gw flux as a power series in @xmath11 , where @xmath12 is a typical velocity in the system and @xmath13 is the speed of light . in the early stages of an inspiral , the radiation reaction time - scale @xmath14 where @xmath15 is the angular velocity and @xmath16 its time - derivative , is much greater than the orbital time - scale @xmath17 . it is during this adiabatic regime that the post - newtonian approximation works best . at present , the pn expansion for the case of comparable - masses is known to order @xmath18 @xcite and @xmath19 @xcite , for the energy and flux functions , respectively . however , at this order an arbitrary parameter exists in the expression for the flux . in order to see how well pn theory performs , we can compare two different systems . if we assume a ns - ns binary of masses ( 1.4,1.4 ) @xmath20 and a lower frequency cutoff of the detector at 40 hz , then the orbital velocity " of the binary is small , @xmath21 , ] when it enters the detector bandwidth and the two stars are still largely separated , @xmath22 . the ratio of time - scales in the most sensitive regime of the detector is in the range @xmath23 . if on the other hand we take a bh - bh binary of masses ( 10,10)@xmath20 , the orbital velocity is quite large , @xmath24 and the separation is quite small , @xmath25 upon entering the detector bandwidth . this is very close to the regime , @xmath26 , @xmath27 , where the background curvature becomes strong and the motion relativistic . once again , comparing time - scales , we obtain @xmath28 , where the final value is taken at the last stable orbit at @xmath29 hz . it is known that pn theory becomes inaccurate at an orbital separation of @xmath30 @xcite . therefore , post - newtonian approximation becomes less valid for higher mass systems in the ligo band but well describes the early stages of the inspiral of a ns - ns system visible in ligo . as previously stated , black hole perturbation theory makes no assumptions about the orbital velocity of the components , but does restrict their masses . one assumes that a test particle of mass @xmath31 is in orbit about a central bh of mass @xmath32 such that @xmath33 . assuming this restriction is satisfied we have an analytical expression for the energy . however , no analytical expression has been worked out for the gravitational wave flux emitted by such a system . using black hole perturbation theory , a series approximation was initially calculated to @xmath34 by poisson for a test particle in circular orbit around a schwarzschild black hole @xcite . the series was extended numerically to @xmath35 by cutler et al . @xcite , and then to @xmath36 by tagoshi and nakamura @xcite and confirmed analytically by tagoshi and sasaki @xcite . the most recent progress is an extension of the series to @xmath37 by tagoshi , tanaka and sasaki @xcite . for a test particle in circular orbit about a kerr black hole , the initial progress was again made by poisson @xcite . the series approximation was improved from @xmath38 to @xmath35 , and subsequently to @xmath39 by tagoshi , tanaka , shibata and sasaki @xcite . several authors @xcite have shown that the convergence of both post - newtonian approximation and black hole perturbation theory is too slow to be useful in constructing accurate templates . more recently , damour , iyer and sathyaprakash ( hereafter dis ) showed for the case of a test - mass in orbit about a schwarzschild bh , that by using properly defined energy and flux functions that have better analytical properties , combined with pad techniques , it was possible to take the existing series expansion and improve its convergence properties @xcite . the new approximation in which pad approximants of new energy and flux functions are used to derive improved templates is called p - approximant . using a fiducially defined `` exact '' waveform , it was shown that the p - approximant templates were both more _ effectual _ ( i.e. larger overlaps with the exact waveform ) and _ faithful _ ( i.e. larger overlaps with the exact waveform and lower biases in the estimation of parameters ) than the corresponding post - newtonian ( hereafter t - approximant ) templates . while in general , more templates are needed for p - approximant templates to cover the same volume of parameter space @xcite , the extra computational cost is preferred for the increased performance in p - approximants . the failure of the pn expansion to converge sufficiently quickly in the case of a test particle orbiting a schwarzschild bh @xcite does not bode well for the modelling of even a test particle inspiralling into a kerr bh . another motivation for this work is that at present there is effort to use _ effective one body _ ( eob ) waveforms @xcite to detect the inspiral and merger signals from two comparable - mass kerr bhs in gw data . as the eob waveforms are based partially on p - approximants , any development and concretization of the benefits of p - approximant templates will boost our confidence in using these improved waveforms as detection templates . in this paper we will extend the p - approximant technique to the case of a test particle orbiting a kerr black hole . the reason for focusing on test - mass systems is that we can use the exact numerical fluxes @xcite from black hole perturbation theory with which to compare our results and thereby reliably demonstrate the usefulness of the technique . we begin in sec . [ sec : waveform ] with a summary of the matched filtering and the signal - to - noise ratio achieved by the first generation of geo , ligo and virgo interferometers for spinning black hole binaries . we then go on to discuss in sec . [ sec : energyandflux ] the current state - of - the - art in our understanding of the evolution of a test particle in orbit around a kerr black hole . in particular , we shall discuss the time - evolution of the orbital energy and gravitational wave flux as a function of the spin of the central black hole at various post - newtonian orders , and the locations of the last stable and unstable circular orbits . we shall see that the post - newtonian expansion of the flux does not show a regular behaviour as we move from low to high orders in the post - newtonian expansion , becoming worse for more rapidly spinning black holes . in order to improve the convergence properties of the flux function , in sec . [ sec : fluxmodel ] we shall introduce a modified form of the flux function and its pad approximant . we shall demonstrate in sec . [ sec : results ] the improved behaviour of the pad approximant , at first graphically and then by showing that the overlaps , of the inspiral waveform based on it with the exact waveform , are close to unity . we shall use a number of different test systems in our comparison : these will range from systems with dissimilar masses , such as a ns - bh binary , to comparable - mass systems , such as a bh - bh binary . to deal with the comparable mass systems , we shall introduce in the energy and flux functions a term dependent on the symmetric mass ratio @xmath40 while not being entirely consistent , because of the fact that no finite mass correction terms are included , it allows us to examine the performance of various templates as we move from the test - mass systems to comparable - mass systems . the emphasis of the current paper is also on the estimation of parameters . we have carried out a detailed study of how good t- and p - approximants are in measuring the parameters of the binary . we shall show in sec . [ sec : results ] that as a result of not being _ effectual _ representations of the exact signal , t - approximants also turn out not to be _ faithful _ representations either . in other words , the systematic error in the estimation of the parameters caused by the wrong phasing of the signal is much larger in the case of t - approximants than in the case of p - approximants . in summary , p - approximants are not only _ effectual _ but they are _ faithful _ too as in the case of non - spinning black hole binaries . in this section we will discuss the nature of the post - newtonian waveform from an inpiralling compact binary and the response of an antenna to such a signal . we will then use the fourier transform of the waveform to compute the signal - to - noise ratio ( snr ) that can be achieved for these signals when they are detected using matched filtering . in the transverse - traceless gauge gravitational waves are represented by two polarisation amplitudes @xmath41 and @xmath42 the response @xmath43 of an antenna to an incoming signal is expressed as a combination of the two polarisation states and the beam pattern functions @xmath44 and @xmath45 of the antenna as @xcite : @xmath46 for a wave from a binary of masses @xmath47 and @xmath48 ( total mass @xmath49 and mass ratio @xmath50 ) that is inclined with respect to the plane of the sky at an angle @xmath51 , propagating in the direction @xmath52 ( see ref . @xcite for exact definitions ) , frequency @xmath53 and polarization angle @xmath54 with respect to the antenna , the polarisation amplitudes , in the so - called _ restricted _ post - newtonian approximation @xcite , are @xmath55 with @xmath56 a velocity parameter , and the beam - pattern functions are @xmath57 using the above expressions for the beam - pattern functions and gravitational wave amplitudes the response takes the form @xmath58,\ ] ] where the amplitude coefficient @xmath59 and phase @xmath60 can be assumed to be constant for signals lasting for a short duration ( say , less than about 30 mins ) : @xmath61 . \label{kt8}\end{aligned}\ ] ] thus , gravitational wave antennas are not able to extract the two polarisations separately and the data analysis problem boils down to matching the time - varying phase @xmath62 and to a lesser extent the amplitude @xmath63 of the antenna response function . post - newtonian theory and the quadrupole formula applied to a binary give the relativistic binding energy @xmath64 per unit mass , and the flux of the waves @xmath65 as series expansions in the parameter @xmath66 once we have the binding energy and the flux we can use the energy balance argument , namely that the flux of gravitational waves is completely balanced by the negative rate of change of the binding energy [ @xmath67 , in order to arrive at a parametrized equation for the evolution of the phase @xmath68 of gravitational waves . integrating the energy balance equation supplemented by @xmath69 one obtains : @xmath70 where @xmath71 @xmath72 is a reference time and @xmath73 is a reference phase of the signal , when the velocity parameter is @xmath74 the numerical integration of the above equations is more economical when the following differential equations are solved instead , @xmath75 the parametric representation , eqs . ( [ eq : timevsv ] ) and ( [ eq : phivsv ] ) , of the phasing formula @xmath76 holds under the assumption of ` adiabatic inspiral ' , _ i.e. _ , that gravitational radiation damping can be treated as an adiabatic perturbation of a circular motion . however , the effective one - body approach @xcite has allowed a treatment of the radiation damping to proceed beyond the adiabatic approximation . in order to extract an inspiral signal that may be buried in noisy data by the method of matched filtering , we need to employ post - newtonian accurate representations for the two functions @xmath77 and @xmath65 that appear in the above phasing formulas . given an approximant @xmath78 , @xmath79 , one can define by replacing @xmath80 , @xmath81 in eqs . ( [ eq : timevsv ] ) and ( [ eq : phivsv ] ) some approximate parametric representation , @xmath82 , @xmath83 , and therefore a corresponding approximate template @xmath84 obtained by replacing @xmath85 , in the following @xmath85-parametric representation of the waveform @xmath86 by the function of time @xmath87 obtained by inverting @xmath82 . there are several ways of performing this inversion which leads to the different t - approximants , _ taylort1 _ [ keep the rational polynomial @xmath88 in eqs . ( [ eq : timevsv ] ) and ( [ eq : phivsv ] ) as is and solve the integrals numerically ] , _ taylort2 _ [ re - expand the rational polynomial as a post - newtonian series , truncate terms to the appropriate order and solve the integrals analytically to get series expansions , @xmath89 and @xmath90 but solve numerically for @xmath68 from the parametric equations ] and _ taylort3 _ [ do as in taylort2 but also invert the series expansions to obtain the phasing as an explicit function of time : @xmath91 , as discussed in ref . @xcite . it is often convenient to deal with the fourier representation of the waveform . in the stationary phase approximation the fourier transform , defined as @xmath92 , for positive frequencies reads @xcite @xmath93 } , \label{d4.6a}\ ] ] and , since @xmath43 is real , @xmath94 here @xmath95 is the stationary point of the phase @xmath96 ( i.e. , @xmath95 is defined by @xmath97 ) , @xmath98 @xmath99 is the time - derivative of the instantaneous gravitational wave frequency evaluated at the stationary point given by , @xmath100 and @xmath101 is the phase of the fourier transform in the stationary phase approximation given by @xmath102 instead of solving the integrals in the above equation it is numerically efficient to solve the following equivalent differential equations for the phasing formula in the fourier domain : @xmath103 and in this study we have used the above differential equations in computing the waveforms . for the energy we have used the exact analytical formula ( see next section ) ; for the flux we have used the exact numerical flux to define the exact waveform and the perturbative expansions , or their re - summed improved versions , to define the approximate waveforms . in order to estimate the signal - to - noise ratio ( snr ) we shall assume that the signal is detected using matched filtering and that the bank of templates used in matched filtering has a waveform that matches the unknown signal very closely . to compute the optimum snr we need to known the noise spectral density of the detectors . for initial ligo , the one - sided noise power spectral density ( psd ) from the design study @xcite is given by @xcite @xmath104\ { \rm hz}^{-1},\ ] ] where @xmath105 and @xmath106hz is the `` knee - frequency '' of the detector . we take the psd to be infinite below the lower frequency cutoff of @xmath107hz . for the virgo detector the psd is given by @xmath108\ { \rm hz}^{-1},\ ] ] where @xmath109hz and @xmath110hz . finally , for the geo detector the expected noise psd is @xmath111\ { \rm hz}^{-1}\ ] ] where @xmath112 hz and the noise is assumed to be infinite below a lower frequency cutoff of @xmath107 hz . next , let us define the scalar product of two waveforms @xmath113 and @xmath114 by @xmath115 , \label{eq : scalarprod}\ ] ] where the * denotes complex conjugate and @xmath116 are the fourier transforms of @xmath117 . the ( square ) of the signal - to - noise ratio obtained by matched filtering a signal @xmath113 with a template @xmath114 is given by @xmath118 if the template perfectly matches the signal then the snr is simply @xmath119 for the inspiral signal given in eq . ( [ d4.6a ] ) , and using the _ newtonian values _ for the energy and flux functions , namely @xmath120 @xmath121 and @xmath122 the snr reduces to : @xmath123^{1/2},\ \\ ] ] here @xmath124 is the frequency where the data analyst chooses to terminate the signal . the cutoff frequency is usually chosen to be the frequency of the waves at the last stable orbit of the binary ( cf . sec . [ subsec : energy ] ) which depends on the total mass of the two stars and their spin magnitude . note that the snrs depend only on the combination @xmath125 the chirp mass and not on individual masses of the system . for a given total mass the snr at a fixed distance ( alternatively the span of the antenna at a fixed snr ) is the highest for binaries of equal masses ( when @xmath126 ) and reduces by the fraction @xmath127 for binaries of unequal masses . the snr decreases inversely as the distance to the source and depends not only on the intrinsic parameters @xmath128 @xmath129 and @xmath130 but also on the relative orientations of the source and the antenna specified by the angles @xmath131 @xmath132 @xmath133 @xmath134 we can compute the snr for a source of rms orientation by averaging the amplitude factor @xmath59 over all the angles . performing the average over the angles leads to : @xmath135 for an ideally oriented source , that is for a source that produces the highest snr , @xmath136 thus , the rms and ideal snrs , obtained by using @xmath137 and @xmath138 respectively , are given by because here we use one - sided psd as opposed to two - sided psd used in ref . @xmath139^{1/2},\ \ \rho_{\rm ideal } = \frac{5}{2 } \rho_{\rm rms}. \label{final snr}\ ] ] in fig . [ fig : snr ] we have plotted the _ rms _ snr for the ligo , virgo and geo detectors , as a function of the spin magnitude and the total mass of the system . clearly , systems with large spins in which the test mass is in prograde orbit produce the highest snrs since their last stable orbit frequency is higher than those of retrograde orbits and they last longer . to capture such signals , however , one must know the phasing of the waves very well since the bodies spend a substantial fraction of their time in the detector band in the strongly non - linear regime of the evolution . we shall show later in this paper that there could be a significant drop in the snr , and a corresponding drop in the number of candidate events that can be detected , while p - approximants lose little snr and therefore capture almost all events . we can see from eqs . ( [ eq : timevsv ] ) and ( [ eq : phivsv ] ) that the phase of the gravitational wave depends both on the energy and flux functions of the binary system . in the test - mass case , that is when one of the masses is far smaller than the other ( @xmath140 ) so that the dynamics is governed entirely by the kerr geometry of the massive body , we have an exact expression for the energy @xmath141 , but only a series representation for the flux @xmath142 . in the comparable - mass case , wherein one can not neglect the perturbation caused by the companion , both functions are represented only by post - newtonian expansion . the standard approximants for @xmath143 and @xmath144 are simply their successive taylor approximants @xmath145 and @xmath146 respectively . the gravitational wave signal constructed using these taylor approximants are called t - approximant signals or waveforms are simply t - approximants . it was shown in ref @xcite that the convergence properties of both the energy and flux functions can be improved by using pad approximation of well - defined energy and flux functions . damour , iyer and sathyaprakash followed a two - pronged approach to construct improved energy and flux functions @xcite : starting from the more basic energy - type and flux - type functions , @xmath147 and @xmath148 ref . @xcite constructs pad - type approximants , say @xmath149 , @xmath150 , of the `` basic '' functions @xmath147 , @xmath151 one can then compute the required energy and flux functions entering the phasing formula . the successive approximants @xmath152 $ ] and @xmath153 $ ] have better convergence properties than their taylor counterparts @xmath154 $ ] and @xmath155 $ ] . in this study we will restrict ourselves to the test mass approximation and , therefore , employ the exact energy function . however , in the same approximation the flux function has been computed analytically only as a taylor series although numerically the flux has been computed exactly . since our aim is to draw conclusions on how effective p - approximants are in the comparable mass case , wherein one has only a taylor expansion of the flux , we construct p - approximants of flux and compare it with the numerical results . the reasons for constructing a new flux function are the following : firstly , it is well known , in the schwarzschild case , that a simple pole exists at @xmath156 @xcite in the expression for the flux . we might , therefore , expect that a similar pole exists in the kerr case . a taylor approximant of the flux function will never produce a pole while the equivalent pad approximant , which is essentially a rational polynomial , will have a pole . although it is well known that pad approximants have the ability to model functions which are subject to singularities @xcite it is not guaranteed that the rational polynomial will reproduce the exact pole . secondly , it is often possible to recover from the _ divergence _ of a taylor series by using pad approximation . again , one can not be certain that the resulting pad approximant , even when it converges , will be closer to being exact than any of the taylor approximants , but experience with the test mass approximation in the schwarzschild case render some optimism to this expectation . we create a pad approximation of a truncated power series of an analytic function containing @xmath157 terms by equating it to a rational function such that the rational function when expanded as a power series and truncated to order @xmath157 coincides with the original power series . more precisely , pad approximation can be thought of as an operator @xmath158 that acts on a polynomial @xmath159 to define a rational function : @xmath160{\frac{\displaystyle # 1}{\displaystyle # 2 } } p^{n}_{m } \left ( \sum_{k=0}^{n } a_{k}v^{k } \right ) = \dfrac{\sum_{k=0}^{n } a_{k}\,v^{k}}{\\ 1+\sum_{k=1}^{m } b_{k}\,v^{k } } , \label{eq : pade}\ ] ] such that the number @xmath161 of coefficients in the rational polynomial on the right hand side is the same as the number @xmath162 of taylor coefficients on the left hand side . by setting @xmath163 with @xmath164 we can define two types of pad approximants : these are the super - diagonal , @xmath165 , and sub - diagonal , @xmath166 , approximants . normally , the sub - diagonal approximants are preferred over super - diagonal approximants . this is because when @xmath167 the right hand side of eq . ( [ eq : pade ] ) can be re - expanded as a continued fraction which have the property that as we go to each new order of the power series only one new coefficient needs to be calculated . conversely , with the super - diagonal approximants , we would have to re - calculate all the @xmath168 s and @xmath169 s in the above equation as we go to higher orders in the taylor expansion . this means that the sub - diagonal pad approximants are more _ stable _ and if we see a trend of convergence in the coefficients the addition of a term is not likely to spoil this convergence . we refer the reader to appendix [ sec : padeapp ] for a more summary of the properties of pad approximations and how to find the diagonal pad coefficients using continued fraction expansion . in the case of both schwarzschild and kerr black holes we have an exact expression for the orbital energy of a test particle in a circular orbit around the parent black hole . for a black hole of mass @xmath32 the energy @xmath170 in terms of the dimensionless magnitude of velocity @xmath171 @xmath172 being is the radial coordinate in the boyer - lindquist coordinates , takes the form @xcite @xmath173 where @xmath0 is a dimensionless spin parameter given in terms of the spin angular momentum @xmath174 of the black hole by @xmath175 with @xmath176 spin angular momentum per unit mass in the kerr metric . it is actually the derivative of the orbital energy that appears in the phasing formula given by : @xmath177 where a prime denotes a derivative with respect to the velocity parameter @xmath12 . for black holes with spin , the positions of the last stable circular orbit is a function of the spin parameter @xmath0 . in order to find the position of the last stable circular orbit we take @xmath178 which gives @xcite : @xmath179\,\left [ 3 + z_{1}(q ) + 2\,z_{2}(q ) \right]}\right],\ ] ] where @xmath180\\ \nonumber \\ z_{2}(q ) & = & \sqrt{3\,q^2 + z_{1}^{2}}.\end{aligned}\ ] ] the @xmath181 @xmath182 sign on @xmath183 corresponds to prograde ( retrograde ) orbits . now the position of the last unstable circular orbit or `` photon ring '' occurs where the energy function exhibits a pole . thus , the photon orbit is found by solving a cubic equation . of the three poles , the physically relevant pole is given by @xcite @xmath184\right],\ ] ] where once again the different signs define a prograde ( retrograde ) orbit . in the limit of @xmath185 , both @xmath183 and @xmath186 move towards the horizon of the black hole , @xmath187 in the maximally rotating case of @xmath188 , we can not distinguish between the last stable orbit and the light ring . this means , the greater the spin of the bh , the closer the particle can approach to it before beginning the plunge phase . on the other hand , in the limit @xmath189 , the position of @xmath183 moves outwards from the bh . thus the particle begins its plunge much earlier than in the prograde case . finally , for a particle in an orbit about a schwarzschild black hole the above equations give the familiar position of @xmath190 for the lso and @xmath191 for the photon ring . as discussed in the introduction it has not been possible to derive an exact analytical expression for the flux of gravitational waves emitted by a binary system although analytic approximate expressions , and exact numerical results , have been computed . in the interesting case of two comparable masses in orbit around each other post - newtonian methods have been used to derive an expression for the flux to quite a high order in the expansion parameter . however , since post - newtonian theory is known to be poorly convergent , especially when the expansion parameter approaches unity , it has been suggested to employ in its place an equivalent rational polynomial , or pad approximation , to the ( modified ) flux function . since the purpose of this paper is to test the effectiveness of such an approximation we need a firm ground on which we can conduct our test . in the case of a test mass in an equatorial orbit around a kerr black hole numerical methods have been used to compute the flux to all post - newtonian orders . this is , of course , valid only in the limit of vanishingly small mass of the test body as compared to the central object . however , all the relativistic corrections , including hereditary effects , such as the back scattering of gravitational waves off the curved background geometry , would be present in this computation . the deformation of the geometry due to the presence of the second body , or the back reaction of the waves on the motion of the test body , which would in turn affect the emission process , will not have been included in such a calculation . the deformation of the geometry , parametrized by the symmetric mass ratio of the system @xmath192 is important only a few orbits before the two objects merge . this should be expected from the fact that both in newtonian and einsteinian gravity it has been possible to construct an effective one - body formalism to describe the dynamics of a binary and the dynamics derived within this formalism is expected to be valid close to the point when the two bodies plunge towards each other . therefore , we expect that the deformation of the system will not bring about a major change in the analytic behaviour of the flux and lessons learnt in the test mass approximation will be applicable , albeit qualitatively , in the comparable mass case . thus , our strategy is to employ the test mass exact flux computed numerically together with the analytic expression for the energy function discussed in the previous section . these two exact functions , together with the energy balance equation , can be used to construct an exact phasing formula for the inspiral signal . we shall also define an approximate phasing formula using the corresponding post - newtonian expansions of the two functions . our approximations at each post - newtonian order will be one of two types : ( 1 ) the standard post - newtonian or ( 2 ) the improved pad approximation . in the next section we will discuss how these approximations can be further improved and in the last section we will measure the quantitative performance of the two approximations . for a test - particle in a circular equatorial orbit the post - newtonian expansion of the flux function has been calculated up to @xmath193 ( @xmath194 ) in the case of a schwarzschild bh @xcite , and to @xmath193(@xmath195 ) in the case of a kerr bh @xcite . the general form of the flux function in both these cases is given by the expression @xmath196 . \label{eq : flux}\ ] ] where the expansion is known to order @xmath197 and 11 , in the kerr and schwarzschild cases , respectively , @xmath0 is the dimensionless spin parameter introduced earlier and @xmath198 is the dominant _ newtonian _ flux function given by @xmath199 here , @xmath200 is the magnitude of the invariant velocity parameter observed at infinity which is related to the angular frequency @xmath201 by @xmath202 . the relation between the parameter @xmath200 and the local linear speed @xmath12 in boyer - lindquist coordinates is given by @xmath203^{1/3},\ ] ] which reduces , in the schwarzschild limit , to @xmath204 note that this local linear velocity is related to the boyer - lindquist radial coordinate @xmath172 by @xmath205 the spin - dependent coefficients , @xmath206 , and the log - term coefficients , @xmath207 , are given by has the wrong sign in some later references but ref . @xcite has the correct sign . ] , @xmath208 where @xmath209 is euler s number . for graphical purposes , it is more appropriate to deal with the newton - normalized fluxes defined by @xmath210 we will now outline the method of calculation the p - approximant of the gw flux as proposed by dis @xcite . we notice from the form of the series expansion for the flux , eq . ( [ eq : flux ] ) , that we begin to encounter logarithmic terms at order @xmath211 and above . in general , series approximations of this form have slow convergence properties . in order to prepare the series representation of the flux for creating the pad approximation , it is convenient if we factor out the logarithmic terms . we can then write eq . ( [ eq : flux ] ) as @xmath212 \left[\sum_{k=0}^{n}\,c_{_{k}}x^{k}\right],\ ] ] where the new coefficients @xmath213 and @xmath214 are functions of the old coefficients @xmath215 and @xmath216 as in ref . @xcite the log - terms have been `` normalized '' using the value of the velocity parameter at the lso ; this helps in reducing the importance of the log - terms . factoring out the logarithmic terms aids in constructing the rational polynomial , or the pad approximation , of the reminder . moreover , since we expect the flux to have a pole at the location of the light ring it is best to factor out the expected pole so that the reminder has good analytical properties . to this end we create the factored flux function , @xmath217 by the operation @xmath218 factoring out the pole also helps to alleviate the problem arising from the absence of the linear term in the pn expansion of the flux . ( note that @xmath219 in both the schwarzschild and kerr cases . ) we can see from appendix [ sec : padeapp ] that in the absence of such a term the continued fraction form of the pad approximation , the so - called diagonal pad approximant , would lead either to zero or infinite pad coefficients . the above operation rectifies this problem by introducing a linear term into the taylor series for the flux . if we write the expression in full we obtain @xmath220 \left[\sum_{k=0}^{n}\,f_{_{k}}x^{k}\right ] , \label{eq : fluxf}\ ] ] where @xmath221 and @xmath222 @xmath223 . we can now construct a new flux function by using the pad approximant of the factored flux given above . indeed , we can construct two variants of the new flux : the first one is what we call the _ direct _ or d - pad approximant , which is obtained by directly starting from the flux function @xmath224 in eq . ( [ eq : fluxf ] ) and constructing the equivalent rational polynomial . this is the approach followed in dis . an alternative approach to this is motivated by the fact that in the gravitational wave phasing formula the flux appears in the denominator . thus , instead of constructing the pad approximant of the flux function one could first construct the polynomial expansion of the inverse of the flux function and construct the pad approximant of the resulting polynomial . we call the approximant constructed this way as _ inverse- _ or i - pad approximant because it is obtained from the taylor expansion of the _ inverse _ of the flux function @xmath224 in eq . ( [ eq : fluxf ] ) . thus , our two improved versions of the flux are defined as follows : the direct pad approximant is defined by @xmath225p^ { m}_{m+\epsilon}\left[\sum_{k=0}^{n}\,f_{_{k}}x^{k}\right],\ ] ] where @xmath226 is the diagonal or sub - diagonal pad approximant , @xmath227 with @xmath228 or @xmath229 depending on whether @xmath157 is even or odd . the inverse pad approximant of flux is defined by @xmath230p^ { m}_{m+\epsilon}\left[\sum_{k=0}^{n}\,d_{_{k}}x^{k}\right],\ ] ] where the coefficients @xmath231 in the taylor expansion are defined by @xmath232 one finds @xmath231 by first expanding the rhs in a binomial series and then identifying the coefficients of the various terms with those on the lhs . having constructed the rational polynomials equivalent to a given truncated taylor series of the modified flux function we can return to the original flux function that appears in the phasing formula . we shall call the flux function so constructed as _ p - approximant , _ and not just pad approximant to remind ourselves that the new flux has been obtained by improving the convergence properties in two steps ( i.e. , definition of a new flux function and the construction of its rational polynomial ) and not just a direct application of pad approximation . thus , we define the _ direct p - approximant _ of flux as @xmath233 and _ inverse p - approximant _ as , @xmath234^{-1}.\ ] ] detailed investigation shows that inverse p - approximants of flux , namely @xmath235 have better convergence properties than direct p - approximants . we shall therefore use only @xmath236 in all our investigations in the rest of this paper . in this section we will look at the convergence properties of the post - newtonian series representations and the improved p - approximants of the flux function discussed in the last section . to this end we shall employ all the post - newtonian orders up to which the taylor expansions have been worked out in the kerr case . we shall investigate the convergence of the approximants in the case of a test particle ( i.e. , @xmath237 limit ) either co - rotating ( i.e. , @xmath238 _ prograde orbits _ ) or counter - rotating ( i.e. , @xmath239 _ retrograde orbits _ ) , with respect to a central black hole whose spin parameter @xmath0 takes nine values over the full range from @xmath240 in this section we shall make a graphical comparison of the analytical and numerical fluxes and assess the performance of the flux with respect to variation in the spin - parameter values at increasingly higher post - newtonian orders . we will find that neither the post - newtonian series nor their improved versions fit the numerical flux exactly when the parameters are chosen to be the same for analytical fluxes as the numerical flux . however , the p - approximant fluxes agree with the numerical fluxes quite closely when their spin values are slightly mis - matched with the true values of the numerical fluxes . the post - newtonian approximants do show this improved behaviour , albeit not to the extent of p - approximants . thus , we can expect the p - approximants to define better signal models than post - newtonian approximants with regard to both faithfulness and effectualness . to @xmath241 for a test particle in prograde orbit around a kerr black hole . the solid lines represent the exact numerical fluxes with thicker lines corresponding to larger spin magnitudes , @xmath242 respectively . the correspondingly thickened dashed lines represent the taylor- and p - approximants.,title="fig:",width=288 ] to @xmath241 for a test particle in prograde orbit around a kerr black hole . the solid lines represent the exact numerical fluxes with thicker lines corresponding to larger spin magnitudes , @xmath242 respectively . the correspondingly thickened dashed lines represent the taylor- and p - approximants.,title="fig:",width=288 ] the numerical fluxes have been computed exactly @xcite , but in the test mass approximation , at a set of test values of the spin parameter @xmath244 for prograde orbits the flux has been computed at @xmath245 . for the sake of completeness , we shall also include in our comparison the numerical flux for a particle in orbit about a schwarzschild bh , i.e. @xmath246 @xcite . in fig . [ fig : ktp ] we have plotted the newton - normalized numerical fluxes alongside analytical fluxes for both the truncated taylor approximants ( four panels on the left ) and ( inverse ) p - approximants ( four panels on the right ) . for each approximant , we consider four post - newtonian orders at \{@xmath247 } and five spin values . the solid lines with increasing thickness correspond to the exact numerical fluxes at progressively larger values of the spin parameter starting at @xmath248 ( thinnest line ) and ending at @xmath249 ( thickest line ) . the corresponding dashed lines represent the analytical approximant to the gw flux . let us first concentrate on the post - newtonian fluxes : we notice immediately that at all values of @xmath0 , and large values of @xmath250 the 2.5 post - newtonian approximant @xmath251 is highly divergent . indeed , this order has been known to exhibit improper behaviour in the schwarzschild case and continues to be so for prograde orbits in kerr . the situation improves at @xmath252 but begins to worsen again at @xmath253 becoming highly divergent at @xmath254 . the plots serve to illustrate the principal problem with the pn expansion : going to higher orders of the approximation does not necessarily correspond to a better accuracy . the sequence of panels on the right - hand - side of fig . [ fig : ktp ] demonstrates that the p - approximant flux has improved convergence properties over the post - newtonian flux for the same range of post - newtonian orders and spins . the extreme divergence which the post - newtonian flux exhibited at @xmath251 and @xmath255 has been cured . let us note , however , that sometimes the rational polynomial approximation of a taylor series can produce spurious poles in the region of interest . an example of this can be seen at @xmath256 which shows a spurious pole at every value of @xmath244 the main factor that stands out in the right panels of fig . [ fig : ktp ] is the following : the p - approximant fluxes display the same characteristic shape as the numerical fluxes although they do nt agree with the exact flux at higher values of the parameter @xmath257 to @xmath241 for a test particle in retrograde orbit around a kerr black hole . the solid lines represent the exact numerical fluxes with thicker lines corresponding to larger spin magnitudes , @xmath258 respectively . the correspondingly thickened dashed lines represent the taylor- and p - approximants.,title="fig:",width=288 ] to @xmath241 for a test particle in retrograde orbit around a kerr black hole . the solid lines represent the exact numerical fluxes with thicker lines corresponding to larger spin magnitudes , @xmath258 respectively . the correspondingly thickened dashed lines represent the taylor- and p - approximants.,title="fig:",width=288 ] in the case of prograde orbits the p - approximants graphically display a closer convergence to the numerical fluxes : while being better than the taylor approximation , they are still not as good as we would like . however , we must remember that what is more important is how the phasing of the waveforms defined by the two approximants match with that defined by the exact flux , rather than a simple graphical agreement of the flux . the graphical display is useful as it gives us a rough idea on any improvement in performance of a particular model . but as the noise of the interferometer weights the inner - product between templates , it may emphasize a particular range of frequency of the waveform more than the other . it may turn out that the divergent region of a particular model is outside the detector bandwidth for most masses . in such cases the approximant may work well in spite of the divergences or poles . we shall return to this question in sec . [ sec : results ] when we consider the matching of waveforms . we now focus on the gw flux from a test - mass counter - rotating with respect to the black hole . we will again use a similar set of spin values , i.e. @xmath260 as well as the schwarzschild flux for comparative purposes . in fig . [ fig : ktr ] , left panels , we plot the taylor approximant flux for retrograde orbits . the solid lines in the figure correspond to the exact numerical fluxes with thicker lines corresponding to smaller spin values @xmath261 and @xmath262 respectively . the correspondingly thickened dashed lines represent the taylor series approximation to the gw flux with the same spin magnitude as the numerical flux . the right panels are the same except that in place of taylor approximant fluxes we plot the inverse p - approximant fluxes . the first thing we notice is that just as in the case of prograde motion the @xmath251 and @xmath255 approximants are again highly divergent although at @xmath255 , the convergence of the pn approximation improves as we go to larger negative spins . this is not entirely surprising : as we go to larger counter - rotating spins the higher order taylor expansion coefficients become smaller and the non - linear general relativistic corrections are not as important as in the case of prograde orbits . moreover , the position of the lso moves outwards from the bh as @xmath0 becomes more negative . at @xmath263 , the lso is at @xmath264 . this corresponds to @xmath265 at @xmath266 the lso has moved outwards to @xmath267 , corresponding to @xmath268 . if we compare this to the equivalent spins in the prograde case we find that for @xmath269 , @xmath270 and @xmath271 . at @xmath249 , @xmath272 and @xmath273 . in other words , retrograde orbits are weakly bounded orbits in comparison to prograde orbits and experience smaller relativistic corrections . recalling that gravitational wave luminosity @xmath274 retrograde flux for @xmath266 at lso is smaller by a factor @xmath275 compared to prograde flux for @xmath249 at lso . consequently , the pn approximation seems to work well in the retrograde case . in the case of retrograde orbits p - approximants provide , at order @xmath276 only a marginal improvement over t - approximants . this is in sharp contrast to the prograde case where the improvement was excellent . at orders @xmath277 and @xmath255 are again quite good , with no poles at @xmath278 for most spin values . while there seems to be a less obvious benefit in using the p - approximant flux in the retrograde case , it is again important to note that the p - approximants show the same general behaviour as the exact flux in the retrograde case too and can therefore be expected to match the true general relativistic signal more closely . in the last two sections we noted how the p - approximant sequence seems to have a character similar to the sequence given by the numerical computations although the two do not agree perfectly . this suggests that if the value of the spin parameter used in the t- or p - approximant fluxes is varied relative to the true value of the spin parameter in the exact flux , we should be able to obtain a better fit to the numerical fluxes . while we will initially do this graphically , we remind the reader that this exercise is purely demonstrative . as the fluxes are newton - normalized , it is difficult to say in what region of the velocity we should attempt to make the best fit . added to this is the fact that when matching templates we should take the weighting of the psd into account . however , if the approximant fluxes agree graphically well with the numerical fluxes then it would mean that we should be able to match the t- or p - approximant flux with the numerical result without introducing any new parameters . to examine how best mis - matching spins might work we will focus on one of the post - newtonian orders , say @xmath252 the order at which the t - approximants work well and p - approximants have no spurious poles . in fig . [ fig : fitt ] , left panel , we have plotted the best - fit for the t - approximant flux for various values of the spin magnitude from @xmath279 ( top most curve ) to @xmath280 ( bottom most curve ) . for each spin magnitude we have fitted the t - approximant with the numerical result as best as possible by varying the spin - magnitude of the approximant . in best fitting the numerical flux we incurred errors ( defined as @xmath281 ) in the spin magnitude of the approximant relative to the exact case . these are listed in table [ table : bestfit ] . although the errors decrease as we approach the extreme prograde flux , we can see from the figure that the fit is not very good for the taylor approximant for @xmath282 schwarzschild case . .the `` best - fit '' value of the spin parameter , @xmath0 , for t and p - approximant fluxes for @xmath279 to @xmath280 . for the schwarzschild case we have an absolute error of 0.11 in the case of t - approximant and @xmath283 in the case of p - approximant . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] [ table : schwarzoverlaps ] we have applied p - approximant templates to the case of a ns orbiting kerr bhs of varying mass and spin . when graphically matching the numerical fluxes we found that the p - approximant fluxes gave a better fit as compared to the pn fluxes , especially when we varied the value of the spin parameter and best fitted the exact flux . however , the true test of how well a template performs is reflected by the fitting factors achieved . using a signal waveform constructed from the exact expression for the orbital energy and numerical fluxes from black hole perturbation theory , we were able to compare the performance of t and p - approximant templates . in the case of retrograde , schwarzschild and prograde orbits , not only did the p - approximants gave better and more reliable fitting factors , they also gave smaller biases in the estimation of parameters . we also saw the true power of the p - approximant templates in that we were able to generate templates right up to the lso . this is something that was not possible with the t - approximant templates due to the approximation for the flux function becoming negative before the lso is reached . this restricted just how much of the numerical waveform we could model . while not being completely correct due to the fact that we omitted the finite - mass correction terms , we also saw that the p - approximant templates gave the best performance when in the equal - mass case . the bias in the estimation of parameters is particularly poor in the case of t - approximants especially for high - mass systems . since high - mass systems can be seen to a greater distance , and therefore likely to be the first sources to be detected , it is important to keep in mind that the bias in the estimation of parameters could be as large as the random errors associated with the measurement . such biases could seriously affect the test of general relativity @xcite that are envisaged to be carried out with the aid of high - mass systems . it is clear that for the type of systems examined in this paper , namely equatorial test - mass circular orbits in kerr , p - approximant templates are to be preferred over their pn counterparts in both detection and measurement . it remains to be seen if p - approximants ( or their improved versions ) prove to be good enough when considering generic orbits in kerr and for comparable mass systems with spinning black holes on generic orbits . we would like to thank prof . m. shibata for his permission to use the numerical fluxes in this work . we would also like to thank t. damour , b. iyer and the members of the cardiff relativity group for the stimulating conversations and helpful suggestions . ekp would like to thank the school of physics and astronomy , cardiff , for its hospitality in the final stages of this work . this research was supported in parts by the particle physics and astronomy research council and the leverhulme trust , both in the uk . 99 a. abromovici et al . , _ science _ * 256 , * 325 ( 1992 ) c. bradaschia et al , _ nucl . instrum . methods phys . _ sec . a , 518 ( 1990 ) j. hough and k. danzman , _ proposal for a @xmath284 laser interferometric gravitational wave antenna _ , ( 1994 ) k. kuroda et al in _ gravitational waves : sources and detectors _ , ed . by i. cinfolini and f. fidecaro , 100 ( world scientific , 1997 ) l. p. grishchuk , v. m. lipunov , k. a. postnov , m. e. prokhorov and b. s. sathyaprakash , _ phys . _ * 44 * , 1 ( 2001 ) e. phinney , _ astrophys . j. _ * 380 * , l17 ( 1991 ) r. narayan , t. piran and a. shemi , _ astrophys . j. _ * 379 * , l17 ( 1991 ) i. h. stairs et al . , _ astrophys . j. _ * 505 * , 352 ( 1998 ) k. belczynski , v. kalogera and t. bulik , astrophys . j. , * 572 * 407 ( 2001 ) . astro - ph/0111452 . k. a. postnov and m. e. prokhorov , _ binary black hole formation and mergings , _ in proceedings of the xxxiv rencontres de moriond `` gravitational waves and experimental gravity '' , january 23 - 30 , 1999 , astro - ph/9903193 . c. w. helstrom , _ statistical theory of signal detection _ ( pergamon press , london 1968 ) s. v. dhurandhar and b. s. sathyaprakash , _ phys . rev . _ d * 49 * , 1707 ( 1994 ) a. buonanno , y. chen , and m. vallisneri , phys . d * 67 , * 024016 ( 2003 ) . a. buonanno , y. chen , and m. vallisneri , phys . d * 67 , * 104025 ( 2003 ) . y. pan , a. buonanno , y. chen , and m. vallisneri , _ a physical template family for gravitational waves from precessing binaries of spinning compact objects : application to single - spin binaries , _ gr - qc/0310034 l. blanchet , t. damour , b. iyer , c. m. will and a. g. wiseman , _ phys . rev . lett . _ * 74 * , 3515 ( 1995 ) l. blanchet , t. damour and b. iyer , _ phys _ d * 51 * , 5360 ( 1995 ) c. m. will and a. g. wiseman , _ phys . _ d * 54 * , 4813 ( 1996 ) l. blanchet , b. iyer , c. m. will and a. g. wiseman , _ class . quantum grav . _ * 13 * , 575 ( 1996 ) l. blanchet , _ phys _ d * 54 * , 1417 ( 1996 ) t. damour , p. jaranowski and g. schfer , _ phys . _ d * 63 * , 044021 ( 2001 ) ; v. c. de andrade , l. blanchet and g. faye , _ class . quantum grav . _ * 18 * , 753 , ( 2001 ) . l. blanchet , g. faye , b. r. iyer and b. joquet , _ phys . _ d * 65 * , 061501 ( 2002 ) ; l. blanchet , b. r. iyer and b. joquet , _ phys . _ d * 65 * , 064005 ( 2002 ) e. poisson , _ phys . _ d * 47 * , 1497 ( 1993 ) c. cutler , l. s. finn , e. poisson and g. j. sussman , _ phys . _ d * 47 * , 1511 ( 1993 ) h. tagoshi and t. nakamura , _ phys . _ d * 49 * , 4016 ( 1994 ) m. sasaki , _ prog . _ * 92 * , 17 ( 1994 ) h. tagoshi and m. sasaki , _ prog . theor . phys . _ * 92 * , 745 ( 1994 ) t. tanaka , h. tagoshi and m. sasaki , _ prog . _ 1087 ( 1996 ) p. r. brady , j. d. e. creighton and k. s. thorne , _ phys . rev . _ d * 58 * , ( 1998 ) e. poisson , _ phys . _ d * 49 * , 1860 ( 1993 ) m. shibata , m. sasaki , h. tagoshi and t. tanaka , _ phys . _ d * 51 * , 1646 ( 1995 ) h. tagoshi , m. shibata , t. tanaka and m. sasaki , _ phys . _ d * 54 * , 1439 ( 1996 ) c. cutler et al . , _ phys . _ * 70 * , 2984 ( 1993 ) e. poisson , _ phys _ d * 52 * , 5719 ( 1995 ) e. poisson , _ phys . _ d * 55 * , 7980 ( 1997 ) t. damour , b. r. iyer and b. s. sathyaprakash , _ phys . _ d * 57 * , 885 ( 1998 ) e. k. porter , _ class . quantum grav . _ * 19 * , 4343 , ( 2002 ) y. mino , m. sasaki , m. shibata , h. tagoshi and t. tanaka prog . theor . suppl . * 128 * 1 ( 1997 ) . a. buonanno and t. damour , _ phys . _ d * 59 * , 084006 ( 1999 ) t. damour , _ phys . rev . _ d * 64 * , 124013 ( 2001 ) t. damour , b. r. iyer and b. s. sathyaprakash , _ phys . _ d * 63 * , 044023 ( 2001 ) t. damour , b. r. iyer , p. jaranowski and b. s. sathyaprakash , _ phys . rev . _ d * 67 * , 064028 ( 2003 ) k. s. thorne , _ gravitational radiation _ , in s. w. hawking and w. israel ( eds ) , _ 300 years of gravitation _ ( cambridge uni . press , cambridge ) b. s. sathyaprakash and s. v. dhurandhar , _ phys . _ d * 44 * , 3819 ( 1991 ) s. v. dhurandhar , j. watkins and b. f. schutz , _ fourier transform of a coalescing binary signal _ ( unpublished preprint , 1989 ) t. damour , b. r. iyer and b. s. sathyaprakash , _ phys . _ d * 62 * , 084036 ( 2000 ) b. s. sathyaprakash , _ phys . _ d * 50 * , r7111 ( 1994 ) w. h. press , b. p. flannery , s. a. teukolsky and w. t. vetterling , _ numerical recipes in c : the art of scientific computing _ ( cambridge uni . press , cambridge 1992 ) 2nd edition . c. m. bender and s. a. orszag , _ advanced mathematical methods for scientists and engineers _ , mcgraw - hill international book company , singapore , 1978 . j. m. bardeen , w. h. press and s. a. teukolsky , _ astrophys . j. _ * 178 * , 374 ( 1972 ) s. chandrasekhar , _ the mathematical theory of black holes _ ( oxford uni . press , new york , 1983 ) m. shibata ( private communication ) c. cutler and e. flanagan , _ phys . _ * d 49 * 2658 ( 1994 ) gr - qc/9402014 . r. balasubramanian , b. s. sathyaprakash , and s. v. dhurandhar , _ phys . _ d * 53 * , 3033 ( 1996 ) ; b. j. owen , phys . rev . d * 53 * , 6749 ( 1996 ) ; b. j. owen and b. s. sathyaprakash , _ phys . _ d * 60 * , 022002 ( 1999 ) . b. f. schutz in _ the detection of gravitational radiation _ , ed . by d. blair ( cambridge uni . press , cambridge 1989 ) l. blanchet and b. s. sathyaprakash , _ phys . _ * 74 * , 1067 ( 1995 ) . let @xmath285 be an analytic function whose taylor expansion in @xmath250 denoted @xmath286 is @xmath287 we construct the pad approximation @xmath288 of @xmath289 by equating the polynomial expansion with a rational function to the same order according to @xmath290 where @xmath160{\frac{\displaystyle # 1}{\displaystyle # 2 } } \frac{b_{n}(x)}{d_{m}(x ) } = \dfrac{\sum_{n=0}^{n } b_{n}\,x^{n}}{\\ \sum_{n=0}^{m } d_{n}\,x^{n } } , \label{eq : ratfunc2}\ ] ] such that @xmath291 . the benefit of using pad approximants over taylor approximants is that the error in the approximation after @xmath292 terms reduces by a factor of @xmath293 @xcite . if we let @xmath294 , where @xmath228 or @xmath295 , we construct the `` diagonal '' and `` sub - diagonal '' pad approximants respectively . the advantage of this is that we can cast the rational function in the form of a continued fraction , i.e. @xmath296 if we use the rational function interpretation , eq . [ eq : ratfunc2 ] , each time we go to a higher order @xmath157 of approximation , we have to recalculate each of the coefficients , @xmath297 and @xmath298 . however , by using the continued fraction formalism , we only have to calculate the newest coefficient , @xmath299 . as an example , consider the taylor expansion @xmath300 by matching this to a pad approximation of the same order we get @xmath301 solving for the pad coefficients , @xmath299 , gives us @xmath302 if we now go to the next order , i.e. @xmath303 and solve for the coefficients @xmath299 once more , we obtain @xmath304 the example shows that as we go to higher and higher orders in @xmath12 , we only have to calculate one new coefficient as all the lower order coefficients remain constant . consequently , there is a sense of stability in the behaviour of the continued fraction coefficients and convergence is easier to test . there are , unfortunately , problems with pad approximation that one should bear in mind : firstly , there are regions where pad approximants diverge while taylor approximants converge . the second , and perhaps most important feature is that when pad approximation does indeed converge , we can not be sure that it converges to the exact function that we had in mind @xcite . it should therefore always be kept in mind that there are inherent difficulties in using pad approximants . finally , pad approximants sometimes exhibit spurious poles in the region of interest . however , as one will be able to test _ a priori _ the existence of poles one can avoid the use of such approximants .	the standard post - newtonian approximation to gravitational waveforms , called t - approximants , from non - spinning black hole binaries are known not to be sufficiently accurate close to the last stable orbit of the system . a new approximation , called p - approximants , is believed to improve the accuracy of the waveforms rendering them applicable up to the last stable orbit . in this study we apply p - approximants to the case of a test - particle in equatorial orbit around a kerr black hole parameterized by a spin parameter @xmath0 that takes values between @xmath1 and @xmath2 in order to assess the performance of the two approximants we measure their _ effectualness _ ( i.e. larger overlaps with the exact signal ) , and _ faithfulness _ ( i.e. smaller biases while measuring the parameters of the signal ) with the exact ( numerical ) waveforms . we find that in the case of prograde orbits , that is orbits whose angular momentum is in the same sense as the spin angular momentum of the black hole , t - approximant templates obtain an effectualness of @xmath3 for spins @xmath4 for @xmath5 the effectualness drops to about 0.82 . the p - approximants achieve effectualness of @xmath6 for all spins up to @xmath7 the bias in the estimation of parameters is much lower in the case of p - approximants than t - approximants . we find that p - approximants are both effectual and faithful and should be more effective than t - approximants as a detection template family when @xmath8 for @xmath9 both t- and p - approximants perform equally well so that either of them could be used as a detection template family .
the way in which a galaxy has assembled its stars is reflected in the distribution of stellar orbits . for example , collisionless @xmath0-body collapse simulations predict a predominance of radial orbits in the final remnant @xcite . in contrast , collisionless galaxy merger simulations predict a variety of orbital compositions , depending on progenitor properties @xcite , the merging geometry @xcite , the progenitor mass ratios @xcite and the presence of dissipational components @xcite . stars in galaxies are approximately collisionless and the orbital structure once a galaxy has approached a quasi - steady - state is conserved for a long time . to a certain extent then , the assembly mechanism of early - type galaxies can be constrained from their present - day orbital structure . a global characteristic of the distribution of stellar orbits is its anisotropy . traditionally , anisotropies of elliptical galaxies have been inferred from the ( @xmath1 ) diagram . in particular , the rotation of bright ellipticals has been shown to be insufficient to account for their flattening @xcite . however , whether fainter , fast - rotating ellipticals are flattened by rotation is less easy to determine from the ( @xmath1 ) diagram , because isotropic as well as anisotropic systems can rotate . in fact , fully general axisymmetric dynamical models recently have revealed an anisotropic orbital structure in even the flattest , fast rotating objects @xcite . one goal of this paper is to investigate numerically the connection between anisotropy , rotation and flattening in spheroidal stellar systems . in addition , we present global anisotropies for a sample of coma early - type galaxies . these anisotropies are derived by analysing long - slit stellar absorption line kinematics with axisymmetric orbit models . our dynamical models include dark matter halos . previous anisotropy determinations for larger samples of ellipticals ( including dark matter ) were restricted to round and non - rotating systems , assuming spherical symmetry @xcite . spherical models do not account for galaxy flattening . in the simplest case , a flattened system is axially symmetric . early axisymmetric models , however , did not cover all possible anisotropies ( and orbital structures , respectively ; e.g. @xcite ) . fully general , orbit - based axisymmetric dynamical models have so far only been applied to the inner regions of ellipticals and the orbital analysis was made under the assumption that mass follows light ( e.g. @xcite and @xcite ) . by the mass - anisotropy degeneracy , the neglect of dark matter could translate to a systematic bias in the corresponding orbital structure ( e.g. @xcite ) . comparison of anisotropies derived with and without dark matter will allow one to quantify such a possible bias . we also discuss anisotropies derived from modelling mock observations of synthetic @xmath0-body merger remnants . one motivation to do so is that dynamical models of axisymmetric systems may not be unique . for example , the deprojection of an axisymmetric galaxy is intrinsically degenerate @xcite . uncertainties in the intrinsic shape thereby propagate into uncertainties on the derived masses and anisotropies ( e.g. @xcite ) . moreover , the reconstruction of an axisymmetric orbital system is suspected to be further degenerate with the recovered mass ( e.g. the discussion in @xcite ) . the case for a generic degeneracy , beyond the effects of noise and incompleteness of the data , is still uncertain ( e.g. @xcite ) . numerical studies of a few idealised axisymmetric toy models indicate degeneracies to be moderate when modelling realistically noisy data sets ( @xcite , @xcite ) . since we know the true structure of our @xmath0-body modelling targets , we can extend on these studies and further investigate potential systematics in the models over a broader sample of test cases . another motivation to model @xmath0-body merger remnants is to probe whether ellipticals have formed by merging . this requires a comparison of the orbital structure in real ellipticals with predictions of @xmath0-body simulations ( e.g. @xcite ) . however , because of the symmetry assumptions in models of real galaxies , it is not straight forward to compare _ intrinsic _ properties of @xmath0-body simulations with _ models _ of real galaxies . to avoid the related systematics , we here compare models of real galaxies with _ similar models _ of synthetic @xmath0-body merger simulations and both are indicative for true differences between real galaxies and merger predictions . the galaxy and @xmath0-body merger samples and the modelling technique are briefly outlined in sec . [ sec : data ] . toy models of various flattening and anisotropy are discussed in sec . [ sec : theory ] . the anisotropies of real galaxies are presented in sec . [ sec : galaxies ] and compared with models of @xmath0-body merger remnants in sec . [ sec : merger ] . implications for the formation process of early - type galaxies are discussed in sec . [ sec : discussion ] and we summarise our results in sec . [ sec : summary ] . the influence of regularisation and the inclusion of dark matter halos on reconstructed galaxy anisotropies is discussed in app . [ sec : app ] . in app . [ app : entropy ] we briefly discuss the connection between anisotropy and the shape of the circular velocity curve in maximum entropy models . we assume that the coma cluster is at a distance of 100 mpc . .summary of coma galaxy anisotropies . ( 1 - 2 ) : galaxy i d ( gmp numbers from @xcite ) ; ( 3 ) : intrinsic ellipticity @xmath2 ; ( 4 - 6 ) : anisotropy parameters @xmath3 , @xmath4 and @xmath5 ( cf . equations [ eq : delta]-[eq : gamma ] ) of the best - fit dynamical model ; ( 7 ) : @xmath6 , i.e. @xmath7 normalised by the approximate value @xmath8 of an ( edge - on ) isotropic rotator with the same flattening . note that @xmath6 is an observable , i.e. it combines _ observed _ ellipticities @xmath9 ( from column 10 of tab . 1 of @xcite ) and observed velocities @xmath10 and @xmath11 , without reference to any dynamical model . [ cols="<,^,^,^,^,^,^,^ " , ] a complete description of a stellar system is given by its distribution function @xmath12 ( df ; the density in 6-dimensional phase - space ) . in a steady - state system the df @xmath12 depends on the phase - space coordinates only through the integrals of motion @xcite . axisymmetric potentials , which are considered here , admit the two classical integrals of motion energy ( @xmath13 ) and z - component of the angular momentum ( @xmath14 ) . in addition , many orbits in astrophysically relevant potentials are characterised by another , non - classical , so - called third integral ( @xmath15 ; @xcite ) . since integrals of motion label orbits and vice - versa , a steady - state system can be viewed as a superposition of orbits , each with constant phase - space density . let @xmath16 denote the phase - space density along orbit @xmath17 , then the total amount of light @xmath18 on the orbit equals @xmath19 ( @xmath20 is the orbit s phase - space volume ) . the df or the weights of a suitable orbit superposition model determine the spatial density @xmath21 and intrinsic velocity dispersions @xmath22 via @xmath23 and @xmath24 @xmath25 in the following we will only consider @xmath26 , where @xmath27 is the short axis of the density distribution , @xmath28 is the azimuth around this axis , @xmath29 is a fixed cartesian coordinate parallel to the equatorial plane and @xmath30 is a cylindrical radius . let @xmath31 denote the total ( unordered ) kinetic energy in coordinate direction @xmath17 , then the global anisotropy of an axisymmetric stellar system can be quantified , for example , by the ratios @xmath32 @xmath33 and @xmath34 @xcite . in axisymmetric systems , the three anisotropy parameters are related via @xmath35 non - rotating , isotropic spherical systems as well as classical isotropic rotators obey @xmath36 . the df of real galaxies is not known , but has to be reconstructed from photometric and kinematic observations . in the next two subsections we will describe the two samples of real and simulated galaxies discussed in this paper and will briefly outline our modelling method . our sample of observed galaxies ( coma in the following ) consists of 19 coma early - types from @xcite . it comprises 2 central cd galaxies , 10 ordinary giant ellipticals and 7 s0 or intermediate galaxies with luminosities @xmath37 ( a single fainter object with @xmath38 is also included in the sample ) . the galaxies are drawn from the luminosity limited sample of @xcite and are distributed all over the cluster . high - resolution radial profiles of surface brightness , ellipticity and isophotal shape parameters @xmath39 and @xmath40 ( up to @xmath41 in some cases ; cf . @xcite for a definition of the isophotal shape parameters ) derived from a combination of hst and ground - based imaging were used to calculate the deprojected 3d luminosity distribution for several inclinations . the photometric data are complemented by long - slit stellar absorption line kinematics along @xmath42 position - angles per galaxy . the kinematic data consists of radial profiles of mean velocity , velocity dispersion and higher - order moments of the line - of - sight velocity distribution and reach out to @xmath43 . details about the photometric and kinematic data can be found in @xcite , @xcite , @xcite , @xcite and @xcite . these data were modelled with our implementation of schwarzschild s ( 1979 ) orbit superposition technique for axisymmetric potentials @xcite . for each galaxy , we probed for a variety of mass models , composed of a stellar mass density ( from the deprojected light profile ) and a parametric dark halo profile . the parameter space for the mass models spans the inclination , the stellar mass - to - light ratio and the dark halo parameters . in each trial potential the best - fit orbit model is calculated by maximising @xmath44 where @xmath45 quantifies deviations between observed and modelled kinematics - body merger remnants . ] . the function @xmath46 is used to smooth the orbit models . in the absence of any other constraints the maximisation of s yields orbital weights @xmath47 @xcite , such that the yet not specified @xmath48 can be regarded as weight - factors for the @xmath18 . when modelling real galaxies or mock observations of @xmath0-body merger remnants , we assume that there is no preferred region is phase - space and each orbit is given an a priori - weight equal to its phase - space volume : @xmath49 . then , @xmath50 equals the boltzmann entropy , which drives models towards a constant density in phase - space . the ( binned ) deprojected luminosity density is used as a boundary condition to solve equation ( [ orbitmaxs ] ) and the regularisation parameter @xmath51 in equation ( [ orbitmaxs ] ) has been calibrated by means of monte - carlo simulations @xcite . the final , best - fit orbit model is obtained from a @xmath45-analysis . we have applied the same modelling code to mock observations of synthetic @xmath0-body merger simulations . in brief , we have modelled six merger remnants , each projected along its three principal axes ( models of projections along the long , intermediate and short axis of the merger remnants will be shortly referred to as x , y and z - models later on ) . the six merger remnants are taken from the sample of collisionless disk+bulge+halo mergers of @xcite . they have mass ratios between 1:1 and 4:1 and sample the entire distribution of intrinsic shapes and orbital make - ups , including extreme cases . an orbital analysis of the @xmath0-body systems is given in @xcite . we have simulated typical coma observations for each projection : the merger remnants were placed at a distance of 100 mpc and photometric and kinematic profiles with similar resolution and spatial coverage as in the coma sample have been extracted . for a detailed discussion of the models the reader is referred to @xcite . oblate stellar systems can owe their shapes to a variety of different orbital configurations . classically , one has often distinguished between two proto - typical cases : flattening by rotation and flattening by anisotropy . thereby , flattening by rotation is used to term an otherwise round and isotropic system which appears flattened and rotating by extra - light on near - equatorial , high angular - momentum orbits ( populated with the same sense of rotation ) . flattening by anisotropy refers to systems with a depression of stars with high velocities perpendicular to the equatorial plane ( @xmath52 ) . however , in fact there are infinitely many orbit superpositions that account for a given galaxy shape . some of these are discussed in @xcite . different orbital structures can be distinguished by their different anisotropies . in the following we will numerically construct ( self - consistent ) toy models that are designed to ( 1 ) reproduce a given , flattened , density distribution exactly , but ( 2 ) have different intrinsic anisotropies . the models are orbit - based and similar to those described in sec . [ sec : data ] . however , here we only require the models to reproduce a given density ( @xmath53 in equation [ orbitmaxs ] ) . various expressions for the factors @xmath48 in equation ( [ sgen ] ) will be used to impose different anisotropy structures ( see below ) . for our simple toy models we assume a stellar density @xmath54 @xcite with @xmath55 equations ( [ eq : dehnen ] ) and ( [ eq : flattening ] ) describe systems with constant flattening @xmath56 . they approximate the light profiles of elliptical galaxies reasonably well . [ [ flattening - and - maximum - entropy . ] ] flattening and maximum entropy . let @xmath57 denote the df that maximises the entropy of equation ( [ maxs ] ) subject to the density constraints . the squares in the top panel of fig . [ aniso_prof ] illustrate the connection between anisotropy and flattening for @xmath57 : the three panels show the anisotropy parameters from equations ( [ eq : delta]-[eq : gamma ] ) as a function of the intrinsic ellipticity @xmath58 ( cf . equation [ eq : flattening ] ) . while @xmath3 and @xmath4 increase with flattening , @xmath5 is roughly constant . in maximum - entropy models the flattening thus arises from a suppression of energy in @xmath27-direction , while the balance between the energies in @xmath30 and @xmath28 is roughly conserved . in this sense , the maximum entropy models @xmath57 resemble the classical case of flattening by anisotropy . the only difference is that @xmath59 ( cf . [ app : entropy ] for a discussion of @xmath5 ) . note that we calculated the toy models with the same library setup as used for the coma galaxy models . [ [ flattening - by - a - classical - ffel_z . ] ] flattening by a classical @xmath60 . a classical two - integral df @xmath61 , which only depends on @xmath13 and @xmath14 , can be approximated via equations ( [ orbitmaxs],[sgen ] ) with @xmath62 @xmath63 and @xmath64 ( @xmath0 is the total number of orbits ) . equation ( [ newv ] ) derives from the constraint that for @xmath65 , the phase - space density of any orbit @xmath17 with energy @xmath66 and angular momentum @xmath67 has to equal the mean phase - space density of all orbits @xmath68 with the same @xmath69 and @xmath70 , i.e. @xmath71 this case is included in fig . [ aniso_prof ] by the circles . that the @xmath48 from equation ( [ newv ] ) indeed yield @xmath72 is demonstrated by @xmath73 . the flattening of the corresponding systems comes from an excess energy in @xmath28-direction with respect to the isotropic case ( @xmath74 ; orbits with high angular momentum are strongly populated ) . the relationship between @xmath3 and @xmath2 is similar as in maximum entropy models . note that dfs @xmath72 develop noticeable phase - space density peaks on orbits with high angular momentum @xcite . it is likely this property that lowers their entropy as compared to the @xmath57 models . flattening by anisotropy mainly involves shell orbits which approach closely the intrinsic minor - axis . their phase - space volumes are much larger than those of equatorial near - circular orbits with high angular momentum . even a small change in the phase - space density along shell orbits can reduce the amount of light near the minor - axis considerably and , thus , result in a significant flattening . the larger fraction of phase - space involved in this type of flattening , compared with a strong overpopulation of the relatively small region in phase - space occupied by near - circular orbits ( as in cases where @xmath75 ) explains why objects which are flattened by anisotropy have the higher entropy . [ [ flattening - with - radial - anisotropy . ] ] flattening with radial anisotropy . model dfs @xmath76 obtained with @xmath77 ^ 4 \times v(i)\ ] ] are biased towards orbits with a large difference @xmath78 between apocentre and pericentre radius ( radially extended orbits ) . such models are radially anisotropic ( @xmath79 ; cf . triangles in fig . [ aniso_prof ] ) . the relationship between @xmath3 and @xmath2 is again similar as in the previous models . the latter is no surprise , as for self - consistent ellipsoids with constant flattening , @xmath80 can be calculated from the tensor virial theorem @xcite : @xmath81 where @xmath82 and @xmath83 the solid line in the upper - left panel of fig . [ aniso_prof ] shows relation ( [ eq : deltavir ] ) . our numerically constructed orbit models follow this line well . note that , if @xmath0 dfs @xmath84 project each to the same spatial density , then any convex linear combination @xmath85 with @xmath86 will do so . the properties of @xmath87 will be intermediate between those of the individual @xmath84 . the just discussed toy models ( and any linear combination of them ) are non - rotating , because in our choices for @xmath48 we havent distinguished between prograde and retrograde orbits . a large variety of rotation patterns can be constructed from any df @xmath12 as follows : each orbit in an axisymmetric potential comes in two flavours , one prograde ( with positive @xmath88 ) and one retrograde ( @xmath89 ) . both share the same spatial shape but differ only in the sign of the velocity component around the axis of symmetry . thus , the spatial density will only depend on the sum @xmath90 of light on corresponding prograde and retrograde orbits . the amount of rotation , instead , will depend on the difference between the population of the prograde and retrograde orbits , respectively . this can be quantified , for example , by the fraction @xmath91 of light on the prograde of each orbit pair : @xmath92 ( @xmath93 ) . for simplicity , let s assume from now on that @xmath91 is the same for all orbits . then , any @xmath94 with @xmath95 $ ] ( to remain positive definite ) will give rise to the same density profile as @xmath12 ( @xmath96 ) , but with different degrees of internal rotation . for example , in case of @xmath97 prograde and retrograde orbits are populated equally and there will be no rotation in the corresponding system . with @xmath98 ( @xmath99 ) only prograde ( retrograde ) orbits are populated ( maximum rotation ) . the bottom panel of fig . [ aniso_prof ] shows anisotropies for the toy models of sec . [ subsec : toy ] with @xmath100 . while @xmath4 is independent of the amount of rotation , @xmath3 decreases and @xmath5 increases with increasing rotation . the latter reflects that in our toy models the total energy in @xmath28-direction is constant . any increase of the ordered motion is thus to the expense of a smaller @xmath101 . [ eps_vsig ] illustrates where the toy models appear in the @xmath102 diagram . the figure shows the three cases @xmath103 ( no rotation ) , @xmath104 ( intermediate rotation ) and @xmath105 ( maximum rotation ) . on the y - axis , the ratio @xmath106 of the maximum rotation velocity ( @xmath10 , along the projected major - axis ) and the central velocity dispersion ( @xmath107 , averaged inside @xmath108 ) is shown . all models are edge - on . the highest rotation rates at a given flattening are obtained for @xmath61 , because of its strongly populated high angular momentum orbits ( @xmath109 ) . however , @xmath57 models , which are not flattened by an excess of light on high angular momentum orbits ( relative to the isotropic case ) but instead by a suppression of orbits with large @xmath27-velocities ( @xmath110 ) can reach @xmath111 as well . the dashed lines in fig . [ eps_vsig ] approximate classical isotropic rotators by @xmath112 @xcite . up to @xmath113 , @xmath57 models can appear in the same region as classical isotropic rotators , although they are not flattened by rotation in the classical sense ( e.g. @xmath114 ) . radially anisotropic models are dominated by orbits with low angular momentum and have generally low rotation rates . a complete picture of an axisymmetric galaxy s flattening mechanism requires knowledge of the amount of rotation ( e.g. @xmath115 ) and at least one anisotropy parameter ( e.g. @xmath4 , @xmath5 or @xmath3 or the parameter @xmath51 in the notation of @xcite ) . alternatively , two anisotropy parameters also specify the global orbital structure . in any case , the full information about the anisotropy and the flattening mechanism can not be provided by the ( @xmath1 ) diagram alone . for example , four among the five intrinsically most flattened coma early - types are very close to the isotropic rotator line in fig . [ eps_vsig ] . however , they are shaped by a combination of @xmath116 and @xmath117 ( cf . tab . [ tab : aniso ] ) . the presence of dark matter around a galaxy affects the shape of the stellar orbits . some of the models just discussed may not exist , if an additional dark matter halo reshapes the potential significantly . to check this , we have recalculated all our toy models in a potential , where a spherical , logarithmic dark halo has been added to the stellar potential . the parameters of the halo ( its core radius and its asymptotic circular velocity ) have been set according to the dark matter scaling relations in coma early - types @xcite . the derived anisotropies in the new potential differ in no case by more than @xmath118 from the original ones ( but see the discussion in app . [ app : entropy ] ) . especially , the relationship between @xmath4 and @xmath2 , that arises from the maximisation of the orbital entropy also appears in potentials with a realistic dark matter halo . this does not necessarily imply that the neglect of dark matter in models of real galaxies has no effect on the derived anisotropies , because it may enforce a redistribution of the orbits ( cf . next sec . [ sec : galaxies ] ) . fig . [ aniso_eps ] shows the connection between anisotropy and flattening in real galaxies . the intrinsic flattening of coma galaxies is expressed in terms of @xmath119 @xcite . here , @xmath30 is the radius along the projected major - axis and @xmath120 and @xmath121 are the surface - brightness profile and ellipticity profile in the edge - on projection . for an axisymmetric system ( with flattening q ) @xmath122 . lines in fig . [ aniso_eps ] trace three different toy models @xmath123 ( cf . [ sec : theory ] ; the three models are designed to rotate by using @xmath124 in equation [ flammu ] ) . dfs @xmath72 are inconsistent with the global orbital structure of most galaxies ( because @xmath125 in observed galaxies ) . most galaxies have orbital properties between those of @xmath57 and @xmath61 ( with some rotation ) . [ aniso_eps ] also includes anisotropies and flattenings of 24 early - types from @xcite . these galaxies are a subsample of the 48 es / s0s of the sauron survey @xcite , which uniformly covers the plane of observed flattening @xmath9 and @xmath126 ( for @xmath127 ) . the galaxies of @xcite are drawn from this survey according to various requirements , among them consistency with axial symmetry ( according to 2d kinematical maps ) . the galaxies of @xcite ( shortly sauron in the following ) are on average fainter than the coma galaxies . although the samples do not match exactly , the anisotropies of coma and sauron galaxies are found in the same range . however , the coma sample contains relatively more anisotropic but nearly round galaxies on the one hand and more highly flattened but isotropic galaxies ( @xmath128 ) on the other . as a result , the trend for @xmath3 and @xmath4 to increase with @xmath2 which is seen in the sauron sample is not obvious when considering the complete coma sample ( even not if the two coma galaxies with the most uncertain anisotropies are ignored the two central galaxies gmp2921 and gmp3329 ) . the relation between @xmath4 and @xmath2 is weaker in the coma galaxies in part due to a few round but anisotropic galaxies for example gmp1750 and gmp5568 with @xmath129 and @xmath130 . both galaxies show weak minor - axis rotation @xcite and could be slightly triaxial systems . in addition to differences among nearly round galaxies , anisotropies in coma and sauron galaxies also slightly differ at high @xmath2 . the latter is most clearly seen in @xmath3 versus @xmath2 : two highly flattened coma galaxies ( gmp1990 and gmp2440 , @xmath131 ) have @xmath128 . one of these galaxies is likely close to edge - on , because of its high observed ellipticity ( @xmath132 , cf . the radial profile in @xcite ) and its significant isophotal shape distortions . we expect the model of gmp1990 to be well constrained , because of the near edge - on inclination ( minimal uncertainties in the deprojection ) and its far - extending multi - slit kinematic data . for the other galaxy ( gmp2440 ) @xcite quote only a modest observed ellipticity @xmath133 at @xmath134 and the intrinsic flattening comes mostly from the low inclination of the model . note that this galaxy is far above the isotropic rotator line in the right panel of fig . [ eps_vsig ] ( gmp2440 is the only non edge - on galaxy above the isotropic rotator line ) . a maximum - entropy like df is ruled out for this galaxy , because even the maximally rotating version of the @xmath57 model would not allow for the high observed rotation rate . thus , even if we would have underestimated the inclination of this system , fig . [ eps_vsig ] shows that its orbital structure must be significantly deviant from maximum - entropy models . all in all then , modelling uncertainties are unlikely to explain the outstanding anisotropies of gmp1990 and gmp2440 . in fact , a comparison with fig . 3 in @xcite reveals that the sauron sample does not include galaxies like gmp1990 and gmp2440 , because ( 1 ) for only one object the observed ellipticity is significantly larger than @xmath135 ( ngc4550 ) and ( 2 ) even the fastest rotators in the sauron sample are closer to the isotropic rotator line than gmp2440 . in addition to differences in the sample selection also the modelling methods differ in the details . @xcite use similar orbit - based dynamical models as we do here , but sauron anisotropies are calculated inside a fixed aperture with a radius of @xmath136 . a fixed aperture encloses different fractions of the stellar mass in different galaxies , depending on system size and distance . for the coma galaxies we give anisotropies inside @xmath134 . in some galaxies local anisotropies vary significantly with radius @xcite , such that the radius of comparison is crucial . in addition , sauron models are based on the assumption that mass follows light . as it has been stated already in the introduction , the assumption of a constant mass - to - light ratio can result in artificially large @xmath28-energies ( @xcite , @xcite ) or low @xmath5 , respectively . regarding fig . [ aniso_eps ] , sauron galaxies do not have systematically lower @xmath5 than coma objects . for the only two exceptions ( ngc4473 and ngc4550 ) @xcite report evidence for counter - rotating , disk - like components that likely cause their large @xmath28-energies . the small effect that the neglect of dark matter has on the anisotropies likely reflects the fact that we only consider anisotropies averaged inside @xmath137 , where the assumption that mass follows light is most closely fulfilled ( e.g. @xcite , @xcite ) . for the coma galaxies a quantitative comparison of models with and without dark matter is made in app . [ sec : app ] . the spatial coverage with kinematic data in the inner regions is sparse in the coma galaxies ( long - slit data ) compared to the sauron objects ( 2d kinematical maps ) . in regions of phase - space that are not well constrained by the observed kinematics , the dynamical models are mainly driven by regularisation . thus , because the spatial coverage is lower in the coma galaxies , their anisotropies could be biased towards the adopted regularisation scheme . specifically , coma galaxy models are regularised towards maximum entropy @xcite . however , the middle panel in the top row of fig . [ aniso_eps ] does not show any bias of the coma models towards the maximum entropy relation . in fact , sauron galaxies are on average closer to this relation than coma galaxies . this indicates that regularisation is not the main driver for the coma galaxy models . also , in app . [ sec : app ] we give an explicit comparison of coma galaxy models with standard and with weak regularisation . we do not find significant differences . both the intrinsic ellipticity and the anisotropy depend on the inclination of the models . for the coma galaxies , we probe three different inclinations and use the one that fits best @xcite , while inclinations for sauron galaxies are derived from two - integral jeans models @xcite . the inclination is best constrained for highly flattened galaxies , because these have to be close to edge - on . for three of the coma galaxies ( gmp0756 , gmp1176 and gmp1990 ) large ellipticities together with significantly discy / boxy isophotes indeed indicate close to edge - on inclinations ( for example , gmp1176 exhibits @xmath138 ; @xcite ) . in contrast , two among the five galaxies with @xmath139 owe their flattening in part from the relatively low inclination of the best - fit model ( gmp0282 , gmp2440 ; cf . tab . [ tab : aniso ] ) . these galaxies provide the smallest and largest anisotropies , respectively , at high @xmath2 ( cf . middle panel in the top row of fig . [ aniso_eps ] ) . this suggests that the method to determine the inclination for the coma galaxies does not result in a specific bias of the derived anisotropies . we conclude that slight differences between the sauron and the coma anisotropies are mostly due to the different sample selections , while differences in the modelling methodology ( including differences in the data coverage ) seem to be negligible the lower panels of fig . [ aniso_eps ] display the models of @xmath0-body merger remnants ( cf . [ sec : data ] ) . in terms of @xmath3 versus @xmath2 and @xmath4 versus @xmath2 , these models do not differ strongly from models of real galaxies ( see also @xcite ) . however , while @xmath140 in models of merger remnants , @xmath5 is often negative in models of real galaxies . is this discrepancy in @xmath5 indicative for the merger remnants having a different orbital structure than real galaxies , or does it merely reflect systematics caused by the symmetry assumptions in our models ? reconstructed and true intrinsic anisotropies in axisymmetric systems , @xmath141 in the merger remnants . for the intrinsic @xmath3 of the merger remnants we use the average @xmath142 instead of @xmath143 in equation ( [ eq : delta ] ) . ] and flattenings of the merger remnants are compared in fig . [ aniso_recon ] . the one merger remnant closest to oblate axial symmetry ( oblate ) , is reconstructed with high accuracy from the x and y - projections ( edge - on ) . this is plausible , because for this remnant the assumption of axial symmetry is a good approximation . furthermore , in the edge - on case the deprojection becomes unique and the intrinsic degeneracies in the dynamics are likely smallest . however , the general trend in the axisymmetric models is to underestimate both , the flattening and the anisotropy of the merger remnants . x and y - projections allow a better reconstruction of shape and anisotropy than z - projections . it has already been discussed in @xcite that the assumption of axial symmetry enforces an inclination mismatch in the z - models : while the triaxial remnants appear flattened in the @xmath144 projection ( face - on ) , axisymmetric systems are necessarily round when seen face - on . then , because the models are forced towards a wrong viewing - angle , ( 1 ) the intrinsic flattening is underestimated and ( 2 ) x , y and z - axes of models and remnants do no longer correspond to each other . for example , a z - model s @xmath5 measures a different energy ratio as the remnant s @xmath5 @xcite . had we compared the z - models with the apparent shape of the remnant in z - projection and with the energy ratios along axes of models and remnants that correspond to each other , then the differences would have been much smaller ( for example @xmath145 ) . real galaxies are seen at random viewing angles . starting from our models of principal projections it is difficult to predict directly the analogous distributions of @xmath80 or @xmath146 for the realistic case of random projections . however , because the projections along principal axes yield extreme kinematical and photometrical properties of the merger remnants @xcite , it can be expected that dynamical models of projections along intermediate viewing - angles will have properties intermediate between those of the models from principal projections . we have verified this for two out of the six merger remnants ( elong and oblate ) by modelling additional 11 projections ( at intermediate viewing - angles ) . assuming that this result can be generalised to other remnants as well , then figs . [ aniso_eps ] and [ aniso_recon ] suggest the following : if real galaxies would resemble the modelled merger remnants , then one would see approximately the same relationships @xmath80 and @xmath146 as in the coma and sauron galaxies . however , it is clear from fig . [ aniso_recon ] that @xmath140 , for a sample of randomly projected objects like our modelled @xmath0-body merger remnants . thus , in this respect , models of many real galaxies differ from our comparison sample of synthetic @xmath0-body merger remnants : models of merger remnants are always radially anisotropic ( @xmath147 ) , while models of real galaxies are characterised by @xmath148 . this fact is further illustrated in fig . [ aniso_aniso ] , which shows correlations among the anisotropy parameters . even though shape and anisotropy can not be recovered simultaneously ( in some cases ) , the anisotropy correlations in the models of the merger remnants and in the merger remnants itself are very similar to each other . again , the main difference between real galaxies and merger models is the offset between their @xmath5 distributions . besides the fact that merger remnants have on average positive @xmath149 , while real galaxies have @xmath150 ( on average ) , fig . [ aniso_aniso ] shows that the distribution of anisotropies in the merger remnants is tighter than in real galaxies . this may reflect the similarity in the initial conditions of the @xmath0-body simulations ( most noteworthy the similarity in the progenitors and the fact that we only consider collisionless mergers ) . the anisotropy parameters defined in sec . [ sec : theory ] are only global measures of the orbital structure . a full understanding of the formation process of early - type galaxies can only be provided by spatially resolved anisotropy profiles . for example , equatorial near circular orbits obey , in the epicycle approximation , the local relation @xmath151 where @xmath152 is the circular velocity @xcite . in a typical galaxy potential the circular velocity curve is flat ( @xmath153 ) and equation ( [ eq : epicyc ] ) predicts @xmath154 . since the epicycle approximation holds for perturbed rotating disks , we do not expect the majority of early - types in our sample to be well described by equation ( [ eq : epicyc ] ) . however , it might be relevant for the most flattened , rotating and discy objects in our sample . instead , at least some of these ( for example gmp1176 and gmp3958 ) have negative @xmath5 ( i.e. @xmath155 ) . this does not rule out a disk heating scenario for these galaxies , however , because locally we find @xmath156 near the equatorial plane in these galaxies ( around @xmath157 ; cf . the radial anisotropy profiles in @xcite ) . in case of the collisionless @xmath0-body merger simulations , already the averaged anisotropy parameters reveal significant differences to the models of real galaxies . which physical processes are responsible for this discrepancy ? the orbital structure of the models of merger remnants is largely driven by a population of central box orbits in the @xmath0-body systems @xcite . they cause the centres of the merger remnants to become triaxial / prolate and are , for example , largely responsible for the wrong viewing angle of the z - models . dissipation during a merger can have a significant effect on the shape and the projected properties of the final remnant @xcite . already 10 percent of gas are sufficient to suppress central box orbits and to produce an approximately axisymmetric remnant in binary mergers @xcite ( but this result is based on simulations without star formation ) . multiple , simultaneous minor mergers likewise produce remnants less triaxial than collisionless binary merger remnants @xcite , but the corresponding kinematics have not yet been studied in detail . successive minor merging does not necessarily lead to different final remnants , at least if the cumulative merged mass becomes similar to the most massive progenitor @xcite . again , detailed predictions for the orbital make - up and the shapes of the line - of - sight velocity distributions have not yet been worked out . note that the central dark matter densities in coma ellipticals are larger than in present - day spirals @xcite . even if ellipticals have formed by some variant of merging , present - day spiral galaxies are unlikely the progenitors for the bulk of giant ellipticals ( see also @xcite ) . @xcite pointed out that @xmath0-body systems , which have assembled hierarchically in their cosmological simulations @xcite , or by binary mergers with star - formation and black - hole feedback are consistent with the trend between @xmath3 and @xmath2 in observed galaxies . we have discussed the relationship between anisotropy and flattening in toy models , in models of real galaxies , in merger remnants and in models of merger remnants . models of observed galaxies generally exhibit @xmath158 and @xmath150 . we do not find strong correlations of the anisotropy parameters @xmath3 , @xmath4 and @xmath5 with intrinsic ellipticity @xmath2 . in toy models with maximum entropy for a given density distribution we find @xmath4 to increase with @xmath2 , while @xmath117 . observed galaxies appear close to these maximum - entropy relations , but exhibit a large degree of individuality . rotation appears in anisotropic ( @xmath158 ) as well as isotropic systems ( @xmath73 ) , suggesting that the flattening of the galaxies largely arises from a suppression of stars with large energies perpendicular to the equatorial plane . this is similar to the classical notion of flattening by anisotropy and rules out dfs @xmath72 for most early - type galaxies . the global similarity between models of observed galaxies and our maximum - entropy toy models suggests that early - type galaxies are largely relaxed stellar systems . however , there are differences in the details that probably contain valuable information about the assembly mechanism of the galaxies and will be addressed in a future paper . numerical simulations indicate that both strongly radially anisotropic ( @xmath159 ) and strongly tangentially anisotropic systems ( @xmath160 ) can become unstable ( e.g. @xcite ) . maximum entropy models have intermediate anisotropies and are likely stable . thus , the anisotropies of observed galaxies may not only be understood as being the most likely ones ( in the sense of yielding the maximum entropy at a given flattening ) but could also reflect stability constraints . so far we lack detailed studies exploring the stability of axisymmetric systems with dark matter halos and various intrinsic anisotropies . since our ( three integral ) toy models can be easily transformed to @xmath0-body systems ( cf . @xcite ) they provide a suitable tool to setup both artificially anisotropic as well as realistic and observationally motivated initial conditions . in models of real galaxies the unordered kinetic energy in the azimuthal direction , @xmath161 , can exceed the radial energy @xmath162 by up to 40 percent . this separates real galaxy models from similar models of collisionless @xmath0-body binary disk mergers , which are instead characterised by radial anisotropy ( @xmath147 ) . because we have applied the same modelling machinery to both , the real galaxies as well as the synthetic @xmath0-body merger remnants , our findings indicate a true difference between their intrinsic properties . especially , we have shown that if real galaxies would resemble our merger remnants , then corresponding dynamical models of real data would be radially anisotropic , irrespective of the systematics introduced by the assumption of axial symmetry . the radial anisotropy of the merger remnants is related to a population of central box orbits . because dissipation during a merger can efficiently suppress box orbits , our results suggest that dissipation played an important role during the formation of intermediate mass to massive early - type galaxies . in this paper we focussed on the comparison of real galaxies with collisionless binary disk merger simulations . a similar analysis , but for gaseous mergers with star formation and/or for galaxies formed in cosmological simulations could give more insight into the actual formation paths of elliptical galaxies this work was supported by dfg sonderforschungsbereich 375 `` astro - teilchenphysik '' and dfg priority program 1177 . emc receives support from grant cpda068415/06 by padua university . 99 barnes j. e. , 1992 , apj , 393 , 484 barnes j. e. , hernquist l. , 1996 , apj , 471 , 115 bender r. , mllenhoff c. , 1987 , a&a , 177 , 71 binney j. , 1978 , mnras , 183 , 501 binney j. , tremaine s. , 1987 , galactic dynamics ( princeton : princeton university press ) binney j. , 2005 , mnras , 363 , 937 bournaud f. , jog c. g. , combes f. , 2007 , a&a , 476 , 1179 burkert a. , naab t. , 2005 , mnras , 363 , 597 burkert a. , naab t. , johansson p. h. , jesseit r. , 2008 , apj , 685 , 897 cappellari m. et al . , 2006 , mnras , 366 , 112 cappellari m. et al . , 2007 , mnras , 379 , 418 carollo c. m. , de zeeuw p. t. , van der marel r. p. , danziger i. j. , qian e. e. , 1995 , apjl , 441 , 25 contopoulos g. , 1963 , aj , 68 , 1 corsini e. m. , wegner g. , saglia r. p. , thomas j. , bender r. , thomas d. , 2008 , apjs , 175 , 462 cox t. j. , jonsson p. , primack j. r. , somerville r. s. , 2006 , mnras , 373 , 1013 dehnen w. , 1993 , mnras , 265 , 250 dehnen w. , gerhard o. e. , 1993 , mnras , 261 , 311 dehnen w. , gerhard o. e. , 1994 , mnras , 268 , 1019 dubinski j. , 1998 , apj , 502 , 141 forestell a. , gebhardt k. , 2008 , astro - ph0803.3626 gebhardt k. et al . , 2000 , aj , 119 , 1157 gebhardt k. et al . , 2003 , apj , 583 , 92 gerhard o. e. , kronawitter a. , saglia r. p. , bender r. , 2001 , aj , 121 , 1936 godwin j. g. , metcalfe n. , peach j. v. , 1983 , mnras , 202 , 113 hernquist l. , 1990 , apj , 356 , 359 hernquist l. , 1992 , apj , 409 , 548 hernquist l. , 1993 , apj , 400 , 460 jesseit r. , naab t. , burkert a. , 2005 , mnras , 360 , 1185 jesseit r. , naab t. , peletier r. f. , burkert a. , 2007 , mnras , 376 , 997 jrgensen i. , franx m. , 1994 , apj , 433 , 553 kormendy j. , 1982 , in martinet l. , major m. , eds , morphology and dynamics of galaxies . geneva observatory , geneva , p. 113 krajnovi d. , cappellari m. , emsellem e. , mcdermid r. m. , de zeeuw p. t. , 2005 , mnras , 357 , 1113 lynden - bell d , 1962 , mnras , 124 , 1 magorrian j. , ballantyne d. , 2001 , mnras , 322 , 702 magorrian j. , 2006 , mnras , 373 , 425 mehlert d. , saglia r. p. , bender r. , wegner g. , 2000 , a&as , 141 , 449 merritt d. , stiavelli m. , 1990 , apj , 358 , 399 naab t. , burkert a. , 2003 , apj , 597 , 893 naab t. , jesseit r. , burkert a. , 2006 , mnras , 372 , 839 naab t. , ostriker j. p. , 2007 , astro , arxiv : astro - ph/0702535 naab t. , johansson p. h. , ostriker j. p. , efstathiou g. , 2007 , apj , 658 , 710 nipoti c. , londrillo p. , ciotti l. , 2002 , mnras , 332 , 901 richstone d. o. , tremaine s. , 1988 , apj , 327 , 82 roberts p. h. , 1962 , apj , 136 , 1108 robertson b. , cox t. j. , hernquist l. , franx m. , hopkins p. f. , martini p. , springel v. , 2006 , apj , 641 , 21 rybicki g. , 1987 , in de zeeuw t. , ed . , 127 , structure and dynamics of elliptical galaxies , reidel , dordrecht , p. 397 schwarzschild m. , 1979 , apj , 232 , 236 sellwood j. a. , merritt d. , 1994 , apj , 425 , 530 thomas j. , saglia r. p. , bender r. , thomas d. , gebhardt k. , magorrian j. , richstone d. , 2004 , mnras , 353 , 391 thomas j. , saglia r. p. , bender r. , thomas d. , gebhardt k. , magorrian j. , corsini e. m. , wegner g. , 2005 , mnras , 360 , 1355 thomas j. , jesseit r. , naab t. , saglia r. p. , burkert a. , bender r. , 2007b , mnras , 381 , 672 thomas j. , saglia r. p. , bender r. , thomas d. , gebhardt k. , magorrian j. , corsini e. m. , wegner g. , 2007a , mnras , 382 , 657 thomas j. , saglia r. p. , bender r. , thomas d. , gebhardt k. , magorrian j. , corsini e. m. , wegner g. , 2008 , apj in press valluri m. , merritt d. , emsellem e. , 2004 , apj , 602 , 66 van albada t. s. , 1982 , mnras , 201 , 939 wegner g. , corsini e. m. , saglia r. p. , bender r. , merkl d. , thomas d. , thomas j. , mehlert d. , 2002 , a&a , 395 , 753 weil m. , hernquist l. , 1996 , apj , 457 , 51 de zeeuw p. t. et al . , 2002 , mnras , 329 , 513 the coma galaxy models discussed in this paper are derived using our standard regularisation parameter @xmath163 ( cf . equation [ orbitmaxs ] ) . the strength of the standard regularisation has been calibrated by means of monte - carlo simulations of isotropic rotators @xcite . to check how much the choice of @xmath51 affects our results , we ( 1 ) determined the best - fit dynamical model at @xmath164 and ( 2 ) recalculated the anisotropies of all galaxies from these weakly regularised models ( at @xmath165 , the minimum @xmath45 is usually reached ) . in the top row of fig . [ aniso_aniso : test ] we show both models with standard and with weak regularisation for comparison . as can be seen , lowering the regularisation has almost no effect on the derived anisotropies . especially , there are still at least four galaxies with significantly negative @xmath166 . in the bottom row of fig . [ aniso_aniso : test ] we make a similar comparison for models with and without dark matter : squares indicate the anisotropies of our best - fit models with a constant mass - to - light ratio ( no dark matter halo ) . as expected , when assuming that mass - follows - light , @xmath5 become smaller ( the amount of @xmath28-energy is increased to compensate for the missing dark mass ) . from the bottom - right panel one would expect that the average @xmath5 becomes negative when the radial increase of the mass - to - light ratio ( caused by a dark halo ) is neglected . this is not the case in the sauron sample , however , although @xcite assumed that the mass - to - light ratio is constant with radius in their models . that neglecting dark matter has a stronger effect in coma galaxies is likely related to the fact that our kinematical data reach out into the region where dark matter becomes noticeable ( @xmath167 ) , which is probably not the case in many sauron galaxies ( where the data extend only out to @xmath168 ) ( top ) and @xmath169 ( middle ) for maximum - entropy models ( @xmath57 ; flattening @xmath170 ) with dark matter halos . left - hand side / squares : halo tuned to result in an approximately flat circular velocity curve ; right - hand side / triangles : halo leading to an increasing circular velocity in the outer parts of the model . for comparison , the case without halo is shown on both sides ( circles ) . circular velocity curves of the models ( scaled to the maximum circular velocity @xmath171 without halo ) are shown in the bottom panels.,width=317 ] the maximum - entropy toy models @xmath57 discussed in sec . [ subsec : toy ] resemble the classical flattening by anisotropy , except that they are only approximately isotropic in @xmath30 and @xmath172 . to investigate where this anisotropy comes from , we have constructed maximum - entropy toy models in potentials that include a dark matter halo . the halo density distribution is assumed to follow @xmath173 @xcite . to mimick realistic halos ( cf . @xcite ) we choose a flat central density core ( @xmath174 ) and we set the flattening @xmath56 of the halo equal to the flattening of the luminous component of the toy model ( cf . equation [ eq : flattening ] ) . we investigated three mass models : ( 1 ) no halo , ( 2 ) a mass model that has a roughly constant circular velocity curve and ( 3 ) a mass model with a rising @xmath175 in the outer parts of the model . the corresponding circular velocity curves for @xmath170 are shown in the bottom panels of fig . [ aniso_vcirc ] . the upper panels of fig . [ aniso_vcirc ] display the radial profiles of the local anisotropies @xmath177 and @xmath178 along the equatorial plane ( averaged within @xmath179 , where @xmath180 is the latitude ) . as one can see , the anisotropy in the meridional plane ( @xmath181 ) does not depend on the shape of the gravitational potential . thus , the relation between @xmath4 and @xmath2 is largely independent from the gravitational potential and closely related to the entropy maximisation . however , beyond @xmath134 , where dark matter starts to influence the shape of the circular velocity curve , @xmath169 is different in the three different potentials . [ aniso_vcirc2 ] shows that the local value of @xmath169 along the equatorial plane is directly connected to the logarithmic slope @xmath182 of the circular velocity curve . in general then , because @xmath5 from equation ( [ eq : gamma ] ) is the spatial average of @xmath169 ( and the local anisotropies along other position angles in the meridional plane ) , its exact value is not set uniquely by the entropy maximisation but also depends on the shape of the circular velocity curve . in practice , however , deviations with respect to the model without halo become noticeable only beyond @xmath134 , such that even the spatially averaged @xmath5 of the toy models does not depend strongly on whether a halo is included or not . note that the relation revealed by fig . [ aniso_vcirc2 ] is different from the epicycle relation ( [ eq : epicyc ] ) . this is expected , because the azimuthal velocity dispersion @xmath101 in the toy models largely results from the fact that they do not rotate . instead , the dispersion predicted by the epicycle approximation arises from perturbations on circular orbits in a rotating disk .	we use oblate axisymmetric dynamical models including dark halos to determine the orbital structure of intermediate mass to massive coma early - type galaxies . we find a large variety of orbital compositions . averaged over all sample galaxies the unordered stellar kinetic energy in the azimuthal and the radial direction are of the same order , but they can differ by up to 40 percent in individual systems . in contrast , both for rotating and non - rotating galaxies the vertical kinetic energy is on average smaller than in the other two directions . this implies that even most of the rotating ellipticals are flattened by an anisotropy in the stellar velocity dispersions . using three - integral axisymmetric toy models we show that flattening by stellar anisotropy maximises the entropy for a given density distribution . collisionless disk merger remnants are radially anisotropic . the apparent lack of strong radial anisotropy in observed early - type galaxies implies that they may not have formed from mergers of disks unless the influence of dissipational processes was significant . [ firstpage ] galaxies : kinematics and dynamics galaxies : elliptical and lenticular , cd galaxies : formation
co - line observations play an essential role in studying the kinematics and interstellar physics in the central regions of spiral galaxies , where the interstellar matter is mostly in the molecular - gas phase and is strongly concentrated ( sofue et al . 1995 ; honma et al . there have been numerous observations of nearby galaxies in the co line emissions with single dish telescopes ( young & scoville 1991 ; braine et al . 1993 ; young et al . 1995 ; nishiyama & nakai 2001 ) . large - scale co line surveys of the virgo galaxies have been obtained using the fcrao 14-m telescope at an angular resolution of 45@xmath6 by kenney & young ( 1988 ) , and the btl 7-m telescope by stark et al . these surveys with single dishes were made with angular resolutions of tens of arcsec , and have given information about the global structure of molecular disks in virgo and nearby galaxies . interferometer observations at high angular resolutions are crucial for studying detailed molecular disk structures within the central few hundred parsecs ( sargent & welch 1993 ) . high - spectral resolution is also crucial to investigate the detailed kinematics of the central gas disks . both high spatial and high spectral resolutions provide us with precise velocity fields and rotation curves , which are the basis for deriving the fundamental parameters such as the mass distribution , bars and related shock phenomena , triggering mechanism of starburst and/or fueling mechanism of massive black holes . interferometer observations have often performed to investigate the individuality of each galactic center and activity . recently , some large surveys of nearby galaxies have started to be reported . the nobeyama mm - wave array ( nma ) and owens valley radio observatory ( ovro ) mm - wave array were used since 1990 s to map the central regions of nearby spiral galaxies with the co line at a typical angular resolution of @xmath7 ( sakamoto et al . the berkely - illinois - maryland association survey of nearby galaxies ( bima song ) has mapped 44 nearby galaxies at typical resolutions of @xmath8 ( regan et al . 2001 ) . interferometer observations of several nearby galaxies have been also conducted from various interests , such as bars ( e.g. , kenney et al . 1992 ; regan et al . 1999 ) , star formation ( e.g. , wong & blitz 2002 ) , and nuclear activity ( e.g. , baker 1999 ; sakamoto et al . 1999a ; kohno et al . 1999 ; schinnerer et al . 1999 ) . the vics ( virgo high - resolution co survey ) project with the nma has been performed in order to obtain a homogeneous high angular- and spectral - resolution database for a large number of co - bright virgo cluster spirals in the @xmath9 line . angular resolutions were @xmath10 after reduction in the conventional clean procedure with natural weighting . the major scientific motivation was to investigate the detailed central kinematics of the galaxies , particularly the innermost rotation curves from analyses of position - velocity diagrams across the nuclei , which would be effective to detect central compact massive objects . the data are also useful for investigation of the kinematics and ism physics of the central molecular disks , and their environmental effect in the cluster circumstance . an advantage to observe the virgo cluster galaxies is their almost identical distance , which has been accurately determined to be 16.1 mpc ( @xmath11 corresponds to 78 pc ) by the cepheid calibrations ( ferrarese et al . since our target galaxies lie within 2 mpc from the virgo center , m87 , the distance ambiguity will be at most 15% , mostly less than 10% . the accurate distance will enable us to estimate physical quantities rather precisely , such as the co and dynamical masses , and linear scales of gas disks . the vics results will be published in a series of papers . in this paper we describe the overall observations and reduction , and present an atlas of the central molecular disks of virgo galaxies . in the forthcoming papers we will describe more details of the observations , analyses , and results for individual galaxies , as well as rotation curves and central kinematics , investigations of the ism physics , and comparison with other wavelengths observations . the database will be opened for public use on our web page . the target galaxies in the survey have been selected from the list of spiral galaxies of the fcrao co - line survey by kenney & young ( 1988 ) by the following criteria . 1 . the sources were chosen in the order of co line peak antenna temperatures @xmath12 at the optical centers . twenty - eight galaxies with the peak antenna temperatures above @xmath13 were selected from 42 galaxies of kenney & young ( 1988 ) . inclination angles were limited to be @xmath14 in order to investigate central gas dynamics . this criterion excluded ngc 4293 ( @xmath15 ) , ngc 4302 ( @xmath16 ) , ngc 4312 ( @xmath17 ) , and ngc 4710 ( @xmath16 ) . 3 . galaxies with morphological type of s0 , i.e. ngc 4293 , ngc 4710 and ngc 4438 , were excluded . interacting galaxies were excluded by a criterion that the galaxies have no companion within @xmath18 radius . pairs of ngc 4568/4567 , ngc 4298/4302 , and ngc 4647 were excluded . peculiar galaxies in optical images , i.e. ngc 4438 and ngc 4064 , were excluded . galaxies observed with the nma since 1994 were excluded . ngc 4321 and ngc 4527 have been observed by sakamoto et al . ( 1995 ) and sofue et al . ( 1999 ) , respectively . sixteen galaxies were selected on the basis of these criteria , and we have observed 15 galaxies except ngc 4450 . the targets are listed in table 1 , which also summarizes the morphological type , b - band total magnitude , optical size , inclination , position angle from optical isophotal contours , and nuclear activity from optical spectroscopy ( ho et al . 1997a , b ) . the table list also the co - line peak temperature , integrated intensity , intensity - weighted mean velocity , and velocity width from the single dish co - line observations ( kenney & young 1988 ) , the selection criterion 1 by the peak antenna temperature was applied because of higher probability of detection in a single channel . we note that the fcrao survey with 45@xmath6 resolution ( kenney & young 1988 ) shows that all galaxies , except one , are centrally co peaked . their data show that the peak temperature and maximum co intensity have approximately a linear correlation , and that the co scale radius is an increasing function of maximum co intensity . this implies that our selection by peak temperature is approximately equivalent to a selection by total integrated co intensity , and hence by total co luminosity for their equal distances . hence , our target galaxies , and some that were not selected by the reason that they were already observed by nma , would represent the most co luminous virgo galaxies . table 1 we have performed aperture synthesis observations of the @xmath9 line emission from the 15 virgo galaxies listed in table 1 in the course of a long - term project during the winter seasons of 2000 ( 1999 dec . -2000 apr . ) , 2001 ( 2000 dec.-2001 apr . ) and 2002 ( 2001 dec.-2002 apr . ) . we made the observations in three available configurations : ab ( long baselines ) , c ( medium ) and d ( short ) configurations . the visibility data covered projected baselines from about 10 to 351 m. the nma consisted of six 10-m antennas , providing a field of view with a fwhp beam width of @xmath19 at 115 ghz . since interferometry observations sample data in a fourier space , the range of collected fourier components , or baselines , determines the detectable sizes of objects . in our observations , the antenna size limited the minimum projected baseline length , restricting the largest detectable size to about @xmath20 . thus our data may miss some fluxes of extended components of the objects . table 2 lists the observation periods and array configurations , observed central frequencies , and the positions of pointing centers as well as phase - reference centers for individual galaxies . table 2 the antenna were equipped with tuner - less sis receivers , which had receiver noise temperatures of about 30 k in double sidebands , and the typical system noise temperatures were about 400 k in single sideband . we used a digital spectro - correlator system ( okumura et al . 2000 ) , which had two spectroscopic modes ( bandwidth of 512 and 1024 mhz ) ; we used the mode covering 512 mhz ( @xmath21 ) with 256 channels and 2 mhz ( @xmath22 ) resolutions . the nearby radio point source 3c 273 was used as the flux and phase calibrator , which was observed every 20 minutes . the band pass response across the channels was also calibrated using the 3c 273 data . the intrinsic flux density of 3c 273 at the observing frequency was calibrated for each observing run ( typically 5 days ) using the planets ( mars etc . ) . the flux of 3c 273 during the three years of observing periods was gradually variable between 9 and 12 jy . the uncertainty in the absolute flux scale for each observing run was @xmath23 15 % , which apply to all results presented in this paper . the raw data were calibrated using uvproc - ii , a first - stage reduction software ( tsutumi et al . 1997 ) , and were fourier - transformed using the nrao astronomical image processing system ( aips ) . we applied the clean method with natural weighting to obtain three dimensional data cubes ( ra , dec , @xmath24 ) . the intensity data were averaged in 2 to 6 bins ( @xmath25 ) of the original channels in the spectro - correlator , and the channel increments were set to 2 to 4 ( @xmath26 ) . the intensity scale at this stage was in jy per synthesized beam , which can be converted to brightness temperature in kelvin . the resultant synthesized beam sizes range in @xmath2 . the typical rms noise scaled for a @xmath5 channel was @xmath27 . table 3 lists the resultant parameters of the data cubes for individual galaxies . table 3 in addition to the above reduction parameter set , we cleaned the data with tapered and uniform weighting functions , which provided low ( @xmath28 ) and high ( @xmath29 ) resolution maps , respectively . we will present those maps in separated papers for discussing individual galaxies . we calculated the fractions of the recovered single - dish flux from our aperture synthesis observations , which are listed in table 3 . we first corrected the data cube for primary beam attenuation , and convolved with a gaussian single - dish beam of fwhm @xmath30 ( comparable to that of the fcrao survey ) , and took the flux at the pointing center of the fcrao observations . the recovered fluxes were typically 80% , which recovered almost all fluxes within the field of view . however , a few galaxies showed exceptionally low and high recovered fluxes . the recovered flux of ngc 4535 was greater than the fcrao flux by about twice . we made careful analyses of the raw data for several times : we made uv data for c , d , c+d , and ab+c+d array configurations , which were obtained in independent observing periods ( table 2 ) . we , then cleaned them separately , but obtained about the same flux for all the configurations . also the rms noises of the reduced cubes and maps are comparable to those for other galaxies observed in the same periods . hence , we conclude that the flux calibration was correct . the flux disagreement could be possible , if the fcrao flux was about significantly under - estimated , and ours was about 15% over - estimated , both within the measurement errors . the recovered flux of ngc 4548 was only 16% of the fcrao flux . this may have happened due to larger sizes of the missing components than the maximum detectable size of our observations ( 2.2 ) . alternatively , it could be due to very low brightness of the extended components . in fact , the fcrao flux is as weak as 6.7 k with 45@xmath6 beam . if the rest 84% is extended in the 45@xmath6 beam , the intensity for our beam ( @xmath31 ) would be 5.6 k . for an assumed line width of about 20 in our beam , the expected brightness is only 30 mk , which is much below our detection limit . we have checked for continuum sources in the galaxies by making channel maps for a wide range of velocity . the channel maps for individual galaxies are shown in figure a1 in the appendix , where the outernost channels can be used to check continuum fluxes . as figure a1 shows , no significant continuum source has been detected in any of the observed galaxies by the present sensitivity , which is typically 20 mjy beam@xmath32 in rms noise as listed in table 3 . nevertheless , such galaxy as ngc 4579 , which contains an agn , could have some continuum emission . so , we applied deeper continuum checking for this galaxy . we cleaned the 2 mhz @xmath33 256-channel data cube of ngc 4579 by binning every 32 channels ( 168 km / s ) . however , no continuum source stronger than 10 mjy was found in the outermost - velocity channels ( @xmath34 ) , where no co line emission is expected . figure 1 ( top - left panels ) shows optical ` looks ' of the observed galaxies for a @xmath35 area taken from the stsci digitized sky survey ( dss ) second generation blue images . the inserted squares show areas for the co maps . we present total integrated intensity maps of the co emission in the bottom left panel . the intensity maps were obtained by using the aips task momnt , which integrated the intensities by velocity only when the intensity exceeds a threshold level . the threshold level was taken to be 2 to 3 times the rms noise in the data cube . channel maps of the observed galaxies are shown in appendix , and will be discussed in more detail in the forthcoming papers of this series on individual galaxies . the bottom - right panel shows intensity - weighted velocity fields for the @xmath36 regions ( our field of view is @xmath19 at fwhp at 115 ghz ) , and the top - right panels show position - velocity diagrams along the major axes ( top right ) , except for ngc 4254 and ngc 4402 , for which @xmath37 regions are shown . the primary - beam correction has not been applied in these maps . the intensity scales in the maps are in kelvin of brightness temperature , rather than in @xmath38 scale that is directly derived from interferometer observations , for convenience to compare with single dish observations and to convert to the molecular hydrogen column density . we measured the peak co intensities using these maps and listed in table 5 together with the co peak brightness temperatures as read from the data cubes . table 5 figure 2 shows the co intensity maps in the same angular and linear scales . each box covers a @xmath39 region , which corresponds to 4.7 kpc @xmath33 4.7 kpc region for an assumed virgo distance of 16.1 mpc ( ferrarese et al . figure 3 shows velocity fields corresponding to figure 2 . in figure 4 we plot the co intensity maps in the same angular scale on the virgo cluster region , where each map scale has been enlarged by 50 times the real angular size . fig . 1 fig . 2 fig . 3 fig . 4 figure 1 ( bottom left panels ) shows intensity - weighted velocity fields for the observed @xmath36 regions , which are the same regions for the integrated - intensity maps at the bottom right panels . for ngc 4254 and ngc 4402 , the @xmath37 regions are presented . figure 3 shows the velocity fields in the same angular scales . the general pattern of the velocity field in figures 1 and 3 is a symmetric spider diagram , indicating a regular circular rotation of the co disk . slight non - circular streaming motions , such as due to spiral arms and bars , are superposed on the regular rotation . however , there are some galaxies that show strong non - circular motion ; ngc 4569 has an extremely large deviation from the circular rotation , indicating either high - velocity streaming or a large - amplitude warping in the central disk . position - velocity ( pv ) diagrams are shown in figure 1 ( top right ) . these diagrams were made by slicing the data cubes along the optical major axes with appropriate widths for individual galaxies . most galaxies have steeply rising rotation velocity in the central 100 to 200 pc radii . such sudden rise of rotation velocity in the close vicinity of the nuclei had not been clearly detected in the lower - resolution observations . one of the major purposes of the present co survey was to obtain the central rotation curves to investigate possible central massive cores , which have been found in many nearby galaxies ( sofue et al . 1999 ; takamiya & sofue 2000 ; sofue and rubin 2001 ; sofue et al 2001 ; koda et al 2002 ) . in a separated paper ( sofue et al . 2003a ) , we describe the result of detailed analyses of the pv diagrams and derivation of accurate rotation curves by applying a new iteration method ( takamiya & sofue 2002 ) , and discuss the central mass distribution . in order to give an overview on the general characteristics of distributions of the molecular gas ( co intensity ) in the observed galaxies , it is helpful to compare the galaxies in a unified scheme . in figure 2 , we present the observed @xmath40 distributions in the same angular and intensity scales . the image sizes are all @xmath41 , corresponding to 4.68 kpc @xmath33 4.68 kpc for an assumed distance of 16.1 mpc . the contours are drawn at the same levels of 5 , 10 , 20 , 40 , 80 , 160 @xmath42 for all galaxies . figure 3 shows the same as figure 2 , but for distributions of intensity weighted velocities . the velocity fields generally show a spider diagram pattern , indicating circular rotation of the molecular disk . the rotation velocity rises rapidly in the central few hundred parsecs , which is more clearly observed in the position - velocity diagrams in figure 1 . in many galaxies , the spider diagrams are more or less distorted , indicating either non - circular streaming motion or warping of the gas disk . in figure 4 we plot the @xmath40 maps on the sky area of the virgo cluster in a similar manner to a plot of hi maps by cayatte et al . the angular scales are enlarged by a factor of 50 for individual galaxies . the co distributions appear to be not strongly correlated with the distance from the center of the cluster at m87 , which is marked by a cross . this property is very different from the hi gas distribution ; the hi disks of inner - cluster galaxies are usually largely distorted and are often truncated by the ram - pressure of the intracluster medium ( cayatte et al . 1990 ) , while the central co disks are not strongly perturbed by the ambient gas in the cluster , probably because they lie deep in their galactic potentials . the calibrated and reduced data presented in this paper will be opened on our web page in the form of fits formatted cubes and maps , and in gif - formatted images at the url , http://www.ioa.s.u - tokyo.ac.jp / radio / virgo/. in order to determine the central positions , we fitted a disk model with a brandt - type rotation curve ( brandt 1965 ) to intensity - weighted velocity fields using the aips task gal . since this task assumes a pure circular rotation , we used only central several arcseconds for the fitting , out of which the isovelocity contours indicate some deviations from circular rotation . the iteration in the task could occasionally provide different results with different initial guesses . we checked this error by changing the initial guess in appropriate ranges , and confirmed that the task suffices to provide the dynamical centers accurate to about @xmath11 in most cases . the derived central positions are listed in table 4 . the thus obtained center positions were in coincident with the ned center positions within an arcsecond in most cases . however , in such cases that the central co distribution is not smooth , or the velocity field is strongly perturbed , and hence above dynamical centers are not reliable , we determined the central positions from the literature . the center of ngc 4212 , which shows patchy co distribution ( figure 1 ) , was adopted from the optical observations by cotton et al . ( 1999 ) , which coincides with our co emission peak within an error of @xmath43 . ngc 4569 shows strong noncircular motions in the velocity field and position - velocity diagram ; we adopted the central position determined by sakamoto et el . ( 1999 ) from their co interferometry observations with a lower resolution . the center of ngc 4579 was taken to coincide with the position of the unresolved radio continuum source ( ho and ulvestad 2001 ) . table 4 figure 5 displays azimuthally averaged radial profiles of co - line intensities in unit of k @xmath44 as projected on the galaxies disks corrected for the inclinations . in order to make these plots , integrated - intensity maps without clipping were corrected for the primary beam attenuation , and we applied the aips task iring around the central positions derived in [ sec : dycen ] . we fixed the inclination and position angles for each galaxy to those provided from optical observations ( table 1 ) . the sampling intervals in the plots were set to be @xmath45 . however , the effective sampling intervals are equal to the beam widths . the number of effective sampling points for the fit increases with radius @xmath46 proportionally to @xmath46 , and hence the statistical error decreases proportionally to @xmath47 . the intensity error at each point in a map is given by @xmath48 ( @xmath49 k ) , where @xmath50 k ) , @xmath51 , and @xmath52 are the rms , number of channels within the expected velocity width , and the velocity interval , respectively , as given in table 3 . therefore , the typical error of the profiles is given approximately by @xmath53 k . fig . 5 the intensity distributions in the central 10 to 15@xmath6 regions are approximately exponential with scale radii 5 to 10@xmath6 , or 400 to 800 pc . these scale radii are a few times smaller than those derived by regan et al . ( 2001 ) . since our survey has three times higher resolution ( 2 `` ) than theirs ( 6 '' ) , we may detect the central cusps of the co distributions . in most cases , the outer regions than 20@xmath6 are not significantly detected , except for some cases with disk components . three of our target galaxies , ngc 4254 , 4501 and 4569 , have been observed in co line with lower sensitivity and resolutions ( sakamoto et al . our radial profiles for ngc 4501 and 4569 are consistent with those from the previous observations . the profile for ngc 4254 is also roughly consistent with the previous result , while there are some small scale deviations . the ellipse fit gives a quantitative presentation of the radial profiles including the outskirts . however , the fit looses linear resolution , depending strongly on the inclination , because it uses the data in the minor axis direction with an equal weight . figure 5 shows that many galaxies have a strong concentration of co gas within the central 10@xmath6 ( 0.8 kpc ) radius , while some have a plateau or a dip at the center . we discuss such galaxies in 5 as a central-/single - peak and twin - peaks types , respectively . the molecular gas distributions show a wealth of variety . although it is difficult to categorize them in a simple way , we can find some characteristic types in the central gas distributions . many galaxies have high concentration of co gas in the central a few kpc region , where the co morphology shows either single peak or twin peaks . more extended components have more variety of morphologies , which can be classified into arm type , bar type , and amorphous type . 1 . central peak and/or single peak : many galaxies show strong concentration of molecular gas around the nuclei in so far as the present maps are concerned . ngc 4212 , ngc 4419 , ngc 4501 , ngc 4535 , and ngc 4536 are the examples . the typical size of these central peaks is about 200 - 400 pc . in most cases , their peaks are single at the present resolution , which we call `` single peak '' . figure 6 shows the co intensity distributions in the central @xmath54 regions ( 1.6 kpc square ) of the central-/single - peak galaxies . the central / single peak galaxies shares a considerable fraction among the observed galaxies . note that the object selection was made by peak antenna temperature in the fcrao survey with a 45@xmath6 beam . hence , the present maps could have a selection effect for galaxies with higher peak - temperatures . however , as argued in 2 , our object selection is approximately equivalent to a selection by co luminosity . we may consider that the statistics with the presently observed galaxies is significant to discuss the general types of central co morphology of the most co luminous virgo galaxies . + fig . 6 2 . twin peaks : typical example of twin - peak molecular gas distribution is seen for ngc 4303 , which shows two offset open molecular arms along the optical bar , which end at a molecular ring with two peaks . kenney et al . ( 1992 ) have reported both single - peak , and twin - peaks types . they selected four barred galaxies with strong co and fir emission , and showed that the barred galaxies have twin peaks in co likely as the consequence of bar - induced inflow . in so far as the present data set is concerned , which includes galaxies of random types with the co emission concentrated in the central @xmath55 ( 3.5 kpc ) regions , twin - peak galaxies shares rather a small fraction . kenney et al . ( 1992 ) showed that the separations between twin peaks are about 200 - 400 pc , while our single peak galaxies do not have double peaks even in the same scale ( @xmath56 ) . note , however , that this classification may depend on the spatial resolution ; it may happen that a single peak at our resolution consists of a more number of inner structures at higher resolution . in fact , ngc 4501 appears to be a single peak type in the present atlas , while a higher resolution image show a small patchy ring with a diameter of @xmath57 ( onodera et al . 2003:private communication ) . pv diagrams may apparently imply spatially - unresolved double peaks with separated velocities . however , a single peak galaxy may apparently show two peaks on a pv diagram , at the positive and negative terminal velocity ends at turnover radii of central rising and outer flat rotation curves , even when the gas distribution has no spatially - separated double peaks ( sofue et al . 1999 ; sakamoto et al . hence , pv diagrams may not be used for the spatial morphological classification . we describe individual galaxies about their co properties obtained from the present observations . an extremely bright co peak is observed at the center , which classifies this galaxy in a `` central - peak '' type , while it appears that the peak may be resolved into two peaks at higher resolution . hence , the classification depends on the resolution . the central peak is surrounded by a bright disk at high inclination at about the same inclination as the optical disk . the position - velocity diagram shows a very high - velocity rotation , whose maximum reaches almost @xmath58 . however , the central co peak is rotating more slowly at about @xmath59 . the co distribution consists of a molecular core and extended straight arms in the direction of the major axis . the core is shifted from the map center toward the ne by a few arcseconds . the velocity field and pv diagram indicate that the rotation is rigid body - like , while the core shows a steeper velocity gradient . the central molecular gas distribution shows a bar - like elongation , while no optical bar feature is seen in the visual - band images . the co intensity has a slight depression at the dynamical center , which coincides with the nucleus . two well - developed spiral arms wind out from the bar ends toward the south and north . the south - eastern arm bifurcates into a tightly - wound dense molecular arm with an almost zero pitch angle . hence , the molecular disk has three arms , and the arms are well correlated with optical dark lanes . the velocity field shows a regular spider pattern , indicating a circular rotation of the disk , on which small - amplitude streaming motion due to the spiral arms are superposed . the pv diagram shows a sharp rise in the central few arcseconds , indicating a massive core , and then the velocity increases gradually . overall distributions and kinematics agree with the previous low resolution observations ( sakamoto et al . a detailed study of this galaxy with consideration of the ram - pressure effect by the intra - cluster medium is presented in sofue et al . ( 2003b ) . co gas is highly concentrated in the nuclear disk within a radius @xmath60 ( 600 pc ) . the nuclear disk comprises the `` twin peaks '' at the eastern and western edges of the nuclear disk , and there appears to exist a diffuse central component around the nucleus between the twin peaks . two prominent bisymmetric spiral arms , or offset ridges , wind out from these twin peaks , and extend toward the north and south along the dark lanes in the optical bar . the pv diagram along the major axis ( north - south ) indicates a rise of rotation velocity within @xmath61 ( 160 pc ) to 160 - 180 . our result is consistent with the high - resolution observations with the ovro interferometer by schinnerer et al . detailed description of this galaxy are given in a separate paper of this series by koda et al . ( 2003 : private communication ) . the co intensity distribution shows a high density nuclear molecular disk of @xmath62 . the nuclear disk is surrounded by a more extended molecular disk of radius @xmath63 ( 2 kpc ) . this outer disk appears to be consisting of two spiral arms , one extends to the west from the southern edge of the nuclear disk , tracing the dark lane , and the other arm toward the east from the north - eastern edge of the nuclear disk . the velocity field shows a usual spider diagram superposed by some streaming motion in the molecular ring / arms . the pv diagram shows a nuclear component and outer ring / arms . the co gas is strongly concentrated in the central @xmath64 radius disk , which is associated with an elongated outer disk component . kenney et al . ( 1990 ) also reported the concentrated co distribution . the outer molecular disk is lopsided toward the north - west . the pv diagram indicates that the central component has a rotation velocity as high as 150 within 5@xmath6 radius , which is followed by a gradually rising disk rotation . a nuclear concentration of the molecular gas of radius @xmath65 is remarkable , as reported by sakamoto et al . ( this is classified as the `` single - peak '' type , although the peak intensity is not particularly high compared with the other typical single peaks . this central peak is surrounded by an extended component elongated in the se to nw direction , with the se end at 5@xmath6 radius being brighter . two prominent molecular arms are running at @xmath66 . the north - eastern arm is much stronger than the south - western arm . both arms are associated with the dark lanes along the optical spiral arms . the velocity field and pv diagram indicate sharp rise of rotation velocity in the nuclear disk . detailed description and modeling by spiral - shock accretion mechanism are given in onodera et al . ( 2003:private communication ) . the molecular gas shows a strong concentration in the central region of @xmath67 radius . this galaxy is a typical `` single - peak '' type . offset bars are extending from the central disk toward the ne and sw , coinciding with the optical dark lanes in the bar . the velocity field shows a usual spider diagram with the zero velocity node at position angle of @xmath68 , coinciding with the optical minor axis . however , the co arms along the dark lanes show some non - circular streaming velocity , indicating inflow along the arms . the pv diagram shows a sharply rising , but rigid body - like behavior within the central molecular disk . this is also a typical `` single - peak '' type galaxy with the molecular gas being concentrated in the nuclear disk of @xmath69 radius and an unresolved compact core exists at the nucleus . the velocity field shows a spider diagram , and the pv diagram indicates that the rotation velocity rises to 200 within the central @xmath70 . there appears no strong non - circular streaming motion . this is a typical barred galaxy . co emission is very weak compared with the other galaxies . the co distribution is highly concentrated near the nucleus , being centrally peaked , and no extended emission is detected . the recovered flux is only 16% of the fcrao flux with 45 " beam ( 2.3 ) , and the rest 84% ( 5.6 k ) could be due to very extended components with sizes greater than our maximum detectable size ( 54@xmath6 ) . the molecular gas is highly concentrated within @xmath71 kpc radius . the co intensity distribution is elongated in the same direction as the optical major axis , and has two peaks with depression at the nucleus , consistent with the earlier co map ( sakamoto et al . thus , the central molecular morphology of this galaxy may be classified in twin peaks at the present resolution . however , a higher resolution co map reveals that the apparent two peaks coincide with both ends of an elliptical molecular ring , while they are not associated with so called offset ridges ( nakanishi et al . the velocity field is strongly disturbed from circular rotation , and the pv diagram indicates significant ` forbidden ' velocities . nakanishi et al.(2003 ) discuss the kinematics of this galaxy in detail , and tried to explain these features using two models of non - circular motion and warping of the inner disk . and they conclude that it is natural that disturbed velocity field and forbidden velocities of the pv diagram are due to non - circular motion . helfer et al . ( 2001 ) have reported an extended co emission from their wide - field mosaic image . jogee et al . ( 2001 ) have also presented a high resolution co image of this galaxy . no significant detection of the co line was obtained for this galaxy , not only in channel maps , but also in an integrated - intensity map . our resultant rms was @xmath72 for a @xmath73 width , that is greater than @xmath74 from single dish observations ( table 1 ) . since the map shows only noises , we do not present the result . the co distribution is elongated in the east - west direction , about @xmath75 displaced from the optical bar axis . there are two major co peaks with asymmetric peak intensities , which are associated with symmetric spiral features as reported by kohno et al . the velocity field shows a higher rotation velocity than 200 in the central few arcseconds , which is more clearly visible in the position - velocity diagram . this galaxy is known for its lopsided structure in the optical as well hi disks , most likely due to the ram - pressure effect of the intracluster gas , blowing from the northwest ( phookun and mundy 1995 ) . the co distribution is also lopsided , in the same direction as that of the hi and optical image tail . the lopsided co distribution suggests that the ram pressure effect is not negligible even in such a central molecular disk . moreover , the co distribution is more elongated than the optical / hi disks . the velocity structure is rigid - body like with mildly increasing rotation velocity with the radius . such rotation characteristics is exceptional among the presently observed pvds . like the optical spiral arm features , the co intensity distribution is amorphous , and is patchy and widely extended . there appears neither spiral arms nor bars in co , and no central peaks are found . the peak @xmath40 amounts only to @xmath76 , the lowest among the observed galaxies . the velocity field indicates a regular rotation pattern , but the central rise of rotation velocity is mild , as indicated by the pv diagram . we have obtained high - resolution co - line survey of 15 virgo spiral galaxies using the nobeyama millimeter - wave array in ab , c and d configurations , and presented the result in the forms of integrated intensities , velocity fields and position - velocity diagrams along the major axes . the galaxies were sampled from the co brightest galaxies in the kenney and young s ( 1988 ) list without any bias . the co properties may be compared with each other without ambiguity of the linear scale , as the distance to the virgo cluster galaxies are safely taken to be 16.1 mpc from the cepheid calibration ( ferrarese et al . for the homogeneity , our data will be useful for investigating correlation between the co properties and other characteristics . we will discuss the correlation of the central peaked co distributions and nuclear activities in a forthcoming paper of this series . in the second paper of this series , we will derive exact rotation curves by analyzing the position - velocity diagrams , and discuss the dynamical properties of the central regions of the virgo galaxies , as well as detailed mass distributions . we summarize the results obtained in this paper as follows , and discuss the implications below . the mean radial profiles of the molecular gas distribution in the inner 10 to 15@xmath6 radius regions are approximately exponential with @xmath77-folding scale radius of 400 to 700 pc . in some galaxies , more extended disk components are detected , whose scale radii are greater , while the present interferometry data are not appropriate to determine the disk radii precisely . a careful inspection of the intensity maps shows that the observed intensity distributions have a variety of types , which may be classified into the following types . the centrally concentrated components can be classified into two types : the central - peak or single - peak type and twin peaks type . the latter shows plateau - like radial profiles near the center . the distributions of more extended components can be classified into spiral arms , bars , and amorphous types . it is of particular interest to consider what causes the variety of molecular gas morphology , as it could be intimately related to the activities in the centers of galaxies . twin - peaks of molecular gas at the ends of a set of two bisymmetric offset molecular bars ( dark lanes ) along the optical bar have been noticed for several barred galaxies in relation to fueling mechanisms of interstellar gas toward the central regions ( kenney et al . 1992 ; sakamoto et al . ngc 4303 is a typical case for the twin - peaks type , and the structure is well explained by a bar - potential and galactic shock hypothesis ( schinnerer et al 2002 ; koda et al 2003:private communication ) . however , the fraction of galaxies having `` twin - peaks '' is not particularly high in so far as the present resolution maps ( @xmath78 or 150 - 300 pc ) are concerned . on the other hand , `` central- '' or `` single - peak '' galaxies share a larger fraction : five galaxies among the fifteen in the present survey at our resolution . examples are ngc 4192 ( sab(s)ab ) ) , ngc 4419 ( sb(s)a ) , ngc 4501 ( sa(rs)b ) , ngc 4535 ( sab(s)c ) and ngc 4536 ( sab(rs)bc ) . twin - peaks are often thought to be a consequence of characteristic gaseous orbits in a bar potential : x1 and x2-orbits in a bar intersect each other at the inner ilr , causing collision of the gas at the two intersecting points , and consequently producing twin peaks . although this simple interpretation of the twin - peaks is attractive , the current study has shown that the single - peak is more popular than the twin - peaks . it is an interesting subject to consider the origin of the single - peak , while we need a more careful simulation and gas dynamics . we here try to speculate a possible mechanism . first of all , the twin - peaks are not the final and stable structure of the gas distribution in a bar . the selfgravity of the gas will cause a collapse in the gas structures , particularly such large clumps as twin - peaks will be gravitationally unstable , and cause further infall of gas into the center due to the friction among the clouds ( wada & habe 1992 ) . moreover , if there exists central massive object , the friction due to stronger shearing motion will accelerate the accretion toward the nucleus ( fukuda et al . in fact , most of the galaxies show very steep rise of rotation curve in the central @xmath79 pc region , indicating the existence of massive compact cores of mass @xmath80 around the nuclei ( sofue et al . hence the central single - peak may be formed after the twin - peaks have developed . the above mechanism could work even if a galaxy shows no prominent bar in the optical / infrared photographs , because even a very weak bisymmetric distortion of a disk potential can cause non - circular motion of gas ( koda & wada 2003 ) . also , onodera et al . ( 2003 : private communication ) discuss another possible mechanism to produce a central single - peak by stellar spiral arms for the case of ngc 4501 , which indeed has no bar , but shows continuous spiral structure from the disk to the nucleus . acknowledgements : the observations were performed as a long - term project from 1999 december till 2002 april at the nobeyama radio observatory ( nro ) of the national astronomical observatories of japan . we are indebted to the staff of nro for their help during the observations . we thank t. takamiya and m. hidaka for their help with the observations and reductions , a. kawamura and m. honma for their help with the observations . the data reductions were made using the nrao aips package . we made use of the data archive from nasa extragalactic database ( ned ) . j. k. was financially supported by the japan society for the promotion of science ( jsps ) for young scientists . jogee , s. , baker , a. j. , sakamoto , k. , scoville , n. z. , and kenney , j. d. p. 2001 , in the central kiloparsec of starbursts and agn : the la palma connection , asp conf . j. h. knapen , j. e. beckman , i. shlosman , and t. j. mahoney ( asp , san fransisco ) , 612 ccccccc & & & & frequency & + ngc & config . & year & & ( ghz ) & reference + ( 1 ) & ( 2 ) & ( 3 ) & ra(j2000 ) & dec(j2000 ) & ( 5 ) & ( 6 ) + 4192 & ab+c+d & 2000 - 2002 & 12 13 48.30 & + 14 54 02.9 & 115.350000 & 1 + 4212 & c+d & 2000 - 2001 & 12 15 39.11 & + 13 54 04.8 & 115.286574 & 2 + 4254 & ab+c+d & 2000 & 12 18 50.03 & + 14 24 52.8 & 114.355710 & 3 + 4303 & ab+c+d & 2000 & 12 21 54.87 & + 04 28 24.9 & 114.659250 & 2 + 4402 & ab+c+d & 2001 - 2002 & 12 26 07.06 & + 13 06 45.7 & 115.200000 & 1 + + 4419 & ab+c+d & 2000 - 2002 & 12 26 56.43 & + 15 02 51.1 & 115.286574 & 1 + 4501 & ab+c+d & 2001 - 2002 & 12 31 59.14 & + 14 25 12.9 & 114.407000 & 4 + 4535 & ab+c+d & 2000 - 2001 & 12 34 20.25 & + 08 11 52.2 & 114.507290 & 5 + 4536 & ab+c+d & 2001 & 12 34 27.07 & + 02 11 18.3 & 114.586000 & 5 + 4548 & ab+c+d & 2001 & 12 35 26.40 & + 14 29 47.0 & 115.098000 & 3 + + 4569 & ab+c+d & 2000 - 2002 & 12 36 49.82 & + 13 09 45.8 & 115.286574 & 4 + 4571 & c+d & 2001 - 2002 & 12 36 56.40 & + 14 13 02.0 & 115.138000 & 3 + 4579 & ab+c+d & 2001 - 2002 & 12 37 43.53 & + 11 49 05.4 & 114.710000 & 6 + 4654 & ab+c+d & 2000 - 2002 & 12 43 55.74 & + 13 07 44.2 & 114.811630 & 1 + 4689 & c+d & 2000 - 2002 & 12 47 45.60 & + 13 45 46.0 & 114.659250 & 3 + notes . col.(1 ) : galaxy name . col.(2 ) : nma configulations used for the observations . col.(3 ) : year of the observations . col.(4 ) : pointing position as well as the phase tracking center . col.(5 ) : observing frequency . col.(6 ) : references for positions : ( 1 ) condon et al . 1990 ; ( 2 ) saikia et al . 1994 ; ( 3 ) ned ; ( 4 ) sakamoto et al . 1999 ; ( 5 ) hummel et al . 1987 ; ( 6 ) kohno et al . 1999 ccccccccccccc & & & & & & @xmath81 for @xmath82 & @xmath83 + ngc & & & & @xmath84 & & ( k ) & ( % ) + ( 1 ) & ( @xmath85 ) & ( @xmath85 ) & ( @xmath86 ) & & ( 3 ) & ( 4 ) & ( 5 ) & ( mjy / beam ) & ( mk ) & ( 7 ) & ( 8) + 4192 & 2.4 & 1.9 & 158 & & 20.8 & 20.8 & 23 & 18 & 363 & 20.2 & 47 + 4212 & 4.0 & 3.7 & 149 & & 20.8 & 10.4 & 22 & 28 & 174 & 6.2 & 76 + 4254 & 3.0 & 2.3 & 148 & & 20.8 & 10.4 & 22 & 25 & 333 & 13.3 & 102 + 4303 & 2.8 & 1.9 & 27 & & 10.4 & 10.4 & 17 & 21 & 363 & 17.3 & 97 + 4402 & 2.8 & 2.3 & 166 & & 20.8 & 10.4 & 20 & 26 & 371 & 14.3 & 106 + + 4419 & 3.5 & 2.7 & 159 & & 10.4 & 10.4 & 38 & 26 & 253 & 9.7 & 83 + 4501 & 5.6 & 3.7 & 160 & & 10.4 & 10.4 & 46 & 17 & 75 & 4.4 & 85 + 4535 & 3.1 & 2.6 & 164 & & 10.4 & 10.4 & 23 & 21 & 240 & 11.4 & 208 + 4536 & 2.5 & 1.8 & 173 & & 10.4 & 10.4 & 34 & 17 & 347 & 20.4 & 68 + 4548 & 2.6 & 2.0 & 154 & & 31.2 & 15.6 & 19 & 22 & 389 & 17.7 & 16 + + 4569 & 4.5 & 3.1 & 146 & & 10.4 & 10.4 & 37 & 25 & 165 & 6.6 & 111 + 4571 & 3.8 & 2.5 & 154 & & - & - & 0 & 33 & 319 & 9.7 & + 4579 & 4.5 & 3.5 & 152 & & 20.8 & 10.4 & & 22 & 128 & 5.8 & 31 + 4654 & 5.2 & 3.7 & 149 & & 20.8 & 10.4 & 23 & 22 & 105 & 4.8 & 142 + 4689 & 5.2 & 4.2 & 134 & & 20.4 & 10.4 & 17 & 20 & 84 & 4.2 & 68 + notes . col.(1 ) : galaxy name . col.(2 ) : major and minor axis sizes and position angle for the synthesized beam . col.(3 ) and ( 4 ) : integrated velocity width of a channel , and sampling width between channels of the cube . col.(5 ) : number of channels where emission are detected . col.(6 ) : rms noise scaled for a @xmath87 channel . col.(7 ) : equivalent antenna temprature for @xmath88 . col.(8 ) : fraction of single - dish flux recovered by the aperture synthesis observations . single - dish data are from kenney & young ( 1988 ) . cccc & & + ngc & & reference + ( 1 ) & ra(j2000 ) & dec(j2000 ) & ( 3 ) + 4192 & 12 13 48.29 & + 14 54 01.9 & 1 + 4212 & 12 15 39.40 & + 13 54 04.6 & 2 + 4254 & 12 18 49.61 & + 14 24 59.6 & 1 + 4303 & 12 21 54.94 & + 04 28 25.6 & 1 + 4402 & 12 26 07.45 & + 13 06 44.7 & 1 + + 4419 & 12 26 56.40 & + 15 02 50.2 & 1 + 4501 & 12 31 59.12 & + 14 25 13.3 & 1 + 4535 & 12 34 20.35 & + 08 11 52.2 & 1 + 4536 & 12 34 27.08 & + 02 11 17.1 & 1 + 4548 & 12 35 26.44 & + 14 29 47.4 & 1 + + 4569 & 12 36 49.82 & + 13 09 45.8 & 3 + 4579 & 12 37 43.53 & + 11 49 05.5 & 4 + 4654 & 12 43 56.67 & + 13 07 36.1 & 1 + 4689 & 12 50 15.86 & + 13 29 27.4 & 1 + notes . col.(1 ) : galaxy name . col.(2 ) : adopted central position . col.(3 ) : reference of the position : ( 1 ) this study . dynamical center derived with the aips / gal package ; ( 2 ) cotton et al . 1999 ; ( 3 ) sakamoto et al . 1999 ; ( 4 ) ho & ulvestad 2001 . cccc + ngc & @xmath89 & @xmath90 & @xmath91 + & [ k ] & [ @xmath92 & [ @xmath92 + + 4569 & 3.54 & 586 & 266 + 4419 & 5.71 & 585 & 229 + 4536 & 5.87 & 575 & 224 + 4303 & 6.48 & 436 & 395 + 4535 & 5.17 & 377 & 275 + + 4192 & 3.26 & 348 & 96 + 4402 & 4.06 & 224 & 58 + 4501 & 1.14 & 210 & 111 + 4548 & 1.46 & 123 & 98 + 4212 & 1.38 & 115 & 78 + + 4254 & 3.88 & 110 & 97 + 4579 & 0.78 & 84 & 67 + 4654 & 1.15 & 44 & 27 + 4689 & 0.62 & 24 & 21 + fig . 1 . atlas of the observed virgo galaxies . top top - left panels show dss second generation blue optical images , each for @xmath35 area . @xmath9 observations were obtained for the central @xmath94 regions . bottom panels show observed co intensity distributions ( left ) and corresponding velocity fields ( center ) . position - velocity diagrams along the major axes are shown in the top right panels . parameters of the galaxies and displayed areas are indicated at the bottom for each galaxy . indicated ra and dec are in j2000 . the north to the top , and the east to the left . v fig . integrated @xmath9-intensity maps of the observed galaxies in the same angular scale . the image sizes are @xmath41 , or 4.68 kpc @xmath122 kpc for an assumed distance of 16.1 mpc . the contours are drawn at 5 , 10 , 20 , 40 , 80 , 160 k @xmath44 . 3 . co line velocity fields of the observed galaxies in the same angular scale corresponding to figure 2 . contours are drawn every 20 @xmath44 relative to the systemic velocity , which is expressed by white thick contours . darker coding represents redshift , and white for blueshift . channel maps of the @xmath127co ( @xmath128 ) line emission of the virgo galaxies . intensity scale is in kelvin of brightness temperature . contours are drawn at @xmath129 times the lowest - contour value ( @xmath130 ) . the lowest contour level ( cl ) is indicated for individual galaxies .	we present the results of the virgo high - resolution co survey ( vics ) obtained with the nobeyama millimeter - wave array ( nma ) . this survey was made in the course of a long - term project at nobeyama from 1999 december through 2002 april . the objects were selected from virgo cluster members , considering co richness from single dish flux , mild inclination , and lack of strong tidal perturbations . the central @xmath0 regions ( @xmath1 kpc ) of 15 spiral galaxies were observed with resolutions of @xmath2 and @xmath3 , and sensitivities of @xmath4 for a @xmath5 channel . the objects lie at the same distance of the virgo cluster ( 16.1 mpc ) , which is advantageous for comparisons among individual galaxies . we describe the details of observations and data reduction , and present an atlas of integrated co intensity maps , velocity fields and position - velocity diagrams along the major axes . the molecular gas morphology in the virgo galaxies shows a wealth of variety , not specifically depending on the hubble types . several galaxies show strong concentration of gas in the central few kpc region , where the co morphology shows either `` single - peak '' or `` twin - peaks '' . morphology of more extended co components can be classified into `` arm - type '' , `` bar - type '' , and `` amorphous - type '' . * key words * : galaxies : spiral galaxies : ism galaxies : structure galaxies : virgo = -5 mm = -5 mm = -10 mm
active galactic nuclei ( agns ) are believed to be highly obscured by dusty dense gas , and this contributes to huge radiative energy in some ultra - luminous infrared galaxies ( ulirg ) @xcite . the central engines of type-2 seyfert galaxies are also likely obscured by optically thick molecular gas , for which a torus - like geometry is often assumed . however , detailed structures of the ism in deep centers of ulirgs and seyfert galaxies are still unclear . there are some indirect ways to infer their geometry and distribution using x - ray spectroscopy ( e.g. , @xcite ) , optical / infrared spectral energy distribution ( sed ) in comparison with model sed assuming the geometry of the torus ( e.g. , @xcite ) , and optical / infrared absorption lines ( e.g. , @xcite ) . lutz et al . ( 2004 ) searched the 4.7 @xmath3 m fundamental ro - vibrational band of co in 31 bright local agns , but found no clear signature of absorption features , even in a compton - thick seyfert 2 nucleus , like ngc 1068 . recently , the _ spitzer _ infrared spectrograph ( irs ) revealed vibration - rotation absorption bands of gaseous c@xmath4h@xmath4 , hcn , and co@xmath4 as well as silicate absorption toward deeply obscured ( u)lirg nuclei @xcite . the absorption lines of c@xmath4h@xmath4 and hcn suggest the presence of warm ( @xmath5 k ) and dense ( @xmath6 @xmath2 ) gas @xcite . they suggest that this gas occupies only a small fraction of the nuclear region ( @xmath7 0.01 pc ) near the intrinsic mid - infrared source . the co absorption features observed in ulirgs are not resolved into individual lines , except iras 08572 + 3915 , therefore the kinematics and structure of the absorbed material are still an open question . @xcite have found a broad co absorption line toward iras 08572 + 3915 nw using ukirt / cgs4 . the observations revealed the following features : 1 ) there are several blue - shifted components with @xmath8 km s@xmath9 for the high @xmath10 lines ( @xmath11 ) , and @xmath12 km s@xmath9 and @xmath13 km s@xmath9 and for the low @xmath10 lines ( @xmath14 and @xmath15 ) . 2 ) the hydrogen column density is approximately @xmath16 @xmath17 . 3 ) mean temperature of the warm blue - shifted component is about 200 k. @xcite have also detected the co absorption in iras 08572 + 3915 nw using subaru / ircs . their results are consistent with those of @xcite , showing a blue - shifted component ( -160 km s@xmath9 ) with the temperature @xmath18 k and cold gas at @xmath19 k at the systematic velocity , assuming lte . they also claim that there is a red - shifted component ( + 100 km s@xmath9 ) seen in high @xmath10 ( @xmath20 ) lines , implying higher temperature gas ( @xmath21 k ) . column density of the warm component is @xmath22 @xmath17 . these observations suggest that the absorption features do not simply arise from a smooth rotating molecular torus , which has been often postulated to explain the type-1 and type-2 agns . however , it is hard to determine the geometry and internal structures of the ism in the central region of the galaxy from this observational information alone . theoretical models used for sed fitting are phenomenological without kinematical information , therefore we can not compare them with the absorption line observations . in this paper , we investigate how the observational features suggested by the co absorption in iras-08572 + 3915 nw can be understood in the context of a three - dimensional hydrodynamic model of the ism around a supermassive black hole ( smbh ) with nuclear starbursts @xcite ( hereafter wn02 ) . this is currently a unique model of the obscuring material around a smbh on a several tens pc scale , characterized by a highly inhomogeneous , multi - phase , and turbulent ism with a globally stable , geometrically thick structure . three - dimensional radiative transfer calculations for this model revealed that co luminosity distribution is also highly non - uniform @xcite . although this is not a confirmed theoretical model applicable to the obscuring matter in all types of agns , it is worth verifying whether the features observed in the co absorption lines in the ulirg can be explained by the full 3-d hydrodynamic models . in the wn02 model , mass , momentum , and energy conservation equations with the poisson equation are numerically solved with energy feedback from supernovae in a fixed gravitational potential . a rotating gas disk in a time - independent spherical potential is solved by 3-d hydrodynamic code . the mass of the bh is @xmath23 . we also assume a cooling function @xmath24 @xmath25 with solar metallicity , heating due to photoelectric heating , and energy feedback from sne . we assume a uniform uv radiation field ten times larger than the local uv field . the hydrodynamic part of the basic equations is solved by ausm ( advection upstream splitting method ) @xcite . ( see details in @xcite ) . we use @xmath26 cartesian grid points covering a @xmath27 pc@xmath28 around the galactic center ( i.e. the spatial resolution is 0.25 pc ) . the poisson equation is solved using the fast fourier transform and the convolution method . the initial condition is an axisymmetric and rotationally supported thin disk ( the scale height is 2.5 pc ) with a uniform radial density profile and a total gas mass of @xmath29 . random density and temperature fluctuations , which are less than 1 % of the unperturbed values , are added to the initial disk . supernova ( sn ) explosions are assumed to occur at random positions on the disk plane . the average sn rate is @xmath30 0.8 yr@xmath9 . the energy of @xmath31 ergs is instantaneously injected into a single cell as thermal energy . thus the three - dimensional evolution of blast waves driven by the sne in an inhomogeneous and non - stationary medium with a global rotation is followed explicitly , taking into account the radiative cooling . wn02 showed that a globally stable concave `` torus '' , in which the gas is highly turbulent and inhomogeneous , is formed ( see also figure 6 ) . figure [ fig : phase ] is a phase - diagram of the ism in the torus at a quasi - steady state . three dominant phases present : hot gases around @xmath32 k caused by sne , warm gases around @xmath33 k , and cold , dense gases at @xmath34 k. in terms of the observed co absorption in iras 08572 + 3915 , we are interested in kinematics and spatial location of the cold , dense media . ( 80mm,80mm)figure1.eps the size of the present torus model ( @xmath35 pc ) is much larger than the size ( @xmath36 pc ) suggested by near- and mid - infrared high - resolution observations of type-2 seyfert , e.g. ngc 1068 @xcite . this difference actually reflects a fact that near / mid - infrared flux is originated from inner a few pc region of the dusty material around the agn . in fact , a recent 3-d radiative transfer model assuming clumpy dust tori for ngc 1068 @xcite suggests that the outer radius of the torus that fits sed of ngc 1068 is 56 pc , while their h- , k- , and n - band images show their radii are 2.0 - 2.7 pc . in terms of ulirgs , @xcite suggested observed deep absorption features in mid - infrared indicate obscuration on scales of a few 100 pc . probably the obscuring material in the central region of agns and ulirgs is extended from a sub - pc to a sub - kpc scales with various physical / chemical conditions and structures . in order to find the blue- and red - shifted warm , dense gases suggested by the co absorption line in the hydrodynamic model , we sample points with the following conditions from the 2000 randomly selected points in a half ( i.e. @xmath37 ) the `` torus '' . 1 . number density is @xmath38 @xmath2 2 . column density from the nuclear source to the observer @xmath39 @xmath17 3 . temperature is @xmath40 k ( 80mm,80mm)figure2.eps figure [ fig : density - temp ] shows number density and temperature of the selected points . it is clear that two components exist : cold gas with @xmath41 k and warm gas with @xmath42 k. figure [ fig : vlos - temp ] shows the line - of - sight velocity of the selected points , @xmath43 , as a function of their temperature . most of the cold and warm components are distributed in the range of @xmath44 km s@xmath9 . spatial distribution of the selected points are shown in figures [ fig : r - temp ] and [ fig : theta - temp ] . the warm components are distributed around @xmath45 pc from the galactic center , and they are located just above the equatorial plane , i.e. @xmath46 . temperature of the warm component is higher for larger @xmath47 , i.e. closer to the surface of the `` torus '' . the cold dense gases are located @xmath48 pc and near the disk plane ( @xmath49 ) . ( 80mm,80mm)figure3.eps ( 80mm,80mm)figure4.eps ( 80mm,80mm)figure5.eps ( 80mm,80mm)figure6.eps ( 100mm,100mm)figure6b.eps in summary , the absorbed material with @xmath50 @xmath17 , @xmath51 @xmath2 , @xmath52 km s@xmath9 , and @xmath53 k sits at the innermost region of the inhomogeneous torus as shown in figure [ fig : torus ] ( the hatched regions ) . for the gas near the equatorial plane , the column density is too high to be observed as an absorption line for the continuum ir source . therefore , if the ism in the central region of ulirgs , like iras 08572 + 3915 , is similar to the structure shown in figure [ fig : torus ] , the viewing angle to observe the warm and dense gases as absorption is restricted . if this is the case , we can expect to observe high velocity , warm components as well as low velocity , cold components as absorption against the background continuum source , whose size is smaller than @xmath54 pc . the high density gases in those regions have various velocity components , as shown in figure [ fig : vlos - temp ] . thus it is not necessary to observe only blue - shifted gas . in fact , @xcite also found a red - shifted gas component ( + 100 km s@xmath9 ) in high-@xmath10 transition , suggesting gas warmer than the blue - shifted component . one should note that geometry of the `` torus '' in the present model depends on the supernovae rate , total gas mass , and the mass of a smbh , since it is determined from a balance between energy input and dissipation rates under the effect of gravitational potential ( wn02 ) . therefore , the absolute locations ( i.e. @xmath55 and @xmath56 ) of the warm , dense gas around an agn should depend on these parameters . in the present model of the torus , the high density ( @xmath57 @xmath2 ) gases have both red- and blue - shifted velocity for a line - of - sight . since a volume fraction of these dense gases is expected to be very small ( see discussion below ) , and the background continuum is emitted from a small region in the vicinity of the central engine , the chance to observe a dense blob with a particular velocity component by an absorption line should be very small . for example , as shown in figure [ fig : vlos - temp ] , the blue - shifted component around @xmath58 km s@xmath9 exists in about 1/500 of the total volume of the torus whose radius is about 30 pc . therefore , one possible explanation about the observed blue - shifted component is that the line - of - sight toward the agn happens to graze the high - density clump which is approaching us . if this is the case , since the motion in the torus is highly turbulent ( wn02 ) , we have nearly equal chance of observing red - shifted warm gas , too . the red - shifted warm gas found by @xcite may be explained by this picture , but observed samples are still too small to further compare with the model . recently @xcite proposed a simple statistical theory of the structure of the inhomogeneous gas disk in galactic disks , which is caused by non - linear development of gravitational and thermal instabilities . using three - dimensional hydrodynamic simulations , they show that the probability distribution function ( pdf ) of density @xmath59 in a globally stable , inhomogeneous ism is well represented by a single log - normal function over a wide density range , that is @xmath60 } d \ln \rho,\end{aligned}\ ] ] where @xmath61 is the characteristic density and @xmath62 is the dispersion . the simulations show that the dispersion of the log - normal pdf is larger for more massive systems . using the pdf , it is straightforward to calculate a volume fraction of high density gas ( @xmath63 ) above a given critical density ( @xmath64 ) : @xmath65\right)\;. \label{eq : fc}\end{aligned}\ ] ] the dispersion @xmath62 is then related to the average density , @xmath66 , i.e. @xmath67 it is also suggested that characteristic density @xmath61 is not very sensitive for changing the total gas mass . suppose the column density toward the continuum source is @xmath68 @xmath17 and the size of the ` torus ' is 30 pc , the average density would be @xmath69 @xmath2 . therefore , using equation ( [ eq : sigma ] ) , we have @xmath70 for @xmath71 @xmath2 . this leads to @xmath72 and @xmath73 for @xmath74 @xmath2 and @xmath75 @xmath2 , respectively . with this small volume - filling factor of the high - density gas in an obscuring material , it might be reasonable to expect there is a small chance of detecting a high - density clump with a particular velocity by absorption lines . @xcite observed 196 agns and ulirgs using _ spitzer_/irs , and found that quasars are characterized by silicate features in emission and seyfert2s are dominated by weak silicate absorption , while ulirgs are characterized by strong silicate absorption . @xcite suggest that this difference in near - infrared spectra reflects a different geometry of the dusty material . they claim that clouds dominate the obscuration of seyfert 1s and 2s , while on the other hand , central sources of ulirgs are obscured mainly by smooth , geometrically and optically thick material with a steep temperature gradient . according to @xcite , @xmath76 for the nucleus of iras 08572@xmath773915 estimated by infrared spectra obtained by _ spitzer _ is @xmath78 @xmath17 . if the scale of the obscuring material is about 30 ( 100 ) pc , the average number density is @xmath79 @xmath2 . using the log - normal density pdf in 3.1 , we can estimate that the volume filling factor of those gases is about 20 ( 30 ) % , suggesting that the dusty gas that contributes to the silicate absorption is more smoothly distributed in the obscuring material than the warm , dense gas found by co absorption . our result presented here is that at least the co absorbed feature of iras 08572 + 3915 is not inconsistent with a model characterizing a dense clumpy medium around an agn . since the velocity field of the ism is highly turbulent in the model , if individual high density clumps are not resolved , a broad absorption feature originating in dense gases with various velocity components can be expected in the obscuring material . if this is the case , the reason why co absorption is not observed in seyfert-2 @xcite can be understood by the low filling factor of the warm , dense gas in the obscuring material . as discussed in 3.1 , if average gas density of the obscuring material around agns is smaller in seyfert galaxies than in ulirgs , a fraction of high - density clumps becomes small . this seems to be inconsistent with the result by @xcite , but probably dusty `` clouds '' in the obscuring material responsible for the near / mid - infrared sed are much more diffuse than those causing the co absorption ( see discussion above ) . the ism around agns in ulirgs and seyfert galaxies are more or less inhomogeneous , and if so its density pdf is expected to be approximately log - normal . therefore , it is not surprising , if the `` smooth '' distribution of dusty gas that is preferable to the sed fitting would include higher - density clumps with a smaller volume - filling factor . in the present hydrodynamic model , heating due to the radiation from the agn to dust is not taken into account . temperatures of optically thick clouds at a distance of a few pc from an agn with @xmath80would be 300 - 1200 k @xcite . therefore the warm clouds near the surface of the clumpy `` torus '' might be warmer than @xmath81 k , which is a result of the present model . effects of x - rays are not included in the model either , which would be especially important for chemistry of molecules @xcite . for highly inhomogeneous media like those seen in figure [ fig : torus ] , hard x - rays could affect not only chemistry of the gas in the vicinity of the central source , but also the clouds inside the `` torus '' , because the column density and the ionization rate are not simply related to the distance from the central source ( wn02 ) . the effects of x - rays on the abundance and line intensities of co and other molecules are investigated using the present hydrodynamic model and 3-d non - lte radiative transfer calculations ( yamada , wada , & tomisaka , in preparation ) . as an alternative interpretation of the blue - shifted warm gas could be originated in outflows from an accretion disk around a supermassive black hole . using two - dimensional radiation - hydrodynamic simulations , @xcite found that an outflow is driven by radiation force due to a luminous accretion disk . however , the outflow velocity is extremely large ( @xmath82 ) , and its temperature is @xmath83 k , therefore the outflow from the accretion disk itself does not explain the warm , blue - shifted gases discussed here . it might be possible however that , the hot outflow gas propagates outward , and it then cools and forms warm , dense cloudlets by the thermal instability at a few pc from the accretion disk . unfortunately long - term evolution of the outflow from the accretion disk and its interaction with the ism on several tens pc scale is still an open question . based on the analysis presented here , we could conclude that the blue - shifted , warm , dense molecular gas in iras 08752 + 3915 nw suggested by recent high - resolution observations by subaru / ircs can be caused by high - density clumps in the inhomogeous inter stellar matter around a supermassive black hole . the small possibility of observing the co absorption in ulirgs and seyfert 2 galaxies could be simply interpreted as indicating that the high density regions occupy a very small volume fraction ( @xmath84 ) in the obscuring material with a turbulent velocity field . this is naturally expected if the inhomogeneous material is caused by non - linear development of instabilities and random processes ( e.g. stochastic explosions of supernovae and interaction with inhomogeneous ism ) in a globally quasi - stable system . however , this model is still a presumption until the absorbed features in near infrared spectra of many ulirgs and seyfert 2 galaxies are resolved and details of the kinematics of the obscuring material are clarified by future observations . it is also necessary to improve numerical models of the ism in the central region of active galaxies by taking into account more realistic treatment of chemical reactions of molecular gas and radiative transfer in terms of agns and star forming regions . the author is grateful to takao nakagawa , mai shirahata , and masa imanishi for fruitful discussions . the anonymous referee s comments were also helpful . numerical computations were carried out on a fujitsu vpp5000 at naoj . kw is supported by grant - in - aids for scientific research [ no . 16204012 ] of jsps . elitzur , m. 2006 , new astronomy review , 50 , 728 fritz , j. , franceschini , a. , & hatziminaoglou , e. 2006 , mnras , 366 , 767 geballe , t. r. , goto , m. , usuda , t. , oka , t. , & mccall , b. j. 2006 , , 644 , 907 hao , l. , weedman , d. w. , spoon , h. w. w. , marshall , j. a. , levenson , n. a. , elitzur , m. , & houck , j. r. 2007 , , 655 , l77 hnig , s. f. , beckert , t. , ohnaka , k. , & weigelt , g. 2006 , , 452 , 459 imanishi , m. , & dudley , c. c. 2000 , , 545 , 701 imanishi , m. , dudley , c. c. , & maloney , p. r. 2006 , , 637 , 114 imanishi , m. , et al . apj in press ( astro - ph/0702136 ) , 404 , 495 jaffe , w. , et al . 2004 , , 429 , 47 lahuis , f. , et al . 2007 , , 659 , 296 levenson , n. a. , weaver , k. a. , & heckman , t. m. 2001 , , 550 , 230 levenson , n. a. , sirocky , m. m. , hao , l. , spoon , h. w. w. , marshall , j. a. , elitzur , m. , & houck , j. r. 2007 , , 654 , l45 liou , m. , steffen , c. , 1993 , j.comp.phys . , 107,23 lutz , d. et al . , 426 , l5 maloney , p. r. , hollenbach , d. j. , & tielens , a. g. g. m. 1996 , , 466 , 561 meijerink , r. , spaans , m. , & israel , f. p. 2007 , , 461 , 793 ohsuga , k. 2006 , , 659 , 205 pier , e. a. , & krolik , j. h. 1993 , , 418 , 673 shirahata , m. , nakagawa , t. , goto , m. , usuda , t. , suto , h. , & geballe , t.r . , 2007 , submitted to apj spoon , h. w. w. , et al . 2004 , , 154 , 184 veilleux , s. , sanders , d. b. , & kim , d .- c . 1999 , , 522 , 139 wada , k. 2001 , , 559 , l41 wada , k. & norman , c. a. 2001 , , 547 , 172 wada , k. & norman , c. a. 2002 , , 566 , l21 ( wn02 ) wada , k. & norman , c. a. 2007 , , 660 no.2 , in press ( astro - ph/0701595 ) wada , k. & tomisaka , k. 2005 , , 619 , 93 meurer , g. , & norman , c. a. 2002 , , 577 , 197 wittkowski , m. , kervella , p. , arsenault , r. , paresce , f. , beckert , t. , & weigelt , g. 2004 , , 418 , l39	possible origins of the molecular absorption discovered in some ulirgs are investigated , based on a 3-d hydrodynamic model of star - forming interstellar gas in a galactic central region . the blue - shifted , warm ( @xmath0 k ) , dense ( @xmath1 @xmath2 ) molecular gas suggested by co absorption in iras 08752 + 3915 could be caused by the innermost region of the inhomogeneous inter - stellar medium ( ism ) around a supermassive black hole . the infrequent observations of the dense gas with absorption in ulirgs and seyfert 2 galaxies could simply suggest that the high - density regions occupy only a very small volume fraction of the obscuring material . this is naturally expected if the inhomogeneous structure of the ism is caused by non - linear development of instabilities . the model predicts a turbulent velocity field in the obscuring material , therefore blue- and red - shifted gases should be observable with nearly the same probability for the large enough statistical samples .
one of the most remarking properties of quantum mechanics is that the state of a quantum system changes not only via the deterministic evolution given by the schrdinger equation but also when it is measured . although the system can be measured directly , in many situations it is probed by an auxiliary system ( ancilla ) that is then detected , providing information about the system of interest . this is a typical situation in cavity qed where atoms and light interact and the detection of either of them alters the state of the other . in the microwave regime , for example , it is usually the atoms that behave as a probe for the field . this configuration has led to a number of experiments on fundamental aspects of quantum mechanics including the measurement of the decoherence of a cat state @xcite , qnd measurement of single photons @xcite , and the observation of quantum jumps of light @xcite . this property becomes particularly interesting for quantum open systems where the environment that the system is coupled to plays the role of a natural bona fide ancilla . the time evolution of a single quantum system can therefore be probed by directly monitoring this reservoir . this dynamics is going to be stochastic , as the evolution is conditioned on the measurements results , and can be mathematically described in terms of quantum trajectories , which are related to different physical ways to monitor the environment and to extract information about the system . the monitoring of the field leaving a damped cavity mode provides a good example of this stochastic dynamics . for instance , if a photodetector is used to collect the output of the cavity , the dynamics is better described in terms of quantum jumps where each click in the detector corresponds to lowering the number of photons inside the cavity by one . however , a completely different dynamics is found if the same propagating field is combined with a local oscillator in a beam splitter and a homodyne measurement is performed , in which case the time evolution of the damped cavity field is appropriately described by a continuous stochastic trajectory ( quantum state diffusion ) . all these trajectories present very interesting scenarios that allow for the production of non - classical states of the cavity field as well as the protection of the purity of the cavity mode or of the entanglement shared by two or more modes . note that if the lost photons are ignored , or the measurements averaged out , one obtains the usual master equation dynamics for the decoherence process that is responsible , for example , for rapidly turning superpositions of coherent states into mixtures @xcite . for this reason , each monitoring scheme described above is said to represent an unravelling of the master equation in terms of stochastic trajectories . note that the unravellings above represent only a limited set of the possibilities to measure the environment . in fact the master equation can be mathematically unravelled in infinite ways in terms of stochastic trajectories @xcite and therefore one could envisage more general ways of monitoring the environment rather than only simple photodetection and homodyning . this freedom in defining unravellings has been recently explored in the context of entanglement decay and protection @xcite , where monitoring schemes that combine different decay channels play a crucial role on the recovery of the mixed state entanglement dynamics in terms of trajectories @xcite and on the protection of entanglement conditioned on measurement outcomes @xcite . it would be interesting then to propose realistic experimental scenarios where this variety of unravellings could be explored . while the monitoring of the field emitted by a leaky cavity mode or by a decaying atom is within current experimental feasibility , this is not necessarily the most efficient way and certainly not the most complete one to generate different unravellings . this scheme proves to be very limited in two very interesting situations : non - zero reservoir temperatures and unravellings that combine detections and no - detections ( jumps and no - jumps ) . the first case presents two problems : first how to determine when a very large reservoir loses a single photon to the system and second how to distinguish a photon that comes from the system from one that already exists in the reservoir . in the second case , one faces the problem of physically superposing a click with a no - click in the detector . these limitations severely hinder the exploration of quantum trajectories in these systems . even though the direct monitoring of a natural thermal environment remains a challenge , in cavity qed one can artificially engineer this and other reservoirs that could produce different kinds of quantum trajectories when measured . in the microwave regime , for example , beams of atoms crossing the cavity can be used to mimic a thermal dissipative reservoir for the cavity field @xcite . the posterior detection of these atoms produces quantum trajectories for the cavity field , as analyzed in @xcite both in terms of jumps and continuous diffusion processes . however , these schemes also present limitations to the production of different unravellings . for example , while the combination of different channels can be easily accomplished within the optical detection scenario by having the photons for each channel arriving at different ports of a beam splitter @xcite , the situation for atomic detection seems far more complex . in all previous proposals to engineer thermal reservoirs for cavities using atomic beams , decay and excitation channels correspond to two - level atoms entering the cavity either in the ground or in the excited state . a combined detection , which is utterly important for the applications proposed in @xcite , would then require some kind of interaction between the atoms after they cross the cavity , which albeit not impossible , seems rather challenging . in this paper we show that simple modifications to these previous proposals lead to an alternative scheme to engineer a thermal reservoir in the context of microwave cavity qed that accommodates more general detections , hence enabling the simulation of a wide set of different classes of trajectories for a monitored atomic reservoir . in particular , we address the limitations raised in the two previous paragraphs . we then propose a complementary experimental setup where we invert the roles , letting the cavity field play the reservoir for the atoms . we rely on the purcell effect to channel the atomic decays into the modes of a lossy cavity and then use the optical detection proposed in @xcite to monitor the reservoir in different ways . in this last case , we introduce an additional measurement possibility to implement the proposal in @xcite that would circumvent the challenging task of collecting broadly emitted photons from atomic decay . we will start by briefly recalling the usual way to simulate a thermal reservoir for a cavity field using a sequence of two - level atoms as discussed in @xcite . the atoms , prepared in either the ground or excited states , cross a cavity where they interact resonantly via a jaynes - cummings hamiltonian for a short enough time . by short time , we mean a time interval @xmath0 much smaller than the inverse of the vacuum rabi frequency of the system @xmath1 such that the probability of an atom making a transition is small . under this model , one can then show that if the atoms are not detected after interacting with the cavity , the cavity field ( of mode annihilation operator @xmath2 ) will evolve according to a master equation of the form @xmath3 \rho + \gamma_+ \mathcal{d}[a^\dagger ] \rho,\ ] ] where @xmath4 , and the relationship between the rates @xmath5 and @xmath6 is given by the ratio between the flux of atoms initially prepared in the ground ( @xmath7 ) and excited state ( @xmath8 ) . a thermal bath is obtained by setting this ratio to be @xmath9 , where @xmath10 is the average number of photons in the modes with energy @xmath11 in the reservoir . if one now considers that the atoms are detected after the interaction , then the evolution of the cavity field is conditioned on the result of this measurement and better described by quantum trajectories . in this case , if an atom enters in the ground ( excited ) state and is detected in the excited ( ground ) state then the cavity field will be modified by a jump operator @xmath12 ( @xmath13 ) corresponding to the annihilation ( creation ) of a photon in the cavity . note that these jumps will be rare since , under our assumption of small interaction time , the atoms will most probably remain in their original state , in which case the cavity field evolution will correspond to the application of a no - jump operator @xmath14 . if the state of the atom is known before and after the cavity then a quantum jump unravelling corresponding to the operators @xmath15 will be produced for the evolution of the cavity field @xcite . note , as mentioned before , that in order to produce any other jump - like unravelling which necessarily combines the operators @xmath16 and @xmath17 , the atoms would need to interact after going through the cavity which is a rather difficult experimental task . we now present an alternative scheme to engineer a thermal reservoir that naturally allows more general detections . the central idea is to encapsulate in a single atom the interactions with the cavity that correspond to the evolution with @xmath16 and @xmath17 . for that we will use three - level atoms with levels organised in a cascade , as shown in fig . [ fig1 ] , and selected in such a way that the energy difference between the intermediate level ( @xmath18 ) and the lower and upper levels ( respectively @xmath19 and @xmath20 ) are close enough to the energy of the chosen cavity mode so that these transitions can be tuned into resonance through the application of external fields , but only one at a time . the atoms are prepared in the @xmath21 state before entering the cavity and in the first stage , the external field shifts the @xmath18 level such that the transition @xmath22 becomes resonant with the cavity frequency @xmath23 . the system then evolves under the jaynes - cummings hamiltonian @xmath24 for a short time @xmath25 . up to second order on @xmath26 the initial atom - field state @xmath27 will be transformed to @xmath28 where @xmath29 is the coupling constant , @xmath30 the initial field state and the tilde indicates that the state is not normalised . if , now , the other transition @xmath31 is tuned to resonance by changing the external field , the hamiltonian will be @xmath32 and , again , making the short time expansion will lead to @xmath33 considering now a time interval @xmath34 that is large enough so that @xmath35 atoms ( @xmath36 ) cross the cavity and yet the probability of occurrence of a quantum jump remains very small , one can recover the master equation result by tracing out the atomic system . this becomes more evident if one identifies the rates as @xmath37 and @xmath38 , where @xmath39 is the number of atoms entering the cavity per unit of time . therefore , the same atom can emulate at the same time both the dissipative and the excitation reservoirs . transition , then with @xmath40 transition . rotations between levels @xmath41 and @xmath42 after the interaction with the cavity and before atomic detection produces different unravellings of the dynamics in terms of quantum jumps.,width=321 ] if instead of tracing out the atoms one measures their levels , then the evolution of the state of the cavity field obeys a quantum trajectory given by the no - jump operator @xmath43 if the atom is detected in level @xmath18 or the jump @xmath44 ( @xmath45 ) if the atom is detected in level @xmath20 ( @xmath19 ) , that corresponds to the usual monitoring discussed previously . the advantage of encoding both jumps in the same atom is that it allows us to explore different unravellings by introducing simple modifications in the detection scheme . if one makes a unitary transformation between levels @xmath46 and @xmath47 after the atoms cross the cavity , for example , then a detection of each state will correspond to a different unravelling in which the jump operators are combinations of @xmath2 and @xmath48 . for instance , in the particular case that @xmath49 , a @xmath50 pulse between levels @xmath19 and @xmath20 produces jump operators proportional to the field quadratures @xmath51 and @xmath52 where the phase @xmath53 is given by the chosen rotation . in fact , since one can apply unitary operations also involving level @xmath18 then it is clear that any unravelling of the type @xmath54 , where @xmath55 is a unitary matrix acting on the vectorial space of the jumps and @xmath56 is the vector @xmath57 , can be produced . note that the unitary @xmath55 is obtained by individually rotating the atoms after they have already interacted with the cavity mode therefore preserving the master equation structure when the atoms are ignored . in the scheme described above , the detection of the atom modifies the quadratures of the field affecting , in principle , all fock states . for example , if the field is initially constrained to fock states @xmath58 , the detection of the atom in its middle ( @xmath18 ) or highest ( @xmath20 ) levels will preserve the initial subspace , but the detection of the atom in its ground state will expand the effective subspace of the cavity mode to include fock state @xmath59 . this can be avoided by choosing a selective interaction @xcite for the @xmath60 transition in which case the creation of an extra photon in the cavity can be set to occur only if the cavity is empty , hence preserving the initial effective subspace and thus fully mimicking a spin - half system for the cavity field @xcite . inverting the roles of the previous session , now , the three level atom is stationary and interacts with two orthogonal cavity modes such that the @xmath61 ( @xmath60 ) transition generates photons circularly polarized to the right ( left ) associated to the annihilation operator @xmath62 ( @xmath63 ) of the cavity mode . we assume a very lossy cavity ( of decay rate @xmath64 ) made of a nearly perfect mirror and a semi - transparent one ( see fig . [ fig2 ] ) . if the rabi frequencies of each transition are large enough then the purcell effect will channel the atomic decay in the cavity modes , and the cavity asymmetry will guarantee that the photons will always leak in a well defined direction . each atomic transition has to be coupled to its respective cavity mode with different coupling constants @xmath65 and @xmath66 , and a @xmath67-polarized classical field of strength @xmath68 has to drive the @xmath69 transition as shown in fig . [ fig2 ] . the relation between the involved rates should be @xmath70 , where @xmath71 and @xmath72 are the natural atomic linewidths . for the scheme to work , much greater typically reads @xmath73 where this particular choice has already been made to fit the analysis that follows . and @xmath65 ) and the cavity decay ( @xmath64 ) will induce an effective atomic decay via the purcell effect . photons from the right and left polarized modes leaving the cavity will then correspond to @xmath74 and @xmath75 transitions , respectively . the corresponding effective decay rates @xmath76 and @xmath77 are assumed to be much larger than the natural spontaneous decay rates @xmath71 and @xmath72.,width=321 ] the overall master equation that describes the dynamics of the whole system is given by : @xmath78 + \gamma \mathcal{d}[\sigma_-^{(eg)}]\rho + \gamma \mathcal{d}[\sigma_-^{(ie)}]\rho + \kappa \mathcal{d}[a],\end{aligned}\ ] ] where @xmath79 with @xmath80 and @xmath81 being the lowering and raising operators for the @xmath82 transition . being @xmath83 there is no rabi oscillation between the corresponding atomic transition and the cavity mode which acts just as a channel for the atomic decay . in other words , as soon as a photon is transferred from the atom to the cavity mode it leaks through the semi - transparent mirror . the effective atomic decay rate is then given by @xmath84 as long as @xmath85 . much in the same way , the effective decay rate for the @xmath61 transition is given by @xmath86 . now , if the external driving field @xmath68 is still much smaller than this effective decay rate , then level @xmath20 can be adiabatically eliminated and the combined effect of the classical field and the strong decay rate @xmath77 will be to generate an effective incoherent pump between levels @xmath19 and @xmath18 of rate @xmath87 . in this way , the scheme produces a decay and an excitation reservoir for the qubit composed of levels @xmath88 and @xmath42 with rates @xmath89 and @xmath90 , respectively . in particular , when the effective rates @xmath91 are the same , _ i.e. _ @xmath92 , one obtains an infinite temperature environment . note that there is still another degree of freedom available if the conditions are not perfectly matched that is the eventual detuning between cavity modes and atomic transitions . for example , let us say that @xmath93 . if @xmath94 , then the effective decay rate is corrected by @xmath95 @xcite . under the above approximations and assuming that the effective rates are much larger than the natural atomic linewidths ( @xmath96 ) , the unconditioned dynamics of the ( two - level ) atom can be finally written as : @xmath97\rho + \gamma_+ \mathcal{d}[\sigma_+]\rho,\end{aligned}\ ] ] where the superscript @xmath98 has been dropped for simplicity . if the photons leaving the cavity are detected , then the atomic dynamics will follow a stochastic dynamics conditioned on the measurement outcomes as described by the quantum trajectory approach . equivalently to the previous session , different detection techniques will lead to different unravellings . for example , if a @xmath99 plate is placed right after the leaking mirror , then each circular polarization is converted into orthogonal linear polarizations , @xmath100 ( @xmath101 ) . these photons are then sent into a polarized beam splitter that separates the linear polarizations into two different propagation modes where detectors are placed to collect them . in this way , the click in the @xmath102 ( @xmath103 ) path identifies a transition @xmath22 ( @xmath60 ) which corresponds to the `` usual '' photodetection unravelling with the jumps @xmath104 and @xmath105 applied to the atomic state . if no photon is detected in the @xmath106 interval , then the no - jump @xmath107 operator is applied to the system . a more interesting situation occurs if the @xmath99 plate is taken out of the setup . in this case , the photons propagating after the pbs will correspond to linear combinations of the photons leaking from the atom , implementing an unravelling with the jumps @xmath108 and @xmath109 ( similar to the quadrature unravelling described in the previous section ) . we have shown in @xcite that if two qubits initially share an entangled state , then this kind of monitoring performed on both subsystems can preserve the entanglement in the system . the above setup is therefore a way to implement the entanglement protection ideas proposed in @xcite . before concluding , let us just remark that the scheme here described for cavity qed could also be applied to other experiments involving three level systems and harmonic oscillators . natural candidates where quantum jumps have been recently observed or proposed are superconducting qubits @xcite and nanoresonators @xcite . finally , in conclusion , we have proposed two experimental ways to produce a wide range of unravellings of a master equation evolution in cavity qed systems . in the first case , three - level atoms are used to simulate a thermal dissipative reservoir and the preparation and posterior detection of these atoms in different basis produce unravellings that correspond to different combinations of the natural jump operators and even between these and the no - jump one . this is a particular case in which `` no - click '' and `` click '' can be physically combined to generate a new class of unravellings . later , we also show how to invert the roles and use the purcell effect to channel the detection of the spontaneous decay and of the incoherent pump ( or even their combination ) for driven three - level atoms . this work adds to the existing class of proposed cavity qed experiments , expanding the range of quantum effects that can be explored in this experimentally successful system . we would like to acknowledge the support from center for quantum technologies at the national university of singapore . arrc acknowledges financial support by the australian research council centre of excellence program and mfs acknowledges the support of the national research foundation and the ministry of education of singapore . englert b .- g . and morigi g. , _ five lectures on dissipative master equations _ in _ coherent evolution in noisy environments _ , edited by buchleitner a. and hornberger k. , vol . 611 of _ lecture notes in physics _ ( springer berlin / heidelberg ) 2002 pp .	the experimental observation of quantum jumps is an example of single open quantum systems that , when monitored , evolve in terms of stochastic trajectories conditioned on measurements results . here we present a proposal that allows the experimental observation of a much larger class of quantum trajectories in cavity qed systems . in particular , our scheme allows for the monitoring of engineered thermal baths that are crucial for recent proposals for probing entanglement decay and also for entanglement protection . the scheme relies on the interaction of a three - level atom and a cavity mode that interchangeably play the roles of system and probe . if the atom is detected the evolution of the cavity fields follows quantum trajectories and vice - versa .
dirac invented his relativistic equation in order to explain the quantum properties of the electron ( spin @xmath4 ) in the relativistic framework : the equation had to ( _ a _ ) display the formula @xmath5 as the eigenenergy of a particle in free motion ( with @xmath6 ) ; ( _ b _ ) be covariant under a lorentz transformation that links the particle dynamical properties between two inertial frames . dirac found that the sound equation had to be expressed , necessarily , in terms of @xmath7 matrices . its more familiar form is @xmath8 , with the hamiltonian being linear in the momentum @xmath9 , @xmath10 and the @xmath7 matrices @xmath11 and @xmath12 , have to satisfy forcefully the relations @xmath13 with @xmath14 for the unit matrix ( in dirac s book @xcite , instead of @xmath15 we find a @xmath7 matrix @xmath16 multiplying the @xmath7 direct product of pauli matrices @xmath17 ) . an usual approach consists in the introduction of the chiral representation , where the components of the matrix vector , @xmath18 and @xmath19 , are respectively in contravariant and covariant forms , in the same fashion that one has @xmath20 and @xmath21 @xcite . the state vector solution to the dirac equation can be written as the sum , @xmath22 of left and right chiral spinors @xmath23 and @xmath24 , @xmath25 . from dirac equation plus eq . ( [ hamdirac ] ) and ( [ state1 ] ) one constructs two coupled differential equations for the spinors @xmath26 and @xmath27 , @xmath28 whose lagrangian is @xcite ( omitting the explicit dependence on @xmath29 ) , @xmath30 interestingly , the dirac equation allows a different insight when written in terms of direct ( or kronecker ) products of pauli matrices . so , daring to interpret quantum mechanics as a special kind of information theory for particles and fields @xcite , in the language of quantum information we may say that the relativistic equation of a spin-1/2 fermion has as solution a state of two _ qubits _ ( two degrees of freedom ) carried by a massive particle whose dynamical evolution in space is represented by a continuous variables that may be the position or the linear momentum @xcite . hereon we will choose the linear momentum representation ( as a c - number ) instead of using the position operator @xmath31 , since we are not introducing a position dependent potential in the hamiltonian . one can appreciate that fact by writing the matrices @xmath15 and @xmath32 in terms of tensor products of pauli matrices @xmath33 where the upperscripts 1 and 2 refer to qubits 1 and 2 , respectively . thus we write the dirac hamiltonian ( [ hamdirac ] ) in terms of the direct product of two - qubit operators , @xmath34 , and the two solutions to dirac equation are @xmath35 , \label{sol1}\end{aligned}\ ] ] where @xmath36 and @xmath3 stand respectively for negative and positive energy solutions , @xmath37 , with @xmath38 . the state @xmath39 is a spinor representing the spatial motion of the free fermion ( @xmath40 in the momentum representation ) coupled to its spin , which describes a structureless magnetic dipole moment . for qubit 1 the kets , @xmath41 and @xmath42 , are identified as the intrinsic parity eigenstates of the fermion . the states are orthogonal , @xmath43 . for the inner product we get @xmath44 , with the normalization factor @xmath45 and we also assume that the local probability distribution for the momenta is normalized , @xmath46 . thus the spinors and @xmath7 matrices stand for the the direct product of the intrinsic degrees of freedom of a massive spin-1/2 fermion , parametrized by the linear momentum @xmath9 , on free motion in space . since @xmath47 , one has @xmath48 that leads to einstein s dispersion relation @xmath49 . as so , the state ( [ sol1 ] ) has no definite intrinsic parity , qubit 1 is in a superposition of both eigenstates . the total parity operator @xmath50 acts on the kronecker product @xmath51 as @xmath52 ; indeed it is the product of two operators , the intrinsic parity @xmath53 ( having two eigenvalues , @xmath54 ) and the spatial parity @xmath55 ( @xmath56 ) . thus , @xmath57 applies on @xmath58 , eq . ( [ sol1 ] ) , and it follows that @xmath59 . regarding the spatial parity operator @xmath60 the @xmath61 ( @xmath62 ) sign stands for axial ( polar ) vectors . complementarily , the @xmath63-matrices are @xmath64 , @xmath65 , @xmath66 , and @xmath67 . there is an asymmetry between the two terms within the brackets in the solution ( [ sol1 ] ) : the first one represents the limit for the non - relativistic state of a spin @xmath4 free fermion , namely , the solution to the schrdinger equation , while the second term is responsible for the relativistic effect ( containing the helicity operator @xmath68 ) . due to the correlation between the parity and helicity qubits , a hypothetical measurement that results in qubit @xmath41 would reduces the dirac solution to the non - relativistic schrdinger equation while if the result of the measurement is qubit @xmath69 the solution is reduced to the purely relativistic term . however , there is no hint that the nature selects one of the two components under any kind of measurement . nevertheless , as we are going to see below , for an ensemble of fermions in contact with a thermal reservoir , one intrinsic parity eigenstate will be selected naturally as an effect of cooling . the helicity eigenvalue equation @xmath70 has orthogonal eigenstates @xmath71 ( @xmath72 , @xmath73 ) where the angles @xmath74 and @xmath75 determine the direction of @xmath76 ( on a spherical surface of radius 1 the tips of the versors @xmath76 and @xmath77 are localized by the angles @xmath78 and @xmath79 ) and the kets @xmath80 , @xmath81 stand for @xmath82 and @xmath83 . so , the spinor @xmath84 can be written as the superposition @xmath85 ( we omit the subscript @xmath86 ) where @xmath87 is the density distribution of the linear momentum . the spinor ( [ h ] ) correlates the linear momentum ( a c - number ) to the helicity eigenstates , however , for simplicity , we are going to assume that the linear momentum is not correlated to the helicity , therefore @xmath88 where a mixing angle , @xmath89 } $ ] , and a relative phase , @xmath90 , have been introduced . the helicity sector of the dirac equation solution will make use of the spinors @xmath91 with @xmath92 that are normalized @xmath93 , however they are orthogonal only for @xmath94 , because @xmath95 . it is worth noting that doing the changes @xmath96 and @xmath97 we get @xmath98 . for a normalized linear momentum distribution , @xmath99 , one has @xmath100 and @xmath101 the simplified form of the time - independent component of eq . ( [ sol1 ] ) becomes @xmath102 where @xmath103 with @xmath104 , and the pure state density matrix is @xmath105 calculating the trace over the qubits the result is @xmath106 = \left\vert \varphi ( \vec{p } ) \right\vert ^{2}$ ] , thus @xmath107 } = 1 $ ] . for an ensemble of free fermions interacting with a thermal environment at temperature @xmath108 , we identify the probability density @xmath109 with a normalized distribution function isotropic in the linear momentum , @xmath110 . integrating eq . ( [ pure ] ) over the linear momentum , the reduced density matrix becomes @xmath111 where we omit the subscripts @xmath3 and @xmath86 in the right - side . as the dependence on the solid angle @xmath112 is exclusively relegated to the helicity states @xmath113 , we get @xmath114 where @xmath115 and @xmath116 from which , by eq . ( [ polar ] ) , one obtains @xmath117 with real coefficients @xmath118 and @xmath119 with complex coefficients @xmath120 noting that @xmath121 , @xmath122 and @xmath123 . the reduced density operator ( [ rho3 ] ) becomes @xmath124 \notag \\ & & + { m}_{--}^{s}\left ( t\right ) \left [ \left\vert -\right\rangle \left\langle -\right\vert \otimes \left ( n_{-}\left\vert \uparrow \right\rangle \left\langle \uparrow \right\vert + n_{+}\left\vert \downarrow \right\rangle \left\langle \downarrow \right\vert \right ) \right ] \notag \\ & & + { m}_{+-}^{s}\left ( t\right ) \left [ \left\vert + \right\rangle \left\langle -\right\vert \otimes \left ( \tilde{n}_{-}\left\vert \uparrow \right\rangle \left\langle \uparrow \right\vert + \tilde{n}_{+}\left\vert \downarrow \right\rangle \left\langle \downarrow \right\vert \right ) \right . , \notag \\ & & + \left . \left\vert -\right\rangle \left\langle + \right\vert \otimes \left ( \tilde{n}_{+}\left\vert \uparrow \right\rangle \left\langle \uparrow \right\vert + \tilde{n}_{-}\left\vert \downarrow \right\rangle \left\langle \downarrow \right\vert \right ) \right ] , \label{dens4}\end{aligned}\ ] ] where we defined the coefficients @xmath125 with @xmath126 , @xmath127 , and @xmath128 . as we admitted that qubit 1 stands for the intrinsic parity of the fermion , we get the correlation density operator for helicity and intrinsic parity as suggested many years ago by t. d. lee and c. n. yang @xcite , in an _ ad - hoc _ procedure . in that case ( c. f. eq . ( 23 ) from ref . @xcite ) the density matrix describes a coherent collection of _ spinorial _ particles that exhibit spin and parity as correlated quantum features , through which a suitable interference phenomena between parity doublets is identified . furthermore , setting @xmath129 the fermi - dirac distribution can be written as @xcite @xmath130 where we have set the boltzmann constant @xmath131 and @xmath132 is a riemann zeta function . the coefficients ( [ m+- ] ) ) can be calculated numerically , @xmath133 with @xmath134 ( temperature per unit mass with @xmath135 ) . in table i we present the values of the @xmath136 for different temperatures . [ cols="^,^,^,^",options="header " , ] in fig . [ fig01 ] we have drawn the coefficients @xmath137 ( for fermions ) as function of @xmath138 , that contain the amount of correlation between spin and intrinsic parity of a fermi gas embedded in a thermalized environment . now we speculate about the meaning of the results . focusing our attention on the cosmological scenario , the temperature of the universe is a parameter that , roughly , parallels the evolution in time : after the initial surge of a very hot and compact seed of energy , the radiation dominated universe expands and cools down , still keeping the thermalized blackbody frequency distribution . the higher ( lower ) the temperature the earlier ( later ) is its age and size , @xmath139 where @xmath140 is the expansion parameter . at quite high temperatures , @xmath141 , or @xmath142 ( for electrons @xmath143 ) , the coefficients @xmath144 , @xmath145 and @xmath146 are close to @xmath147 , while as @xmath148 , @xmath149 and @xmath145 , and @xmath150 go to zero . thus , at early times , when the universe was quite hot , the fermions ( and also antifermions ) existed with positive and negative intrinsic parity ( @xmath151 @xmath152 ) , whereas the transition amplitudes , from positive to negative parity , and vice versa , were almost the same . as the universe gone expanding the temperature was reducing , then one parity ( positive ) began to dominate over the other , @xmath153 , @xmath154 ) , and the transition amplitude @xmath155 was reducing too , so the negative parity fermions became scarcer . as the universe cooled further , say at @xmath156 ( @xmath157 ) for @xmath158 , the fraction of negative to positive parity fermions became @xmath159 , while the inverse comes out for the @xmath36 antifermions . at the present epoch the positive parity fermions dominate , constituting the building blocks of the observed universe , while those having negative parity are reduced to almost beyond observation . by its turns if we consider the negative energy solution ( @xmath36 ) , for the antifermions , the inverse occurs , negative parity prevails at low temperature while the positive parity fermions become quite scarce . so after our calculations the separation between positive parity fermions and negative parity antifermions in the present cold universe finds a plausible explanation . as a function of the temperature parameter @xmath138.,title="fig : " ] the charge conjugation operation changes matter into antimatter and it is represented by the operator is @xmath160 , where @xmath161 stands for the complex conjugation operator and @xmath162 . it transforms a dirac spinor as @xmath163 thus a state is invariant under @xmath164 operation whenever @xmath165 and @xmath166 . applying @xmath164 on ( [ sol1 ] ) one obtains @xmath167 implying also the change @xmath168 . now , applying @xmath169 on the right and @xmath164 on the left of state ( [ dens4 ] ) , the coefficients do not change , while @xmath170 and @xmath171 , thus @xmath172 differs from @xmath173 by the following interchanges @xmath174 and @xmath175 , or , fermion ( @xmath176 ) and antifermion ( @xmath177 ) interchange their role . the parity - helicity density matrix is @xmath178 from which we verify that , under the peres - horodecki criterion @xcite , there is no entanglement between intrinsic parity and helicity since the partially transposed matrix , @xmath179 , coincides with @xmath180 . the eigenvalues of @xmath181 are @xmath182 that we shall use bellow . we have omitted the explicit dependence on @xmath138 . the reduced normalized state for the intrinsic parity is @xmath183 where the nondiagonal term stands for the transition probabilities @xmath184 , @xmath185 , so the strength of a transition depends on the mixing angle @xmath186 . as the fermi gas cools , @xmath187 , independently of the value of @xmath188 and @xmath189 , so the state decoheres at a lower rate than it takes for the system to reduce the negative ( @xmath190 ) or positive parity ( @xmath191 ) contribution to the mixture . for @xmath192 we find @xmath193 . the eigenvalues of intrinsic parity state ( [ red1 ] ) are @xmath194 by its turn , the normalized density operator for the helicity is diagonal @xmath195 and the coefficients are @xmath196 and @xmath197 for @xmath198 or @xmath199 , @xmath200 , @xmath201 @xmath202 , therefore , there is no natural preference for any helicity direction , neither a dependence on the temperature . for @xmath203 or @xmath204 , and @xmath158 , the probabilities are unbalanced and show a dependence on the temperature , @xmath205 , since at low temperatures @xmath206 , see table ii . in the present epoch the helicity does not show any directional preference for the fermions , they are found in positive and negative helicity equally likely , so @xmath207 is the most plausible choice for any value for the mixture angle @xmath189 and temperature @xmath138 . the von - neumann entropy of a density operator @xmath208 is defined as @xmath209 , where @xmath210 are the eigenvalues , thus we calculate the mutual information between intrinsic parity and helicity as @xmath211 for several values of the mixture angles @xmath189 and the phase @xmath212 , as depicted in fig . [ fig02 ] . as a function of the temperature parameter @xmath138 . the plots are for several choices of the mixing angle @xmath189 . different lines represent different phases : @xmath213 ( black ) , @xmath214(red ) , and @xmath215 ( blue ) . for each @xmath186 and for different @xmath216 , the curves show very tiny mutual deviations from each other . @xmath217 decreases to zero at the non - relativistic ( @xmath218 ) limit and increases to its maximal value at the ultra - relativistic ( @xmath141 ) regime.,title="fig : " ] the variation of @xmath212 implies into some very tiny mutual deviations from each other at the transition regime ( from ur to nr ) , i.e. , @xmath212 has not any relevant qualitative effect onto the mutual information between parity and helicity , thus reinforcing our previous hint for the choice @xmath204 . as expected , the mutual information @xmath217 is insignificant at low temperature , @xmath219 , while it is maximal for @xmath141 . it is common knowledge @xcite that at the very early universe ( @xmath220 ) the photons had enough energy to become electron - positron pairs , so electrons and positrons existed in thermal equilibrium with the radiation . at about 400 000 years after the big bang there is change , radiation is free to pass through the universe as its expansion changes it from opaque to transparent . as the universe expanded it cooled , and when the temperature reduced to @xmath221 photons had not enough energy to create @xmath222 pairs , so electrons and positrons were no longer in thermal equilibrium but radiation acquired a thermalized blackbody distribution . a fundamental question is : why matter ( positive intrinsic parity ) eventually dominated over anti - matter ( negative intrinsic parity ) , which , presumably , were initially in equal footing ? our calculations and results can not explain the `` disappearance '' of the antifermions that existed in the early universe , however it hints of why at the present epoch the fermions the quarks and leptons that constitute matter have positive intrinsic parity ( our estimate is @xmath2 positive for 1 negative parity fermions ) and any produced antifermion has a negative intrinsic parity , although the calculations show that at the early universe fermions and antifermions existed in a superposition of both parities entangled the helicity states . otherwise , on the non - relativistic limit the mutual information is null : any quantum correlation between the particle / antiparticle character and the state spin - polarization vanishes . it corresponds to an issue that can be reproduced , from the mathematical point of view , by a foldy - wouthuysen unitary transformation @xcite ] . finally , we point out the essentiality of the present framework @xcite , where it was assumed that the dirac equation and the spinors describe the dynamics and the state of a massive particle carrying two qubits of information , the helicity and the intrinsic parity . that approach permits quantifying the quantum correlation and the entanglement between the particle / antiparticle degrees of freedom . moreover , we believe that it might be relevant discussing the destruction of the `` mirror '' symmetry ( external parity or _ left / right_-handed character ) in particle decays involving electroweak interactions , a point that certainly deserves to be scrutinized in the subsequent investigations . aeb acknowledges financial support from cnpq ( grant 300809/2013 - 1 ) . ssm acknowledges financial support from cnpq and from inct - iq . in this context , it is worth mentioning that the photon can also be cast in terms of a quantum wave equation @xmath224 , where @xmath225 is a spinor having three components , and their real and imaginary part are the electric and magnetic fields respectively . @xmath226 components are the spin 1 matrices and @xmath9 is the linear momentum . this approach for the photon was studied in @xcite .	dirac equation describes the dynamics of a relativistic spin-1/2 particle regarding its spatial motion and intrinsic degrees of freedom . here we adopt the point of view that the spinors describe the state of a massive particle carrying two qubits of information : helicity and intrinsic parity . we show that the density matrix for a gas of free fermions , in thermal equilibrium , correlates helicity and intrinsic parity . our results introduce the basic elements for discussing the spin - parity correlation for a fermi gas : ( 1 ) at the ultra - relativistic domains , when the temperature is quite high , @xmath0 , the fermions have no definite intrinsic parity ( 50% : 50% ) , which is maximally correlated with the helicity ; ( 2 ) at very low temperature , @xmath1 , a unique parity dominates ( conventionally chosen positive ) , by @xmath2 to @xmath3 , while the helicity goes into a mixed state for spin up and down , and the quantum correlation decoheres . for the anti - fermions we get the opposite behavior . in the framework of quantum information , our result could be considered as a plausible explanation of why we do accept , as a fact ( consistent with the experimental observation ) , that fermions ( and anti - fermions ) , in our present epoch of a cool universe , have a unique intrinsic parity . the framework for constructing spin - parity entangled states is established .
the study of stellar dynamics around massive black holes begins in 1970s in application to ( still hypothetical ) intermediate - mass black holes in globular clusters @xcite . the early studies established the importance of two - body relaxation which changes angular momenta of individual stars and drives them into the loss cone the region in phase space in which a star would be captured by the black hole in at most one orbital period . later , the loss cone theory was applied to galactic nuclei @xcite . if the black hole mass is not too large ( @xmath1 ) , a star lost into the black hole could produce a tidal disruption flare @xcite , and several candidate events have been observed to date ( e.g. * ? ? ? * ; * ? ? ? to deliver a star into the loss cone , it is necessary to reduce its orbital angular momentum to a very small value . in a spherical geometry , this can only be achieved by two - body relaxation , but relaxation times in galactic nuclei are typically very long , especially in massive galaxies . the predicted rate of capture events is generally rather low , of order @xmath2yr per galaxy , weakly depending on the black hole mass @xmath3 . on the other hand , if a galaxy is not precisely spherically symmetric , then angular momenta of individual stars are not conserved , so that they may be driven into the loss cone by _ collisionless _ torques ( as opposed to _ collisional _ two - body relaxation ) . this effect was recognized quite early @xcite , and more recent studies have suggested that particular properties of triaxial galactic nuclei , namely the existence of a large population of centrophilic orbits , may keep the capture rate well above that of a similar spherical system @xcite . these conclusions were obtained based on the properties of orbits , and not on the full - scale dynamical simulations , which are fairly non - trivial to conduct for such a problem . there are several numerical methods that have been used for studying stellar systems with massive black holes . fokker - planck models are usually restricted to spherical @xcite or at most axisymmetric @xcite geometries , as are gaseous @xcite or monte - carlo @xcite models . @xmath4-body simulations ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) do not have such limitation , but as we will show below , it is extremely hard to operate them in a regime with realistic proportion between collisional and collisionless effects . we have developed a new variant of monte - carlo code that is applicable in any geometry and have used it to study how the non - spherical effects change the rate of star captures by supermassive black holes . we begin by reviewing the basic properties of orbits in galactic nuclei of various geometries in [ sec : orbits ] . then in [ sec : relaxation ] we set up the loss - cone problem and consider the interplay between collisional and collisionless relaxation processes . [ sec : scaling ] is devoted to order - of - magnitude estimates and scaling relations . in [ sec : montecarlo ] we describe the novel monte - carlo method for simulating the dynamics of near - equilibrium stellar systems of arbitrary geometry , and in [ sec : results ] apply this method for the problem of star capture by supermassive black holes in axisymmetric and triaxial galactic nuclei . [ sec : conclusions ] presents the conclusions . throughout this paper , we consider motion of stars in a time - independent potential , which is composed of a newtonian potential of the supermassive black hole at origin ( we ignore relativistic effects for the reasons described later ) and the potential of extended distribution of stars : @xmath5 it is clear that depending on the relative contribution of these two terms ( or , rather , their derivatives that determine acceleration ) , there are two limiting regimes and a transition zone between them . close to the black hole , the motion can be described as a perturbed keplerian orbit . at larger radii , the black hole almost unnoticed by the majority of stars except for those with low angular momenta , which are able to approach close to the black hole ; nevertheless , even these orbits are mostly determined by the extended stellar distribution rather than the point mass . the boundary between the two cases is conveniently defined by the black hole influence radius @xmath6 , which contains stars with the total mass @xmath7 . another commonly used definition of the influence radius is @xmath8 , where @xmath9 is the stellar velocity dispersion ; it is more readily computed from observations but has a more indirect meaning , as @xmath9 is determined by the distribution of matter in the entire galaxy and not just close to the black hole , and may well be a function of radius itself . the two definitions of influence radius give the same value for a singular isothermal density profile , but may differ by a factor of few for less cuspy profiles ( * ? ? ? * section 2.2 ) . we will use the first definition henceforth . in general , the only classical integral of motion in a time - independent potential of arbitrary geometry is the total energy ( per unit mass ) of an orbit @xmath10 . obviously , in the spherical and axisymmetric cases there exist additional integrals three and one components of the angular momentum vector @xmath11 , correspondingly . in the vicinity of the black hole , however , the orbits look like almost closed keplerian ellipses , with the oscillations in radius between peri- and apoapses ( @xmath12 ) occuring on a much shorter timescale ( radial period @xmath13 ) than the changes in the orientation of the ellipse ( precession timescale ) . under these conditions , one may use the method of orbit averaging ( e.g. * ? ? ? * ) to obtain another conserved quantity @xmath14 , where @xmath15 is the semimajor axis by @xmath16 . thus the total hamiltonian is split into the keplerian and the perturbing part , each one being conserved independently . the perturbing potential @xmath17 , averaged over the radial oscillation period , becomes an additional integral of motion , the so - called secular hamiltonian . the existence of this additional integral does not add new features in the spherical case , but is important in other cases . in particular , the motion in an axisymmetric nuclear star cluster becomes fully integrable , while in the weakly triaxial case another non - classical integral of motion can be shown to exist @xcite , so that the system is again fully integrable . empirical evidence shows that even for rather large deviation from spherical symmetry , orbits in the triaxial case remain regular in the vicinity of the black hole , if one neglects relativistic effects . @xmath18{fig_orbits.png}\ ] ] in a generic triaxial potential there usually exist several types of tube orbits , which conserve the sign of one component of angular momentum , and box orbits , which do not have a definite sense of rotation around any axis @xcite . the latter have stationary points , in which the velocity ( and hence , the angular momentum ) comes to zero , and are generally centrophilic , meaning that a star can pass at an arbitrarily small distance from the origin . centrophilic orbits , by necessity , can not conserve the sign of any component of angular momentum , however the opposite is not true there exist various non - tube resonant orbit families which avoid coming too close to the center . these orbits play an important role in the structure of triaxial galaxies with density cusps and/or central black holes @xcite , as the regular box orbits are destroyed by the sharp gradients in the potential near origin and replaced by chaotic orbits and resonant orbit families . thus the centrophilic orbits outside a few influence radii and up to a radius containing @xmath19 are generally chaotic @xcite , although in some special cases the motion remains fully integrable even in the presence of the black hole @xcite . on the other hand , the motion inside @xmath6 is generally regular , with the box orbits being replaced by a related family of pyramid orbits ( * ? ? ? * fig.11a ) . these are also centrophilic , and as such , are natural candidates for the capture by the black hole @xcite . tube orbits also exist near the black hole , and a special class of short - axis tubes has a noticeably conical shape and are called saucers @xcite . figure [ fig : orbits ] shows the four major classes of orbits in a triaxial nucleus , of which two exist also in the oblate axisymmetric case . in an even more general case without a triaxial symmetry ( for instance , if the black hole is displaced from the center of stellar potential ) , there are other classes of orbits ( e.g. lens orbits , see * ? ? ? * ) , some of them also being centrophilic . in what follows , we consider systems with at least triaxial symmetry . given the non - conservation of angular momentum in non - spherical potentials , it is important to know the amplitude and timescale of its variation . let us describe the deviation from spherical symmetry by a parameter @xmath20 , roughly equal to @xmath21 , where @xmath22 and @xmath23 are the major and minor axes ( more precise definition is given by eq.11b in @xcite , where it is called @xmath24 ) . for a triaxial system , there is another parameter @xmath25 which describes the degree of triaxiality ( @xmath26 is the intermediate axis ) . it ranges from @xmath27 for oblate to @xmath28 for prolate axisymmetric models ; in the rest of the paper we will assume @xmath29 ( maximal triaxiality ) , having in mind that the importance of triaxial effects decreases as the system approaches axisymmetry . let @xmath30 be the angular momentum of a circular orbit with the given energy , and define the dimensionless squared angular momentum @xmath31 . it turns out that for tube orbits with high enough @xmath32 the variation in @xmath32 is rather small , but for saucer and pyramid orbits , both of which appear at @xmath33 , this variation can be by order of @xmath32 itself . ( note , however , that in the axisymmetric case the magnitude of angular momentum is bounded from below by its conserved component @xmath34 ) . the timescale for such variation is of order @xmath35 times the precession time @xmath36 @xcite . the above estimates were valid for the motion in the sphere of influence ; outside @xmath6 the chaotic centrophilic orbits also typically have @xmath33 and the variation of angular momentum by order of itself occurs on the timescale of radial period @xmath13 . there is ample evidence that a substantial fraction of elliptical galaxies and bulges of disk galaxies might be triaxial @xcite , with the minor - to - major axis ratio in the range @xmath37 and a large spread in the triaxiality parameter . nuclear star clusters also are known to exhibit significant departures from spherical symmetry ( see * ? ? ? * for the case of milky way ) , which is also a natural consequence of one of the proposed scenarios of their formation @xcite . even if the mass distribution in the nuclear star cluster itself is spherical , the distortion of the gravitational potential by asymmetries on larger scales ( such as nuclear bars ) would still create the centrophilic orbit families . we will take the axis ratio @xmath38 and triaxiality @xmath29 as our fiducial parameters , but we note that the capture rates do not depend substantially on these values . the black hole tidally disrupts or directly captures stars which pass at a distance smaller than a critical radius @xmath39 , or , equivalently , pass the periapse while having orbital angular momentum smaller than @xmath40 . the former fate occurs for main - sequence stars if @xmath41 , depending on the mass of the star ( see fig.5 in * ? ? ? * ) , and for giant stars even for much larger @xmath3 ; the latter case applies to compact stars , stellar - mass black holes , or for the largest @xmath3 . we shall consider only non - rotating black holes ; see the paper by l.dai in this volume for the problem of tidal disruption of stars by kerr black holes . the critical angular momentum is @xmath42 , \ ] ] where @xmath43 and @xmath44 are the mass and radius of the star and the coefficient @xmath45 depends on the internal structure of the star . the region of phase space with @xmath46 is called the loss cone . a detailed review of the loss cone theory is given by @xcite ; here we remind several key points . substantial changes in both energy and angular momentum distributions occur on the timescale of relaxation time @xmath47 , but the relaxation in angular momentum is more important for the overall rate of capture @xcite , since near the loss cone boundary a comparatively small change in @xmath48 can drive a star into the black hole . the relaxation in @xmath49 would produce a @xmath50 cusp @xcite in a time @xmath51 ; since many galaxies are not that dynamically old , we will not assume this particular density profile and let it to be an unspecified function . the isotropic distribution function of stars in energy @xmath52 is given by the eddington formula , while the density of states @xmath53 provides the link between @xmath54 and the number of stars in energy interval @xmath55 : @xmath56 ( * ? ? ? * eq.4.46 , 4.56 ) . in what follows , we will use interchangeably the energy @xmath49 or the semimajor axis @xmath15 ( for a non - keplerian potential it should be more appropriately called the radius of a circular orbit with the given energy ) . in the spherical case , the only mechanism producing changes in angular momentum is the two - body relaxation . it can be substantially enhanced if there are massive perturbing objects such as molecular clouds @xcite , but the basic physical effect remains the same . resonant relaxation also enhances the exchange of angular momentum between stars @xcite , but has a rather moderate effect on the overall rate of capture , so we also ignore it in the discussion . relativistic effects play an important role close to the black hole ( * ? ? ? * c.will , this volume ) , but since the bulk of captured stars come from radii comparable to @xmath6 , relativity can be neglected for them . on the other hand , in non - spherical systems the angular momentum changes due to both two - body relaxation and torques from non - spherical mass distribution , the interplay between these two effects being quite complex . in the case of spherical symmetry , the main parameter that determines the flux of stars into the loss cone is the ratio @xmath57 of mean - square change in angular momentum per one radial period to the width of the loss cone : @xmath58 where @xmath59 is the limiting value of diffusion coefficient in @xmath32 for @xmath60 , roughly equivalent to @xmath61 . this quantity determines the boundary condition for the diffusion in @xmath32 in the following way @xcite . when @xmath62 , the relaxation is rather weak and as soon as the star reaches the boundary of the loss cone , it is captured within one orbital period ; thus the density of stars inside the loss cone is zero this is called the empty loss cone regime . on the other hand , if @xmath63 , the changes in angular momentum occur so rapidly that in the time - averaged sense , the loss cone remains fully populated : the stars with @xmath64 are still being captured once per radial period , but this occurs only when the star actually passes the periapse , and for the rest of its orbit it can stay inside the loss cone without impunity . if we introduce the time - averaged value of distribution function in the loss cone as @xmath65 , then the rate of capture is given by the number of stars inside the loss cone divided by radial period : @xmath66 in the spherical case , the boundary condition fully determines the steady - state solution of the diffusion equation : assuming a particular form of isotropized ( averaged over angular momenta ) distribution function of stars @xmath52 , the actual two - dimensional function @xmath67 has a nearly logarithmic profile in @xmath32 , and the rate of capture is @xcite @xmath68 it is easy to see that in the full loss cone limit ( @xmath69 ) , this quantity tends to ( [ eq : flux_full_lc ] ) , and @xmath70 . close to the black hole the loss cone is empty , and further out it may or may not reach a full loss cone regime ; for @xmath71 in lower - density galactic nuclei this never happens , while for smaller @xmath3 the transition between empty and full loss cone regimes may lie around @xmath6 . for non - spherical system , the treatment of two - body relaxation is more difficult and has only been carried out rigorously in the axisymmetric case inside the sphere of influence @xcite , where the dynamics is fully integrable . the main conclusion is that the steady - state solution at @xmath72 is still well described by a logarithmic profile , but at smaller @xmath32 there exists a more extended `` loss wedge '' @xcite , in which the orbits on average have @xmath73 but may get into the loss cone ( below @xmath74 ) due to torques from non - spherical potential , roughly after one precession time @xmath75 . the volume of the loss wedge is larger than the loss cone by a factor @xmath76 , and the boundary condition corresponds to a full loss cone regime ( i.e. the distribution function inside the loss cone is almost the same as on the boundary of the loss wedge ) . in other words , the quantity @xmath77 , analogous to ( [ eq : q_spher ] ) , is @xmath78 for most orbits , because the change of angular momentum due to precession during one radial period is typically larger than the size of the loss cone . however , now the global shape of the solution is not simply determined by the boundary condition , but rather by the average two - body relaxation rate at the given energy : the non - spherical torques can not bring an orbit all the way from @xmath79 to @xmath74 this still remains the duty of relaxation . overall , the larger size of the loss wedge increases the steady - state capture rate by a factor of few , compared to the spherical case , but only in the regime when the spherical system would have empty loss cone . roughly speaking , the star only needs to get down to @xmath80 by collisional diffusion , and then it will be driven into the loss cone by collisionless effects ; but as the dependence of the flux on the effective size of the loss region is logarithmic ( [ eq : flux_spherical ] ) , this does not greatly increase the loss rate . a similar picture should be describing triaxial systems ; however , for them , and to a lesser extent for axisymmetric systems , another effect may be more important , namely the draining of the loss region . this term refers to the gradual depletion of the region in phase space populated by orbits that may attain angular momentum lower than @xmath40 at some point of their evolution in the smooth stationary potential ( i.e. in the absense of relaxation ) . in the triaxial case , this region contains the entire population of centrophilic orbits ( which constitute approximately a fraction @xmath81 of total mass ) , whether these are regular pyramids inside @xmath6 or chaotic orbits further out . the lifetime of orbits in the loss region is much longer in the triaxial case ; it is roughly given by ( * ? ? ? * eq.117 ) : @xmath82 the latter equation used the radius of direct capture and not tidal disruption and is valid for @xmath83 . an analogous expression for draining of chaotic orbits in the axisymmetric case ( * ? ? ? * eq.65 ) gives @xmath84 thus it is clear that the timescale for draining the loss region can be quite long , especially in the triaxial case ; during this time , the capture rate at a given energy is roughly given by the expression ( [ eq : flux_full_lc ] ) for the full loss cone . the necessary conditions for this are that ( i ) the initial distribution of stars in @xmath32 is close to uniform ( isotropic ) , and ( ii ) the typical change in angular momentum during one radial period is larger than the loss cone size . the first condition is the most neutral , if not natural , choice for the velocity distribution function , although it may be substantially overestimating the capture rate if the low angular momentum part of the phase space is initially depleted due to the slingshot ejection of stars by a binary supermassive black hole @xcite . the second condition is satisfied for most orbits which are not too close to the black hole ( essentially most orbits that contribute to the total flux ) , as shown by ( * ? ? ? * eq.55 ) and ( * ? ? ? * eq.59 ) . it is also worth noting that the above discussion assumed a steady - state solution for the distribution function ( to the extent that we ignore the loss of stars into the black hole , which is reasonable on the hubble timescale ) . a time - dependent solution to the fokker - planck equation may give the loss rate higher than the steady - state by a factor of few @xcite , or , conversely , lower @xcite , depending on the initial conditions . there are two separate factors that complicate matters in the non - stationary case . the first is that if the relaxation time is much longer than the hubble time , then the gradient of the distribution function near the capture boundary , which determines the capture rate in the empty loss cone regime , is not uniquely determined by the isotropized distribution function @xmath52 , as in the steady state ( eq . [ eq : flux_spherical ] ) . rather , it depends on the prior evolution and , ultimately , the initial state . figure 13 in @xcite shows that under the assumption of initially isotropic distribution of stars , the time - dependent capture rate for a spherical galaxy may be @xmath85 times higher than the steady - state rate for the most massive black holes . on the other hand , these two rates are roughly equal for @xmath86 , when the relaxation is fast enough to establish a nearly steady - state profile . the second factor , more important for non - spherical systems , is related to the collisionless draining of the loss region , which could increase the captured mass by more than an order of magnitude compared to the steady - state solution . however , one should keep in mind that both factors sensitively depend on the initial population of stars in and near the loss region ( i.e. , on the details of galaxy formation ) , while the steady - state rates , which differ only by a factor of two between spherical and axisymmetric systems , provide a more robust estimate . we can obtain a very rough estimate of the total steady - state capture rate for a spherical galaxy by noting that @xmath87 peaks around @xmath88 , so that if the system is in the empty loss cone regime at this energy , then the total flux is @xmath89 to arrive at a single - parameter formula , we use the @xmath90 relation , which is typically written as @xmath91 with @xmath92 , @xmath93 being the typical values found in the literature ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the influence radius then scales as @xmath94 with the parameter @xmath95 depending on the density profile . taking @xmath9 to be the line - of - sight velocity , averaged over the effective radius , we obtain @xmath96 pc for a @xcite model with a @xmath97 cusp . then the estimate ( [ eq : ftotal_spher ] ) gives @xmath98yr for more massive galaxies which are in the empty loss cone regime , comparable with more accurate calculations ( * ? ? ? * fig.13 ) . for an axisymmetric galaxy , the steady - state capture rate is only moderately larger than for a spherical one , and have roughly the same functional dependence on galaxy parameters @xcite . the statement that axisymmetric galaxies have much higher capture rates is often attributed to this paper , but in fact a more careful reading reveals that the effect of axisymmetry increases the capture rates due to relaxation only by a factor of two or so . this ignores the draining of the loss wedge , which is justified since the draining time is substantially shorter than the hubble time at the radii of interest , as is the case for most realistic situations : substituting ( [ eq : rinfl_m ] ) into ( [ eq : tdrain_axi ] ) , we obtain that @xmath99 yr at the radius of influence . on the other hand , the mass of stars in the loss wedge , and hence their draining rate , is roughly proportional to the mass of the black hole , while the capture rate due to relaxation scales more slowly with @xmath3 . hence , @xcite suggest that for the most massive galaxies it is indeed the draining of the loss wedge that is the main source of captured stars , and the capture rate exceed the steady - state value by orders of magnitude , which is apparent in comparing their figures 3 and 5 . thus their study presented two very different conclusions about capture rates in axisymmetric galaxies , without a clear explanation of the discrepancy . it is worth reiterating that the first one , about a moderate increase in the relaxation flux , is more robust , as it does not depend on the initial conditions , while the second one , about a much higher capture rate due to draining , is based on the assumption that the loss wedge was fully populated at the beginning of evolution . both conclusions were confirmed by @xcite using a more elaborate calculation . for a triaxial galaxy , however , the draining times for most galaxies are comparable to or exceed the hubble time , and the capture rate at a given energy is comparable to the full loss cone rate ( [ eq : flux_full_lc ] ) if @xmath100 ( again under the assumption of initially fully populated loss region ) . we may make a crude estimate by integrating the expression for the full loss cone flux from the critical energy up to infinity , where the critical energy @xmath101 is given by the condition that the draining time ( [ eq : tdrain_tri ] ) is equal to the hubble time . hence , the total capture rate is @xmath102 from ( [ eq : tdrain_tri],[eq : rinfl_m ] ) we see that , evaluated at the radius of influence , @xmath103 yr , so that unless the triaxiality is very small , most of the draining flux comes from inside @xmath6 . we may make the above estimate more quantitative by taking the expressions for @xmath104 and @xmath105 in the newtonian potential ( * ? ? ? * eqs.3.49 , 3.51 , 5.182 ) . this yields @xmath106 the capture rate increases approximately linearly with @xmath3 , a situation rather different from the spherical case . for a @xmath97 cusp , the present - day flux is @xmath107 . note that , as the more bound orbits are drained more quickly , the total rate declines with time , although rather slowly . even with such a high flux , it is not quite correct to say that the capture rate is comparable or exceeds the full loss cone rate , as the latter quantity is poorly defined by itself : the undepleted full loss cone draining rate would _ increase _ with increasing @xmath108 , producing a divergent overall total flux ( * ? ? ? * chapter 6 ) . a rough estimate obtained for a singular isothermal sphere ( @xmath109 density cusp ) , taking the full loss cone rate at the radius of influence , produces another two orders of magnitude larger flux @xcite . the scaling relation ( [ eq : ftotal_triax ] ) is somewhat different from the one obtained by @xcite based on their model of scale - free @xmath97 cusp with @xmath110 . there are several reasons for this . first , their models , having a power - law density profile , could not extend to infinity , so were cut off at a few influence radii ; second , their plots of capture rate versus energy used the definition of potential without the contribution of black hole , which is strictly valid only well outside influence radius . in short , there is no range of radii such that @xmath111 , so it is hard to interpret their power - law fits . in particular , the dashed line on their fig.3 , marking the full loss cone flux , is correct for @xmath112 , and it is indeed apparent that for large energies the points start to follow that asymptote , instead of the power - law fit for smaller energies . on the other hand , a correct expression for the full loss cone flux valid both inside and outside has a peak at roughly @xmath113 and declines towards lower energies as @xmath114 ; however , as it extends to arbitrarily negative values of @xmath49 , the overall flux would diverge at low , not at high @xmath49 as they claim . this divergence is removed by imposing a lower limit at energy @xmath101 , which introduces the time dependence in the total flux that was derived in the above formulae . lastly , their normalization is somewhat lower because the capture radius was taken to be at one schwarzschild radius instead of four . despite these complications , the basic statement , that the draining rate may be equal to the full loss cone rate during a hubble time , remains valid if we take into account the proper cutoff at energies below @xmath101 , and even the numerical estimates are correct to within an order of magnitude . we now discuss scaling relations that may facilitate numerical study of loss rates . state - of - the - art @xmath4-body codes on the present - day hardware can simulate systems with @xmath115 stars for @xmath115 dynamical times , which takes months to years of wall - clock time ( d.heggie , priv.comm . ) . a simulation of even the relatively low - mass milky way nuclear star cluster would need to follow some @xmath116 stars for @xmath117 dynamical times , clearly beyond the reach of direct @xmath4-body simulations in the near future . however , we may design ( almost ) equivalent systems with suitably scaled parameters . we start from spherical case , in which the important quantity that needs to be preserved by the scaling is the radius ( or energy ) of transition between empty and full loss cone regimes , i.e. at which @xmath118 ( eq.[eq : q_spher ] ) . then it is clear that we need to change @xmath74 and @xmath119 by the same factor @xmath120 . the flux in the full loss cone regime ( eq.[eq : flux_full_lc ] ) will be also multiplied by @xmath120 , and the flux in the empty loss cone regime ( eq.[eq : flux_spherical ] with @xmath121 ) will be multiplied by almost the same factor , except for the change in the logarithmic factor in the denominator . the total simulation time , in units of dynamical time , should then be scaled by @xmath122 . in the axisymmetric case the scaling is largely the same , if we ignore the draining of the loss wedge ; furthermore , in this case @xmath74 under the logarithm in the denominator is replaced with an effective constant which depends only on @xmath20 @xcite . the draining , however , scales differently with @xmath120 , so we may only ignore its contribution if the scaled draining time is less than the scaled simulation time . in the triaxial case , the flux from draining of centrophilic orbits can not be ignored as it gives the dominant contribution to the total capture rate . we need to preserve the ratio of draining time ( [ eq : tdrain_tri ] ) to the simulation time , which , fortunately , means that both should change by the same factor @xmath122 . thus the relative importance of collisional and collisionless capture rate will not change under the same scaling of @xmath123 . unfortunately , even with this scaling it is not possible to design a feasible @xmath4-body simulation for most of galaxies : since the relaxation rate @xmath124 , and we need to consider galaxies with @xmath125 , @xmath120 should be of order @xmath126 . thus instead of a system which is @xmath127 dynamical times old at @xmath6 , we get a system which has an age of @xmath128 dynamical time , and a capture radius which is a substantial fraction of @xmath6 ; clearly , such an extrapolation is too brave . in addition , for the system with a binary black hole the scaling is impossible to follow , as the stars are lost due to three - body scattering and not to the capture , so that the relevant radius @xmath39 is proportional to the size of the binary orbit and can not be changed at will . in such case , increasing the relaxation rate without adjusting the loss cone size moves the system more into a full loss cone regime , and decreases the relative importance of draining . the complexity of interplay between collisional relaxation and collisionless changes in angular momentum makes a direct solution of fokker - planck equation impractical in the general case of triaxial system with a mixture of regular and chaotic orbits . on the other hand , direct @xmath4-body simulations are presently incapable of achieving the necessary regime of relatively weak two - body relaxation , compared to the collisionless effects . to address this problem , we have developed a novel monte - carlo method . the idea is to follow the motion of sample orbits in a smooth , non - spherical potential while adding a random perturbation to the equations of motion , which would mimick two - body relaxation with an _ adjustable _ magnitude ( unrelated to the actual number of test particles ) . the approach is actually not so novel , dating back to @xcite , who used it for a spherical system with diffusion coefficients being computed under an assumption of maxwellian distribution of background ( field ) star velocities . the perturbations were applied directly to the velocity of the test star , its motion being numerically integrated in the given potential . later on , the idea was modified in @xcite by using orbit - averaged diffusion coefficients in energy , and further developed by @xcite and @xcite to study star clusters around supermassive black holes , using two - dimensional orbit - averaged diffusion in @xmath129 space with a special treatment of loss - cone ensuring a correct form of the boundary condition . the diffusion coefficients were recomputed regularly , using the distribution function of field stars equal to the isotropized distribution of test stars . at the same time , another branch of monte - carlo methods was developed by @xcite , in which the perturbations were computed between pairs of stars , thus equating the test and the field star population and ensuring conservation of global integrals of motion in each interaction . this method is employed in several presently used monte - carlo codes for the evolution of spherically symmetric star clusters ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? unfortunately , it is difficult to generalize it for non - spherical systems , while the spitzer s formulation is relatively straightforward to apply in any geometry . the detailed description of the new monte - carlo code is given elsewhere @xcite ; here we outline its main features . as in all previous studies , we consider scattering of test particles by an isotropic spherically symmetric distribution of field particles , which are described by a distribution function @xmath52 consistent with the spherical part of the potential in which the stars move . the scattering is described in terms of local ( position - dependent ) velocity diffusion coefficients @xmath130 ( e.g. * ? ? ? * eqs.5.23 , 5.55 ) , which represent mean and mean - squared changes in velocity per unit time . the trajectory of a test star is integrated numerically in the given smooth potential , and the perturbations to velocity are applied after each timestep of the ode integrator , typically several tens of steps per radial period , using locally computed drag and diffusion coefficients . we have checked that the prescription for adding velocity perturbation gives correct rates of orbit - averaged diffusion coefficients in energy and angular momentum in the spherical case . the loss cone is taken into account by computing the radius of each periapse from the numerically integrated trajectory and eliminating the particle if this radius is below @xmath39 . this automatically implies an appropriate boundary condition for the angular momentum diffusion in terms of @xmath57 , which applies for both regular changes of @xmath48 due to precession in the smooth potential and the random two - body relaxation . in the present study , we neglect the changes in the background potential in the course of evolution . most galactic nuclei have long enough relaxation times that the change in density profile due to relaxation in energy is negligible during the hubble time ( a few counterexamples exist , notably the milky way @xcite ) . the change in shape which is responsible for the collisionless changes in angular momentum is more difficult to assess . early studies @xcite have found that for a large enough black hole ( @xmath131 of the galaxy mass ) the triaxiality of the system is rapidly destroyed , presumably , due to diffusion of chaotic orbits . however , later papers were successful at creating strongly triaxial models with high fraction of chaotic orbits which nevertheless maintained their shape during many dynamical times @xcite . the changes in shape are quite straightforward to account for in our approach one needs to regularly update the smooth potential model as the system evolves ; however , for the simplicity of interpretation we decided not to explore this possibility , having checked that the systems under study do not noticeably evolve in shape in the absense of relaxation and capture . ( noise can substantially increase the chaotic diffusion rate , we leave the impact of noise on the chaotic orbits for a future study ) . we prepared a number of test models , for three geometries ( spherical , axisymmetric and triaxial ) , constructed by the @xcite orbit superposition method , using the publically available _ smile _ software @xcite . a few modifications were introduced in order to enhance the quality of models . first , to increase the effective `` resolution '' of the orbit library , we have changed the scheme for generating initial conditions in such a way that it creates more orbits with lower energies ( closer to the black hole ) and more orbits with low angular momenta . the actual weights of orbits are assigned by the optimization procedure to satisfy the density and kinematic constraints while attaining the most uniform possible distribution of orbit weights ; this scheme was modified to correctly account for non - equal priors for orbit weights . second , to make the non - spherical models as close as possible to spherical in terms of distribution in angular momentum , we binned the distribution function in @xmath132 space on a non - uniform grid ( with smaller cells closer to loss - cone boundary ) and required the distribution in @xmath32 to be flat in each energy shell . this produces a more isotropic model in the angular momentum space than just a requirement of velocity isotropy . the models were constructed using @xmath117 orbits , @xmath133 constraints for the spherical - harmonic decomposition of density and @xmath134 constraints for angular momentum distribution . the increase in effective resolution allowed us to have @xmath135 factor of 10 more centrophilic orbits compared to a uniform sampling ; accordingly , the typical weights of these orbits were also a factor of 10 less than average . to validate the predictions of capture rates given by the monte - carlo code , we compared it to a number of direct @xmath4-body simulations , in the regime where the latter are feasible . as in our previous studies , we used the efficient direct @xmath4-body code @xmath136grape - chain @xcite . it combines direct summation , using hardware acceleration of force computation provided by the _ sapporo _ library @xcite , with chain regularization algorithm for accurate treatment of close encounters of stars with the black hole . the accuracy parameter of hermite integrator was @xmath137 and the softening length was set to zero . as explained in the previous section , it is very difficult if not impossible to construct even a scaled @xmath4-body model for a realistic galactic nucleus . furthermore , for the difference between three geometries to be most prominent , we need to have a situation when the transition between empty and full loss cone regimes occurs well outside influence radius , and the draining time of saucer orbits at @xmath6 is much less than the simulation time ( otherwise we do nt expect to have a substantial difference between axisymmetric and triaxial cases both would have almost full loss cone draining rates at @xmath6 ) . an extensive search through the parameter space has led us to adopt the following parameters for the comparison between @xmath4-body and monte - calso simulations : @xmath138 @xcite model with total mass of unity and @xmath139 ( this gives @xmath140 and @xmath141 ) , capture radius @xmath142 , and simulation time of 20 @xmath4-body time units . the non - spherical models had axis ratios @xmath143 and @xmath144 for the axisymmetric and triaxial case , correspondingly , while having the same average @xmath145 profile as the spherical one . we stress again that these models do not correspond to any realistic galactic nuclei , but allow us to test the code . we have set up three @xmath4-body simulations with @xmath146 particles using @xmath136grape , and ran the monte - carlo code on the same initial snapshots , using the value of coulomb logarithm @xmath147 ( calibrated against the energy diffusion rate measured in the @xmath4-body simulation ) . @xmath148 figure [ fig : comparison_nb_mc_fp ] shows the comparison between @xmath4-body and monte - carlo codes for the three geometries . we considered the time dependence of the number of captured particles and their distribution in semimajor axis @xmath15 . the agreement between two codes is fairly good , with the @xmath4-body simulation giving slightly larger number of captured stars for @xmath149 ; this enhancement may be a result of resonant relaxation , which is not accounted for in the monte - carlo code , or large - angle deflections , which are also neglected . we also show the change in @xmath150 during one radial period just before capture ; as expected , this change is larger for non - spherical models , and the transition between empty and full loss cone regimes occurs well within @xmath6 for most realistic parameters @xcite . we note that relativistic precession starts playing role for radii smaller than roughly the radius of transition to the full loss cone regime @xcite , and therefore our neglect of relativity does not lead to noticeable errors in the overall capture rate . as a final remark , the monte - carlo simulation of this scale takes only half an hour on a typical multi - core workstation , while the @xmath4-body simulation takes approximately 10 days with a modern gpu . @xmath151 @xmath152 @xmath153 having tested the monte - carlo code on a `` toy '' model , we now apply it for the models describing realistic galactic nuclei . we take the dehnen density profile with @xmath97 , a black hole mass of either @xmath154 or @xmath155 of the galaxy mass , and axis ratios of @xmath156 and @xmath157 , corresponding to @xmath158 . the orbit composition of models is shown on figure [ fig : model_orbits ] . the fraction of `` orbits of interest '' saucers , pyramids and chaotic orbits is @xmath159% of mass near @xmath6 , but their total mass is lower : the mass of chaotic orbits is @xmath160 of the total galaxy mass and the mass of saucers ( pyramids ) is @xmath161 for the axisymmetric ( triaxial ) models . thanks to the improved mass sampling scheme , the number fraction of these orbits is much larger than their mass fraction . the models are scaled to physical scale using three values of @xmath162 and @xmath163 , capture radius @xmath164 pc and the influence radius @xmath165 pc . the simulation time is @xmath166 yr and the dynamical time at the radius of influence is @xmath167 yr . we do not include physical collisions in our treatment , since for our models the collision timescale is much longer than the hubble time at all radii of interest , although the situation could well be different for less massive black holes in dense nuclei @xcite . likewise , the relaxation time is also long enough to neglect the changes in density profile . we consider a population of equal - mass main - sequence stars with solar mass and radius ; this is surely a crude simplification , as the red giants have much larger physical radii and are more readily torn apart by tidal forces @xcite . nevertheless , the basic model should capture qualitatively the main features of more realistic galactic nuclei . the results of monte - carlo simulations agree well with the qualitative picture described above . the difference between three geometries is modest at low @xmath3 , when even the spherical system is largely in the full loss cone regime , but becomes more pronounced for @xmath71 ( figure [ fig : summary ] ) . the peak of distribution of captured stars in energy corresponds to semimajor axis @xmath168 ( as the orbits of captured stars are almost radial , one may say that these stars come from radii @xmath169 ) . at this energy , the draining time of loss region is @xmath170 yr for the triaxial model and @xmath171 yr for the axisymmetric one , in both cases weakly depending on @xmath3 . thus in spherical and axisymmetric systems the flux at present time is mostly determined by relaxation , while for triaxial systems it is dominated by draining of centrophilic orbits ( mostly pyramids within @xmath6 ) . the difference between the first two comes from the larger size of the effective capture boundary in the axisymmetric case , which enhances the flux by a logarithmic factor of few ( @xmath172 , @xcite ) , but only in the case when most of the flux in the spherical case occurs in the empty loss cone regime . again , one should keep in mind that this refers to the steady - state flux , while in the above figure we plot the solution of the time - dependent fokker - planck equation , in which the difference between spherical and axisymmetric systems is much larger if the evolution time is shorter than @xmath173 ( i.e. , for @xmath174 ) . figure [ fig : captured ] shows the properties of captured stars for a @xmath175 black hole in three geometries . in this case the difference in the total captured mass between spherical and axisymmetric cases is already an order of magnitude , while the triaxial system adds another factor of two . the figure also shows that , contrary to a naive expectation , the flux at almost all energies is substantially lower than the full loss cone flux ; nevertheless , the crude estimate based on a full loss cone rate above the `` draining '' energy ( [ eq : ftotal_triax ] ) is remarkably close to the more refined calculation . chaotic orbits contribute a minor fraction to the captured mass in both axisymmetric and triaxial cases . the figure was plotted for the model with @xmath176 of the total galaxy mass ; the model with @xmath177 has very similar properties ( to within 10% ) but less reliable statistics , as the number of captured orbits is lower by a factor of 3 . an important question about non - spherical models with black holes is whether the scattering of chaotic orbits by the central mass may lead to the elimination or reduction of triaxiality . we do not see any evidence for this , even though the models were evolved for @xmath127 dynamical times at the radius of influence . the opposite conclusion reached in earlier studies @xcite might have resulted from considering evolutionary rather than steady - state models , in which a black hole grew in a pre - existing triaxial cusp . we have also considered a model with a larger deviation from spherical symmetry , namely with axis ratio @xmath178 , same as in @xcite , and the black hole mass of 1% of galaxy mass . however , it turned out to be impossible to create a quasi - isotropic model with these properties : there is no difficulty to arrange for the given shape and density profile , but the region outside a few @xmath6 can not have a uniform distribution of stars in @xmath32 . instead a radially anisotropic system is formed with a scarcity of orbits at intermediate @xmath32 and an excess of them at @xmath179 and even more at @xmath180 . naturally , such system may have a draining rate larger than the full loss cone rate , which might explain the conclusions reached by @xcite . instead , we opted for a model with variable triaxiality : the axis ratio @xmath181 increased from 0.8 to 0.9 between 5 and 7@xmath182 , i.e. outside the region that gives the main contribution to the total flux . as expected , this model yielded a larger captured mass , roughly by a factor 1.5 . we also confirmed the scaling relations derived in [ sec : scaling ] , by performing a simulation with @xmath183 and @xmath184 all scaled by the same factor of 10 . this resulted in a very similar behaviour for axisymmetric and triaxial systems , while somewhat increasing the flux in the spherical case ( due to a logarithmic dependence on @xmath74 ) . we have considered the dynamics of stars in galactic nuclei containing supermassive black holes , with the focus on the rates of capture of stars by the black hole , in particular , in non - spherical nuclei . the supply of stars into the loss cone the region of phase space from where the capture is possible is driven by changes in angular momentum due to two effects : collisional two - body relaxation , which operates in any geometry , and collisionless torques which arise in the case that the potential is non - spherical . we have reviewed the structure of orbits in galactic nuclei and identified the classes of orbits that are important for the loss cone problem saucer orbits in axisymmetric systems and pyramid orbits in triaxial systems , as well as chaotic orbits which exist outside the black hole radius of influence @xmath6 in both cases . these orbits form the loss region , from which a star may get into the loss cone and be captured by the black hole even in the absence of two - body relaxation , due to collisionless torques only . in a triaxial case , chaotic and pyramid orbits are centrophilic , i.e. they may attain arbitrarily low values of angular momentum ; in an axisymmetric case , the loss region is composed of orbits with the conserved component of angular momentum , @xmath34 , being low enough to let them enter the loss cone . in both cases the volume of the loss region is substantially larger than the loss cone proper , thus we expect a higher capture rate due to two factors : ( i ) the stars in the loss region are drained over a timescale much longer than the dynamical time , and ( ii ) two - body relaxation is required only to drive a star into the loss region , instead of a much smaller loss cone , for it to be eventually captured . the interplay between collisonless and collisional effects is poorly handled by present - day @xmath4-body simulations , due to the necessity of a very large number of particles for the collisional effects to be realistically small . to circumvent this trouble , we have developed a monte - carlo method that can simulate these systems with a rather modest number of particles while ensuring a faithful balance between the two effects . it follows the orbits of stars in a smooth potential of any geometry while perturbing them according to the standard two - body relaxation theory . we applied this method to the loss - cone problem in non - spherical galaxies , with the aim of determining the total mass of stars captured by the black hole during the hubble time . we assumed a one - parameter family of models with a @xmath185 density cusp and the scale radius of the model determied by the @xmath90 relation . our main results are the following : * most stars find their way into the black hole from regular orbits inside @xmath6 . in a triaxial system , the timescale for collisionless draining of centrophilic orbits is larger than the galaxy age , unless the triaxiality is very small . thus the bulk of the capture events are not caused by two - body encounters , but are due to the torques from the smooth non - spherical mass distribution . in an axisymmetric system , the draining time is short compared to the galaxy age , but the larger size of the loss region compared to the loss cone proper means that the overall capture rate due to collisional relaxation is increased by a factor of few with respect to the spherical case . the difference between three geometries is larger for larger black holes , being more than a factor of ten for @xmath71 , but only a factor of two for @xmath186 . thus we have confirmed earlier studies ( e.g. * ? ? ? * ; * ? ? ? * ) that have suggested the importance of draining of the loss region for the capture rates in non - spherical galaxies . * the initial draining rate for a system with a uniform distribution in @xmath32 ( i.e. isotropic in velocity space ) is roughly equal to the full loss cone rate . however , as the draining time is strongly dependent on energy , the total draining flux , integrated over all energies , decreases with time , as the most bound orbits are progressively depleted . there is no globally well - defined full loss cone rate , but even the value of full loss cone flux at the radius of influence overestimates the actual present - day flux by a factor of few . a similar conclusion was reached by @xcite for the case of a binary black hole in a triaxial galaxy . the draining of centrophilic orbits in the triaxial system may account for the total captured mass of a few percent of @xmath3 , scaling roughly linearly with the latter and weakly depending on the degree of triaxiality . * in non - spherical systems , stars in the loss cone have a more uniform distribution in periapsis radius than in the spherical case ; in other words , the boundary condition corresponds to the full loss cone regime for most captured orbits . this could have observational implications for the properties of the tidal disruption flares @xcite , although recent studies suggest that the value of impact parameter has little influence on the dynamics of the tidal disruption event @xcite . * chaos in the population of low angular momentum orbits , induced by the central black hole , does not necessarily lead to the loss of triaxiality over the lifetime of the galaxy . our models , set up initially in nearly perfect equilibrium , retained their shape for @xmath127 dynamical times at @xmath6 ( for a @xmath175 black hole ) . these conclusions were confirmed for a limited set of models but are likely to hold for more general cases . the most important assumption in the present study was that the initial distribution of stars was close to isotropic in the angular momentum space , which results in a rather large draining flux , and an increased capture rate due to relaxation , compared to the steady - state value . if a galaxy has an already depleted stellar population at low angular momenta , for instance , due to ejection of these orbits by a binary supermassive black hole @xcite , then the difference between the three geometries is likely to be reduced , and the overall flux will be lower . thus our calculations should be taken rather as upper estimates , which nevertheless predict a rather high capture rate ( @xmath187yr ) for the most massive black holes in non - spherical galaxies . on the other hand , the number of detectable tidal disruption flares for these black holes is much smaller , as for black holes more massive than @xmath175 most of the captured stars are swallowed entirely without producing any flare . figure 15 in @xcite suggests that for @xmath188 the tidal disruption flares of giant stars could constitute less than 5% of total capture events . a more quantitative estimate is beyond the scope of this paper , we only note for more realistic assumptions about stellar structure and evolution there are several other processes that could be important in the context of star black hole interaction , such as `` growth '' of a giant star into the loss cone @xcite , or tidal dissipation induced capture @xcite . it is also worth noting that we considered rather low - density galactic nuclei , in which we could neglect physical collisions between stars and the changes in density profile due to relaxation in energy , both processes occuring on timescales much longer than hubble time . for less massive black holes in denser nuclei ( like the milky way ) the situation is more complicated , but in these systems we do nt expect non - spherical effects to play a significant role in the capture rate . i am grateful to the referees for their valuable comments which helped to improve the presentation . this work was partly supported by the national aeronautics and space administration under grant no . nnx13ag92 g . the software for estimating capture rates from fokker - planck models , including draining of the loss region , is available from http://td.lpi.ru/~eugvas / losscone/.	we consider the problem of star consumption by supermassive black holes in non - spherical ( axisymmetric , triaxial ) galactic nuclei . we review the previous studies of the loss - cone problem and present a novel simulation method which allows to separate out the collisional ( relaxation - related ) and collisionless ( related to non - conservation of angular momentum ) processes and determine their relative importance for the capture rates in different geometries . we show that for black holes more massive than @xmath0 , the enhancement of capture rate in non - spherical galaxies is substantial , with even modest triaxiality being capable of keeping the capture rate at the level of few percent of black hole mass per hubble time .
determining distances by use of multiple methods has a long and distinguished history in astronomy from antiquity to the present day . aristarchus of samos first determined the moon - earth distance from the lunar eclipse . a century later , hipparchus checked aristarchus values using the independent method of terrestrial parallax : the position of the lunar limb during solar eclipse as seen from alexandria and hellespont . more recently , astronomers have demanded that multiple methods be employed for zeroing in on the precise parameters governing the currently observed and mysterious accelerated expansion of the universe @xcite . pulsating variables have enjoyed a privileged role in the local volume . due to their characteristic light curves and relatively bright absolute magnitudes compared to the bulk of main sequence stars , they can easily be identified and readily measured in nearby galaxies . with a local zeropoint for these systems , it is straightforward , modulo metallicity and reddening issues , to determine the distances to external systems . cepheids , and their more common , albeit fainter , relatives within the instability strip , the x ( rrl ) variables , have been two key elements in the historical endeavor to launch humanity out of the solar system and milky way and into the local cosmos . indeed the cepheids currently serve as the anchor of the cosmological distance scale , having allowed the most precise measurement the local rate of expansion of the universe , @xmath8 , to date ( freedman et al . 1994 , freedman et al . 2001 ) . in principle , once properly calibrated , all stars of known absolute magnitude should yield identical measurements of the distances to nearby galaxies . this has not historically been the case . in particular , the rrl @xmath9 calibration has varied by almost half a magnitude depending on the method adopted , and , therefore , consistency with the cepheid distance scale as well as other distance metrics has been difficult to establish . as a result , attempts to determine rrl absolute magnitudes were largely abandoned over the past decade with a few notable exceptions ( e.g. , dambis et al . 2009 ) . however , recently there have been new efforts . benedict et al . ( 2011 ) used hst trigonometric parallaxes of 5 rrl stars ( including 4 rrab and 1 rrc variable ) to obtain an average @xmath10 . klein et al . ( 2011 ) recently used mid - ir data from the wise satellite to infer a mid - ir period - luminosity relation . in this work , we present a third measurement of rrl absolute magnitudes using the method of statistical parallax ( s@xmath11 ) . the large number of rrl that have been discovered in the last decade allow us to make a fresh assault on this issue . historically , rrl star distance measurements from s@xmath11 have come in systematically shorter than other distance indicators , in particular cepheids ( barnes & hawley 1986 , hawley et al . 1986 , layden et al . ( hereafter , l96 ) , popowski & gould 1998a , popowski & gould 1998b , gould & popowski 1998 ( hereafter , collectively pg@xmath12 ) , dambis et al . 2009 ) . it is not yet fully understood why either this method or these two classes of objects should yield different distances to the same galaxies . thanks to automated synoptic all - sky surveys like the all sky automated survey ( asas , pojmanski 2002 ) , the number of rrl stars that have reliable light curves has increased by a factor @xmath135 relative to the previous state - of - the - art " . the asas program has identified approximately 2000 rrl stars with 300 - 500 epochs of photometry . by obtaining high - resolution spectra for these targets , we can both measure the radial velocities and metallicities needed for s@xmath11 and address outstanding issues of systematics . the light curves allow us to accurately determine pulsation phases and permit the measurement of the radial velocity at a single fiducial phase at which the pulsation velocity equals the star s systemic velocity ( see kollmeier et al . traditionally , obtaining the radial velocity component for rrl was laborious , requiring multiple epochs of spectroscopic observation . the determination of the phase - velocity relationship for rrabc variables ( liu 1991 , kollmeier et al . 2009 , preston et al . 2011 ) allows a far more efficient strategy for obtaining critical radial velocity information with which to compute s@xmath11 as we discuss further in section 3 . historically , rrc variables have been either excluded from s@xmath11 analyses or only approximately analyzed . this is primarily due to two factors . first , their hotter temperatures make it more challenging to determine abundances from low - resolution , low signal - to - noise ratio ( snr ) spectra ( see layden 1994 ) and , as a result , these objects can not be robustly classified by population ( halo / disk ) as required by modern s@xmath11 . however , high - resolution echelle observations circumvent this issue and allow , for the first time , a definitive s@xmath11 analysis from rrc variables alone . second , there are fewer rrcs relative to rrabs , and it is only now that samples are large enough to perform a robust , self - consistent , pure rrc s@xmath11 analysis . we analyze our full ( rrab + rrc ) sample in a future work ( kollmeier et al . in preparation ) and restrict our attention here to our rrc sample . in section 2 we present a brief overview of s@xmath11 to remind the reader of the basic principles of the technique . in section 3 we present our sample selection , observations , data reduction , and analysis methods . in section 4 we review our updated methodology for determining s@xmath11 , the results of which are discussed in section 5 and compared to previous s@xmath11 results in section 6 . finally , in section 7 we discuss our results in light of recent and historical works on the absolute magnitude scale of rrl variables . the basic principle that underlies statistical parallax is that the absolute magnitude of any stellar population characterized by a particular velocity ellipsoid should have a single true value when derived from either transverse or radial kinematics of that tracer population . the radial velocity ( rv ) determination of the ellipsoid is independent of any assumptions about distance , but the transverse determination requires an assumed value of the absolute magnitude . the rvs alone yield values for the the velocity ellipsoid in units of @xmath14 . the proper motions , when scaled by the square root of the flux ( @xmath15 ) , yield values for the velocity ellipsoid in units of @xmath16 . the ratio of these two is a distance , d , which is the distance that an rrl star would have to be in order to have @xmath17 . and hence , the absolute magnitude of the rrl stars is @xmath18 . the basic formulae for computing statistical parallax in the presence of observational errors have been well - established ( clube & dawe 1980a , b ; murray 1983 ; hawley et al . 1986 ; strugnell , reid , & murray 1986 , pg@xmath12 ) . the method involves a 10-parameter maximum likelihood fit to the kinematic and photometric data : the distance scale @xmath19 , the 3 first moments ( `` bulk velocity '' ) @xmath20 of the sample velocity distribution , and the 6 independent second moments , @xmath21 . the key parameter of interest for distance scale determination is the value of the true versus assumed fiducial absolute magnitude of the tracer population . @xmath22 the error in determining this distance scaling is given by ( popowski & gould 1998a ) : @xmath23 where @xmath24 and the last expression is for the ( more intuitive ) isotropic case @xmath25 , @xmath26 is the bulk motion , @xmath27 is the velocity dispersion , and @xmath28 where @xmath29 is the number of stars with accurate rv measurements and @xmath30 is the number of stars with accurate proper motion measurements . the evaluation in equation ( [ eqn : spi_err ] ) is for @xmath31 , which is typical of halo rrl samples . the distance scale accuracy improves as @xmath32 , i.e. , the sample of stars with _ accurate _ proper motions and radial velocities is increased in size . asas has provided a public catalog of rrl stars identified via high - cadence light curve analysis @xcite . in addition to photometry from the asas survey , these stars have proper motions from the second usno ccd astrograph catalog ( ucac-2 ; zacharias et al . 2004 ) , which covers the entire southern hemisphere . radial velocity information is not published for the majority of the southern sky and what is therefore required are spectroscopically determined velocities for this large sample of objects that has corresponding proper motions and photometry . before beginning our extensive observational program , we evaluated the suitability of the asas sample for measuring s@xmath11 considering known systematic uncertainties . we discuss each of these below . the key underlying assumption of s@xmath11 is that the rrl are a faithful tracer population of a single velocity ellipsoid and do not exhibit poorly mixed kinematics , for example , from coherent stellar streams . clearly , as one goes to very large distance ( i.e. , the outer halo ) this assumption breaks down as rrl populations become increasingly contaminated by satellite debris on non - mixed orbits . indeed , this consideration alone prevents us from extending the survey to extremely faint magnitudes where the distances probed preferentially contain such kinematics ( e.g. kollmeier et al . the distribution of distances for the asas rrc sample is shown in figure [ fig : distances ] . as can be seen , the majority of our sample lies within a distance of 4@xmath33 . the solar neighborhood on these scales is known to be sufficiently smooth that our statistics are unaffected by kinematic substructure ( e.g. , gould 2003 ) . the proper motion errors from ucac-2 are typically 5 - 6 @xmath16 for the majority of the sample corresponding to @xmath34 . at first sight , the proper motion errors may seem too large to be useful , given that the median distance is about @xmath35kpc ( see fig . [ fig : distances ] , determined using an initial value for the absolute magnitude scale of @xmath36 ) , and the direction - averaged dispersion is about @xmath37 . together , these imply that the errors are typically a large fraction of the quantity being measured . in fact , s@xmath11 is extremely robust with respect to observational errors . equation ( 17 ) of popowski & gould ( 1998a ) ( hereafter pg98a ) shows that the statistical error should increase by _ only a few percent _ relative to equation ( [ eqn : spi_err ] ) , given these errors . we verify that this is actually the case in section [ sec : results ] . equation ( 18 ) of pg98a shows that one must be careful about systematic errors that can result from _ misestimating _ the size of these errors . indeed , the s@xmath11 methodology contains an internal check on the proper motion errors . since s@xmath11 uses apparent magnitudes , it is important to have robust extinction measurements for objects in the sample . we use the reddenings provided by schlegel , finkbeiner & davis ( sfd , 1998 ) and a standard extinction coefficient , @xmath38 to determine @xmath39 . of @xmath40 . we adopt the sfd values until a definitive calibration is agreed upon . ] because many of our sources are at high galactic latitude ( see figure [ fig : skydist ] ) , this directly gives adequate values for these sources . we adjust the reported reddening values to account for the finite distance from the plane of our objects . in some instances , our objects are too close on the sky with other field stars in projection to yield reliable asas photometry ( see pojmanski ( 2002 ) for details of asas photometry ) . these objects are eliminated from our analysis . with secure photometry and transverse kinematic measurements in hand , we began our spectroscopic survey . the observations presented here were made in 2011 and 2012 with the echelle spectrograph mounted on the 2.5 m du pont telescope at las campanas observatory as part of the larger carnegie rrl survey ( carrs ; kollmeier et al . in preparation ) . upon completion , our survey will have moderate signal - to - noise ratio spectra for approximately 1200 rrl stars observable from the southern hemisphere . the sky distribution of our full survey sample is shown in figure [ fig : skydist ] . the grey points show rrab stars and the blue points show rrc stars . observations were designed to reach snr@xmath41 in the order containing the mg i triplet at 5170 at the target phase . exposures for a given target bracketed observations of a thar lamp through a 1.5x4-arcsec slit . the photospheric velocity of rrl stars changes by many @xmath14 over the course of a pulsation cycle . traditionally , one has taken multiple observations through the pulsation cycle and fit the resultant velocity function ( when the phase is known ) in order to obtain the systemic velocity of the target ( e.g. layden 1994 ) . we adopt the time - averaged velocity of the pulsation curve as the center - of - mass velocity of the star as detailed in @xcite and @xcite for rrabs . integration of detailed velocity curves of the rrc variables t sex , tv boo , dh peg , and yz cap , all measured with the du pont echelle , shows that the pulsation velocity is equal to the star s time - average velocity at phase 0.32 , reckoned relative to maximum light . our observations were all made as close to this phase as possible , and velocity corrections were applied adopting the following correction : @xmath42 figure [ fig : phases ] shows the resultant phase distribution for our targets . note that these velocity corrections are no larger than @xmath43 for any given star . in order to ensure uniform quality across our sizeable database , a data reduction pipeline was constructed following kelson ( 2003 ) . the pipeline divides the reduction process into first processing the calibration frames , and then using these to reduce the science frames and extract spectra . the calibration stages involve processing the frames for flat - fielding , determining the order edges and y - distortion using sky flats , and obtaining wavelength solutions using a th - ar lamp frame . once these frames are produced , the science spectra are reduced . the science spectra are first divided by the processed and normalized flat - field and then wavelength calibrated . the spectra are then sky - subtracted and the spatial profiles of the resultant 2-d spectra are used to find extraction apertures . the pipeline takes advantage of parallel processing , which speeds up the reduction process significantly for a large number of objects . post - extraction processing of the spectra was done with the iraf echelle package . radial velocities were measured relative to a high snr template of the metal - poor star hd140283 using the iraf fxcor routine . the template spectrum was taken with the du pont echelle spectrograph in the same configuration as survey stars . the heliocentric radial velocity for hd140283 has been determined , independently in four separate high - resolution studies , to be @xmath44 ( e.g. tsangarides et al . 2004 , aoki et al . 2002 , lucatello et al . 2005 , latham et al . the cross correlations were made on the wavelength interval 4900 5500 . objects that were originally classified as rrl stars based on their light curves but , upon inspection of their spectra , were found to be binaries or other non - rrl variables were removed from the sample . the radial velocity error estimates returned by fxcor are extremely small , so we estimate the true errors by making repeat measurements of the _ corrected _ velocities on a subset of 11 stars . from these we determine @xmath45 . we add this in quadrature to our mostly tiny formal errors . we note that , from pg@xmath46 , the velocity errors enter the final result as @xmath47 and so have no practical effect in any case . we nevertheless include this correction for completeness . the measurement of absolute abundances for large samples of rrc stars has not been undertaken previously . we therefore must calibrate our pipeline abundance measurements ( which yield consistent relative abundance determinations ) to the scant data currently available . based on analysis of the high snr du pont echelle spectrum of rrc variable yz cap , we adopted a single set of atmospheric parameters ( @xmath487000k , @xmath49 , @xmath50 , [ @xmath51/fe]=+0.35 ) for all survey stars . the value of @xmath52 is somewhat constrained by the small area occupied by rrc stars in the color - magnitude diagram , and our survey spectra are not of sufficiently high snr to determine this parameter independently . the abundance measurements in rrl stars are relatively insensitive to the adopted values for @xmath53 and @xmath54 . for our final calibration , we rely primarily on a detailed study ( govea et al . in preparation ) that examines the sensitivity of rrc abundance determinations with respect to stellar parameters using a small sample of high snr spectra also obtained with the du pont echelle . this is the only study we know of in which the sensitivity of rrc abundance measurements are systematically studied . we also compare to the known distribution of rrab metallicities , as a secondary calibration . we defer a more detailed analysis of this metallicity calibration to a future work where more high snr , high - resolution data are available for comparison ( kollmeier et al . in preparation ) . pipeline abundance measurements were performed by fitting to a grid of synthetic spectra generated by moog . spectra reduced by our spectroscopic pipeline were cleaned of remaining noise spikes and smoothed using a 3-pixel boxcar filter . owing to the relatively broad lines in rrl stars , this smoothing has essentially no effect on the abundance determination . rrl stars are warmer than the sun and generally metal - poor and therefore contain few spectral features beyond 5500 . as our snr is maximized between approximately 4400 and 5500 , we therefore used two broad spectral ranges for our synthetic fits : 4400 - 4680 and 5150 - 5450 . the latter region contains primarily neutral - species transitions as well as the mg i b lines . the former region has numerous singly - ionized species including ba ii 4554 . the synthetic spectral grid covers a range of effective temperature ( @xmath55 ) , surface gravity ( @xmath53 ) , metal abundance and microturbulent velocity ( @xmath54 ) . the atomic line lists for the syntheses were begun with the kurucz line database , then refined until a good match was achieved for the spectrum of the sun . the smoothed synthetic spectra were compared with the observed spectra and the parameters of those with lowest @xmath56 were chosen as fits . in figure [ fig : abundances ] we show a histogram of our pipeline - derived and calibrated abundances averaging the ( independently ) derived abundances for the two broad spectral windows . determinations for our sample of 247 rrc variables . abundances were determined by averaging two broad spectral windows which had maximal signal to noise as well as numerous spectral features . , width=336 ] with the kinematic and chemical information measured as described above , we are prepared to segregate our sample into the halo and disk populations . we convert our proper motions and radial velocities to radial , rotational , and vertical velocities ( @xmath57 ) as per l96 . we correct these velocities for the dynamical solar motion ( + 9 , + 250 , + 7 @xmath14 ) and we rotate the sun - centric ( @xmath57 ) 3-space velocity to the local frame of the star assuming cylindrical symmetry of the galaxy . the cumulative distribution of these rotation angles is shown in figure [ fig : angles ] . note that they span the range of roughly @xmath58 . the resultant velocities @xmath59 for our sample are shown in the left panels of figure [ fig : disk_halo ] for direct comparison with figure 3 of l96 which shows the same quantities for their sample of rrab stars . similarly , the right hand panel of figure [ fig : disk_halo ] shows the rotational velocity component as a function of measured abundance for direct comparison with figure 4 of l96 . it is important to remind the reader again that we are using rrc stars rather than rrab stars as tracers of the velocity ellipsoid , in contrast to l96 . despite this difference in tracer population , the kinematics of this sample are similar to those of l96 . while it is not possible to know with certainty whether any particular star is a member of the disk or halo , we can clearly see a track of stars that has chemical and kinematic properties similar to a disk - like population . these stars are relatively high metallicity and are rotating in the direction of the sun with similar speed . the exact demarcation of the disk / halo line is necessarily uncertain . following l96 , we therefore perform our analysis on subsamples with the aim of determining how sensitive our results are to these uncertainties . as we will show , they are not . we identify two halo / disk subsamples . the first ( halo-1/disk-1 ) considers all stars rightward ( upward ) of the line @xmath60 } - 225 $ ] to be members of the disk / thick disk and they are excluded from the halo population of interest . of our 247 objects , this procedure assigns 34 to the disk and 213 to the halo . these objects are shown by the red squares in figure [ fig : disk_halo ] . the second ( halo-2/disk-2 ) considers all stars with [ fe / h]@xmath61 _ and _ @xmath62 to be disk stars . objects satisfying these criteria are shown as blue triangles in figure [ fig : disk_halo ] . of our 247 objects , this classification eliminates 30 leaving 217 for our s@xmath11 analysis . we note that our halo-1/disk-1 and halo-2/disk-2 designation is slightly modified from l96 given our sample . however , our results are relatively insensitive to these distinctions , although our errors will necessarily scale with the square root of the number of retained objects . and @xmath63 , that is , the velocities in the radial direction , the direction of galactic rotation , and toward the galactic pole respectively . these velocities are corrected for rotation w.r.t . @xmath64 the right panel shows the rotational velocity component as a function of [ fe / h ] . in each panel , red squares denote the disk-1 " population and blue triangles show the disk-2 " population . these stars are eliminated from our ml analysis as described in the text . , width=336 ] we carry out a 10-parameter markov chain monte carlo ( mcmc ) maximum - likelihood fit to the data . these include the nine kinematic parameters describing the velocity ellipsoid ( three diagonal elements corresponding to the bulk motions and 6 dispersions ) and one parameter describing the absolute magnitude scaling @xmath19 . table [ tab : results ] gives the values and errors of the 10 s@xmath11 parameters for our samples . for reasons explained in appendix a , their errors are almost perfectly gaussian . therefore the 10-dimensional likelihood surface is completely specified by these numbers , together with the correlation coefficients , which are also given in appendix a. the fact that the parameters derived from these two different fits are consistent at well below @xmath65 shows that our choice of disk / halo separation does not significantly affect our results . c r r r r r r r r r r halo-1 & 1.007 & 11.8 & 32.5 & 8.3 & 155.6 & 102.4 & 93.8 & 0.037 & -0.006 & -0.114 + - & 0.048 & 11.1 & 9.9 & 6.9 & 10.2 & 6.0 & 5.6 & 0.075 & 0.075 & 0.076 + analytic - err & 0.045 & 10.7 & 9.6 & 6.4 & 8.9 & 5.8 & 5.3 & 0.069 & 0.069 & 0.069 + halo-2 & 1.002 & 10.1 & 37.4 & 6.1 & 153.8 & 104.8 & 93.8 & 0.018 & 0.008 & -0.144 + - & 0.048 & 10.9 & 10.0 & 6.8 & 10.1 & 6.1 & 5.6 & 0.074 & 0.075 & 0.075 + disk-1 & 0.728 & 4.0 & 217.1 & -24.9 & 62.1 & 42.6 & 49.4 & 0.000 & -0.192 & 0.008 + - & 0.142 & 11.3 & 9.5 & 10.8 & 10.2 & 8.7 & 12.8 & 0.198 & 0.193 & 0.207 + disk-2 & 0.796 & 13.1 & 209.3 & -19.2 & 64.9 & 56.5 & 53.1 & -0.020 & -0.397 & -0.088 + - & 0.158 & 12.6 & 12.9 & 11.7 & 11.2 & 12.2 & 14.1 & 0.203 & 0.190 & 0.207 + for our halo-1 sample , we obtain a value of the absolute magnitude of rrcs of @xmath66 and for our halo-2 sample , this value is @xmath67 . we discuss these values in the context of previous measurements in section [ sec : comp ] . in the analysis reported above , we adopted the proper - motion error estimates in the ucac-2 catalog . as discussed by pg98a , if the size of these error estimates were systematically too big , it would induce a systematic error in our distance - scale estimate , making the rrl appear to be systematically further ( hence more luminous ) than they actually are . this is a potential concern because zacharias et al . ( 2004 ) report external tests only on their mean proper motion estimates and not on their error estimates . fortunately , we are able to perform an internal test on these error estimates . this test shows first that the ucac-2 error estimates are probably overestimated by about 25% , and second that this level of mis - estimation has almost no effect on our results . we note that our rv errors are too small to have any effect , whether correctly estimated or not . they will therefore be ignored in the following discussion . we first review the basic physics that permit such a test and then present results . what is now called `` statistical parallax '' was formerly ( in the first half of the last century ) divided into two effects . one , called `` secular parallax '' , compared the mean bulk motion ( @xmath20 ) in rv with that in ( flux - scaled ) proper motions . the other , also called `` statistical parallax '' compared the amplitude of the dispersions ( @xmath68 ) in rv and proper motions . in modern s@xmath11 , these are done simultaneously in a single fit . however , the errors in the proper - motion error estimates enter very differently into these two components . for the @xmath68 comparison an overestimate of the errors leads to a corresponding underestimate of the intrinsic proper motions of the stars , which can be reconciled with the dispersions measured in rv only by placing them further away . however , for the @xmath20 measurement , which is first - order ( not second - order ) in the proper motions , there is no such effect . if one wished to be completely safe from any such error - bar misestimates , one could in principle choose to just compute the `` secular parallax '' . however , the maximum likelihood formulation of s@xmath11 permits a more sophisticated approach . the likelihood function is a direct and sensitive indicator of the consistency of the two components of the s@xmath11 determination when the error bars are systematically rescaled . table [ tab : rescale ] shows results for several such rescalings . the log - likelihood is maximized for a rescaling factor @xmath69 , while @xmath70 yields a log - likelihood that is lower than this @xmath71 . the probability of this occurring by chance is about @xmath72 . while this is certainly not an impossible statistical fluctuation , neither is there any compelling prior that the ucac-2 errors are exactly as given . we note that at @xmath73 , @xmath74 , which is very plausibly due to statistical fluctuations . because we do have some prior that these errors are not grossly overestimated , and because there is no statistical evidence that @xmath75 , we adopt this as our best estimate of the errors . we note from table 1 that the result of changing @xmath76 from 1.0 to 0.8 is to reduce the distance scale by approximately 1% , which is about 20% of our statistical error . the remaining parameter estimates barely change as well . c r r r r r r r r r r r 1.0 & 3.91 & 1.007 & 11.8 & 32.5 & 8.3 & 155.6 & 102.4 & 93.8 & 0.037 & -0.006 & -0.114 + 0.8 & 1.51 & 0.994 & 12.0 & 34.1 & 8.2 & 155.8 & 102.9 & 94.4 & 0.033 & -0.003 & -0.106 + 0.7 & 0.74 & 0.987 & 12.1 & 35.0 & 8.2 & 155.8 & 103.1 & 94.8 & 0.031 & -0.002 & -0.102 + 0.6 & 0.30 & 0.981 & 12.2 & 35.7 & 8.2 & 156.0 & 103.3 & 95.4 & 0.029 & -0.001 & -0.098 + 0.5 & 0.09 & 0.974 & 12.4 & 36.6 & 8.1 & 155.9 & 103.4 & 95.8 & 0.028 & 0.000 & -0.094 + because the sample is magnitude limited and the target population has an intrinsic dispersion in absolute magnitude , @xmath77 , the sample contains more stars with brighter - than - average luminosity than fainter - than - average . this leads to a correction @xmath78 we modify the pg98a estimate to obtain @xmath79 as follows . pg98a began with the observed dispersion in the large magellanic cloud ( lmc ) of 0.17 mag . they estimated that 0.09 mag was due geometric dispersion , and so estimated an intrinsic dispersion of 0.14 mag . we further note that for an estimated metallicity dispersion of 0.5 mag , and assuming a slope of @xmath80 mag dex@xmath81 in the rrl @xmath9-[fe / h ] relation , an additional 0.1 mag of scatter can be accounted for by metallicity variation . subtracting this in quadrature , we obtain 0.1 mag intrinsic dispersion . this leads to a correction @xmath82 . note that pg98a identified another effect due to the dispersion of absolute magnitudes , which scales @xmath83 . we briefly describe this effect and why it is proper _ not _ to include it here . if stars do not have exactly the same absolute magnitude , then their mean square scaled ( by flux ) proper motions will be greater than their squared scaled ( by mean flux ) proper motions . this will mimic a larger proper - motion dispersion and so cause the stars to appear closer ( so dimmer ) than they actually are . however , because the mechanism of this effect is similar to that caused by misestimation of the proper - motion errors ( section [ sec : ucacpm ] ) , this small correction is already absorbed into the correction adopted there . to obtain our final result , we incorporate the small corrections for malmquist bias and for probable overestimation of the ucac-2 proper motion errors and take the average between our two halo samples , @xmath84 } \rangle = -1.59 \label{eqn : mvfinal}\ ] ] the systematic error ( 0.031 mag ) is determined by combining in quadrature the variance from different definitions of the halo ( 0.010 mag ) , uncertainty in the degree of overestimation of the proper motion errors ( 0.025 mag ) , and uncertainty in the extinction scale ( 0.014 mag ) . there is also a systematic error of about 0.05 dex in the metallicity scale , and so the mean metallicity at which this estimate is valid . because the statistical errors are about 3.5 times larger than the systematic errors , a detailed investigation of the latter is not warranted at the present time this issue will become somewhat more pressing when we combine the analysis of carrs rrab and rrc stars . table 1 shows a comparison between the errors derived from the the mcmc and those predicted analytically using the formulae of pg98a , which assume zero measurement errors and isotropic sky coverage . the actual errors are only slightly larger than the ideal errors despite the relatively large proper motion errors . in appendix a , we derive the corresponding analytic estimates of the correlation coefficients among the 10 s@xmath11 parameters and show that these are in good agreement with the mcmc determinations . we can run our maximum likelihood machinery just as easily on a pure disk population as we can on a pure halo population , albeit with the substantially reduced numbers in our disk sample . we perform this analysis and include the results in table [ tab : results ] . interestingly , the value we obtain for absolute magnitude ( averaged between the two `` disk '' definitions ) is @xmath85 at [ fe / h]@xmath86 . after taking account of the slope @xmath87 in the @xmath9-[fe / h ] relation , this is consistent at the @xmath65 level with our result for the halo sample . also of interest are the detailed values of the velocity ellipsoid for the disk population : @xmath6 with dispersions @xmath7 . indeed , it is thick - disklike , which gives us confidence that we are in fact removing a kinematically distinct population via our chemodynamical criteria . since the work of clube & dawe ( 1980a , b ) there have been multiple attempts to derive the rrl absolute magnitude scale from s@xmath11 ( e.g. hawley et al . ( 1986 ) , strugnell , reid & murray ( 1986 ) , layden et al . ( 1996 ) , pg@xmath12 , luri et al . ( 1998 ) , dambis ( 2009 ) ) . the most recent s@xmath11 estimate of @xmath88 for a sample of 364 galactic rrl stars in the northern hemisphere from targets in the beers ( 2000 ) catalog using 2mass photometry to correct the infrared - inferred period - luminosity relation for rrl stars , finding @xmath89 . this work does not estimate @xmath88 and can thus not be directly compared here . ] ( and so directly comparable to the work presented here ) is by pg@xmath12 who found @xmath90 } \rangle = -1.6\ ] ] from a sample of 182 halo rrab stars taken from layden et al . ( 1996 ) . ] benedict et al . ( 2011 ) used hst to obtain trigonometric parallaxes for 5 rrl ( 4 rrab and 1 rrc ) . combining their reported values and measurement errors ( together with their adopted intrinsic dispersion of @xmath91 ) yields @xmath92 } \rangle = -1.50\ ] ] while , using @xmath93 ( adopted here ) yields @xmath94 at @xmath95 } \rangle = -1.52 $ ] . to obtain these results , we weight the individual measured magnitude offsets from the mean by @xmath96}^2)^{-1}$ ] , where @xmath80 is the slope of the @xmath9-[fe / h ] relation . note that benedict et al . ( 2011 ) incorrectly quote somewhat smaller errors . see appendix b. efforts to measure @xmath9 from rrc stars have been much more limited . hawley et al . ( 1986 ) and strugnell , reid & murray ( 1986 ) obtained s@xmath11 results from 17 and 26 candidate rrc stars respectively , but were forced to make a restricted analysis because of small number statistics , which they reported only for completeness " . in addition , as mentioned above , benedict et al . ( 2011 ) obtained @xmath97 for a single rrc , rz cep at [ fe / h]=-1.77 . our results on the kinematics of the halo ( @xmath4 with dispersions @xmath5 ) are in excellent agreement with previous work . comparing with pg@xmath12 , our results are consistent at well under 1-@xmath27 despite an entirely different sample . comparing with the most recent ( and largest ) s@xmath11 analysis of dambis et al . ( 2009 ) our results are also generally in excellent agreement , as almost all parameters of the halo _ and _ disk velocity ellipsoids agree within 1 or 2-@xmath27 with the exception of vertical velocity dispersions of the halo which are in tension at the 3-@xmath27 level . we do not know the origin of this discrepancy , but will have more leverage to investigate this with our larger rrab sample . we have performed the first decisive analysis of the absolute magnitude for rrc variables via statistical parallax using the first data from the carnegie rr lyrae survey ( carrs ) . our current measurements for rrc variables yield a 5% distance error which is similar to that obtained from modern techniques applied to rrab samples . we find a velocity ellipsoid for our disk and halo population that is in good agreement with previous measurements . already , carrs provides competitive distance accuracy to other surveys and techniques . at the conclusion of carrs , we anticipate a factor of @xmath98 increase in the number of tracers and , consequently , 2% distance errors . in future work , we will analyze this far larger database and have the statistical potency to divide our sample into finer metallicity and kinematic bins than can be done presently . this will allow a precision measurement for comparison with other techniques with the hope of a unified " rrl distance scale . in the gaia era , where space - based parallaxes will be available for many of these objects , our database of high - resolution spectra should provide useful complementary information for going beyond distances and gaining further understanding of rrl as astrophysical objects , rather than merely as test particles . in the main text , we showed that the analytic error estimates derived by pg98a for the 10 s@xmath11 parameters closely approximate the numerical errors in our sample , despite the fact that the analytic treatment assumes zero measurement errors , uniform sky coverage , and no rotation with galactic position . here , we extend the pg98a analytic treatment to the off - diagonal elements of the parameter determinations and compare the resulting correlation coefficients to our numerical values . integrating equations ( 30)-(35 ) from pg98a over @xmath99 objects uniformly distributed on the sky , we obtain the inverse covariance matrix of the 10 s@xmath11 parameters @xmath100 @xmath101 @xmath102 where @xmath103 @xmath104 and where we have adopted a reference frame in which @xmath105 is diagonal . ( see equations ( 37)-(42 ) of pg98a . ) matrices of this form can be analytically inverted , @xmath106 as @xmath107 pg98a ( equations ( 43)-(46 ) ) have already evaluated the diagonal elements of this matrix ( i.e. , the variances of the parameters ) @xmath108 where @xmath109 . here we use equation([eqn : cov ] ) to evaluate the off - diagonal elements , or equivalently the correlation coefficients @xmath110 @xmath111 where @xmath112 and all other terms vanish . note that in making these evaluations , one must return to the `` sun frame '' , i.e. add @xmath113 back to the @xmath20 . in the matrix below , we compare the analytic correlation coefficients ( above diagonal ) to the actual ones ( below diagonal ) . there is good overall agreement . we note that the errors in these parameters are almost perfectly gaussian , and therefore completely described by their means and covariances . this can be seen as follows . in any linear fit , the parameter estimates , @xmath115 can be written as a linear function of the data @xmath116 : @xmath117 , @xmath118 , where @xmath119 are the trial functions , @xmath120 is the inverse covariance matrix of the data , @xmath106 , and @xmath121 . therefore if the data @xmath116 have gaussian errors , then so do the parameters @xmath115 . note that this statement does not depend in any way on the central limit theorem , as is sometimes supposed , but only on the fact that linear combinations of gaussians are gaussians . in the present problem , @xmath122 looks similar in structure to a standard linear - fit @xmath56 except , first , the parameters appear in non - linear combinations , second , the second moments of the velocity ellipsoid appear in an additional term containing a log determinant , and third , the scaling factor @xmath19 also appears in a log term . nevertheless , the linearized problem ( treated explicitly by pg98a ) is extremely close in structure to a standard linear - fit @xmath56 . therefore , one expects similar mathematical properties . this is the underlying reason that the pg98a linearized analysis matches numerical results so closely . here we derive the appropriate weighting scheme for an ensemble of rrl parallax measurements with different errors in both @xmath9 and [ fe / h ] . let us consider @xmath123 stars with measured absolute magnitudes @xmath124 and errors @xmath125 and each with perfectly known metallicity @xmath126 . and let us initially assume that the slope of the @xmath127 relation is known but the zero - point is not . that is , where @xmath129 is some arbitrarily chosen fiducial metallicity , @xmath130 is the known slope and @xmath131 is the unknown zero point . then @xmath132 this @xmath56 is minimized by setting its derivative to zero , i.e. , @xmath133 we now choose the `` arbitrary '' fiducial metallicity so that the relation is independent of the choice of @xmath130 . this can be done if we choose @xmath134 in which case @xmath56 is minimized at @xmath135 regardless of @xmath130 . the error in this estimate is the point at which @xmath56 rises by unity , i.e. , @xmath136 hence , equation([eqn : q ] ) gives the effective metallicity at which the measurement is made . now let us suppose that the metallicities are given to us with error bars @xmath137 , so that we can consider simultaneously fitting for both the zero - point ( now called @xmath138 ) and the five true metallicities , called @xmath115 . we can write @xmath56 @xmath139 setting the @xmath140 derivatives of @xmath56 to zero ( wrt to @xmath141 ) yields the equation @xmath142 where @xmath143 @xmath144 and @xmath145 the inverse of this matrix , @xmath106 is given by equation([eqn : cov ] ) , which then allows us to evaluate @xmath146 , @xmath147\ ] ] this simplfies to @xmath148 where we note that @xmath149 is the error in @xmath138 ( e.g. , gould 2003a ) . this looks identical to our previous expression that ignored the @xmath150 errors , except the inverse weights are increased fractionally by @xmath151 . this result formally confirms one s naive idea that the metallicity error is `` equivalent '' to an additional error in the absolute magnitude , propagated by the slope @xmath152 of the relation . note , however , that in contrast to the case of perfectly known metallicities , one can not enforce complete independence of the result from choice of @xmath130 simply by adopting @xmath129 as the average metallicity weighted by @xmath153^{-1}$ ] , since the weighting itself depends on @xmath130 . however , in the present case @xmath154 , so this has no practical impact .	we present the first definitive measurement of the absolute magnitude of rr lyrae c - type variable stars ( rrc ) determined purely from statistical parallax . we use a sample of 247 rrc selected from the all sky automated survey ( asas ) for which high - quality light curves , photometry and proper motions are available . we obtain high - resolution echelle spectra for these objects to determine radial velocities and abundances as part of the carnegie rr lyrae survey ( carrs ) . we find that @xmath0 at a mean metallicity of @xmath1 } = -1.59 $ ] . this is to be compared with previous estimates for rrab stars ( @xmath2 ) and the only _ direct _ measurement of an rrc absolute magnitude ( rz cephei , @xmath3 ) . we find the bulk velocity of the halo to be @xmath4 in the radial , rotational and vertical directions with dispersions @xmath5 . for the disk , we find @xmath6 with dispersions @xmath7 . finally , we suggest that ucac2 proper motion errors may be overestimated by about 25% .
the differential growth of thin elastic sheets can generate highly non - trivial configurations such as the multiple generation of waves along the edge of a torn elastic sheet @xcite and the wrinkled patterns on leaves @xcite . recently , environmentally responsive gel disks of uniform thickness have been created in the laboratory that mimic this type of growth by differentially shrinking in the radial direction when thermally activated in a hot water bath @xcite . the rate of shrinkage can be controlled , allowing for the creation of regions in the disk that shrink at different rates upon activation . rippling occurs in regions that shrink slower than the center of the disk while regions that shrink faster obtain a spherical dome like shape . geometrically , these regions have positive and negative gaussian curvature respectively and by tuning the composition of the gel , different curvatures can be selected . this type of growth can be used to generate a multitude of different shapes of varying geometries , but a mathematical framework that can accurately predict the equilibrium shapes is lacking . classically , thin elastic plates have been modeled by the fppl von krmn ( fvk ) equations that reduce the full three dimensional equations of elasticity to a coupled system of partial differential equations defined on the mid - surface of the plate @xcite . these equations are asymptotically valid in the small thickness regime if the thickness scales in a particular manner with external loadings @xcite , and different scalings lead to a hierarchy of different plate models that includes fvk @xcite . the fvk equations can be obtained as the variation of an energy functional . this energy is the sum of stretching and bending terms such that the ratio of bending to stretching energy scales like the thickness squared @xcite . consequently , thin plates energetically prefer to bend out of the plane rather than stretch a phenomenon easily observed in stretched sheets @xcite and this could be the mechanism used by growing sheets to buckle out of the plane . non - uniform growth is difficult to model using the fvk equations since it commonly generates structures with residual stresses @xcite and thus it is unclear what is an appropriate reference configuration from which strains can be measured . one technique for defining a reference configuration is to specify a two dimensional `` target metric '' @xmath5 on the mid - surface of the sheet from which strains are measured @xcite . the metric @xmath5 is a geometric quantity that locally determines all of the intrinsic properties of an abstract surface such as its gaussian curvature and geodesics @xcite . thus , we are naturally led to consider the problem of finding isometric immersions of @xmath5 into @xmath0 @xcite . but , the problem is that in general isometric immersions of @xmath5 may not exist or be unique and this model does not incorporate important three dimensional effects such as a mechanism for penalizing highly bent configurations . instead of modeling non - uniform growth as isometric immersions , a non- euclidean " elastic energy functional has been recently introduced that incorporates @xmath5 and reduces to the fvk functional when @xmath5 is the identity matrix @xcite . this functional has the same energy scaling as fvk with the stretching energy measuring @xmath1 deviations of the immersion from @xmath5 . as with the fvk functional , if the thickness is sufficiently small the stretching energy dominates and we expect the elastic sheet to bend out of the plane to match the metric . in particular , if @xmath2 isometric immersions exist then in the limit of vanishing thickness minimizers of the non - euclidean fvk functional converge to a minimizer of the bending energy over the class of @xmath2 isometric immersions of @xmath5 @xcite . thus in this model the particular isometric immersion is selected by the bending energy . an alternative is to decompose the deformation of the sheet into a growth tensor and an elastic response , and define the elastic energy in these terms . lewicka , mahadevan and pakzad @xcite have shown that the two approaches are essentially equivalent in the limit that the thickness of the sheet vanishes . in this paper we study isometric immersions of @xmath5 in the disk geometry when @xmath5 specifies a constant negative gaussian curvature @xmath6 . experimentally , disks of constant negative curvature obtain an equilibrium configuration with a periodic profile of one wavelength with the number of nodes increasing with decreasing thickness @xcite . consequently , we will look for low bending energy test functions that match the target metric and have a periodic profile . finding such test functions corresponds to the classical problem of finding isometric immersions of the hyperbolic plane @xmath3 into @xmath0 . hilbert @xcite proved that there are no real analytic isometric immersions of the entire hyperbolic plane into @xmath0 , but this does not exclude the possibility of the existence of isometric immersions with weaker smoothness criteria . in fact , the nash - kuiper theorem states that through the technique of convex integration there exists @xmath7 isometric immersions of @xmath3 @xcite . but , such an immersion would be too rough since the bending energy requires the surface to be at least weakly second differentiable . furthermore , by efimov s theorem , we know that there are no @xmath8 isometric immersions of @xmath3 @xcite . the admissible surfaces with finite elastic fvk energy are @xmath2 and this motivates the search for @xmath2 isometric immersions which lie between @xmath7 and @xmath8 in smoothness . in general finding an isometric immersion of an arbitrary metric is a non - trivial problem . but , by minding s theorem @xcite , which states that any two abstract surfaces having the same constant gaussian curvature are locally isometric , we can cut out subsets of surfaces of constant negative curvature which will be isometric to pieces of @xmath3 . in particular , if @xmath9 is a surface of constant negative curvature and @xmath10 then @xmath11 is isometric to a disk of radius @xmath12 in @xmath3 , provided @xmath13 is smooth . furthermore , it was proven by amsler that the singularities where @xmath9 fails to be smooth form curves on @xmath9 @xcite . these curves form a boundary beyond which no disks locally isometric to @xmath3 can be cut out of the surface . this paper is organized as follows . in sec . [ sec : model ] we present the mathematical model for thin elastic sheets with a specified metric @xmath5 in the form of a variational problem with a fvk type functional . we show that for isometric immersions the functional reduces to the willmore functional and thus in the vanishing thickness limit if an isometric immersion exists the minimizing configurations should converge to a minimizer of the willmore energy over the class of @xmath2 isometric immersions @xcite . in sec . [ sec : hyperbolic ] we review some of the theory of hyperbolic surfaces that will allow us to construct isometric immersions in @xmath0 . in particular , we make use of chebyshev nets ( c - nets ) which are local parameterizations of a surface whose coordinate curves form quadrilaterals on the surface with opposite sides having equal length @xcite . originally , chebyshev used c - nets to describe the arrangement of fibers in clothing and he showed that a sphere must be covered by at least two cuts of cloth @xcite ( see @xcite for a modern translation ) . as evidenced by the sphere , c - nets in general are only local parameterizations of a surface and a global c - net exists if the integral of the magnitude of the gaussian curvature over a quadrilateral is less than @xmath14 @xcite . c - nets are useful parameterizations in that they lead to a natural way to discretize a surface @xcite and projections of c - nets onto paper retain the three dimensional properties of the surface @xcite . remarkably , for surfaces with constant negative curvature , every sufficiently smooth isometric immersion generates a c - net on the surface obtained by taking the asymptotic curves as the coordinate curves and conversely every c - net on a surface of constant negative curvature generates an immersion whose asymptotic curves are the coordinate lines @xcite . the extrinsic and intrinsic geometry of the particular immersion is entirely determined by the angle @xmath15 between the asymptotic curves and a parameterization is a c - net if and only if @xmath15 is a solution of the sine - gordon equation @xcite . thus , solutions of the sine - gordon equation generate surfaces of negative curvature and the corresponding immersion will fail to be smooth when @xmath15 is an integer multiple of @xmath16 . in sec . [ sec : pseudosphere ] we show that isometric immersions of geodesic discs in the hyperbolic plane can by constructed by cutting out subsets of a classical hyperbolic surface , the pseudosphere . this is contrary to the claims in @xcite where the experimentally observed buckling and refinement in wavelength with decreasing thickness is attributed to the non - existence of smooth isometric immersions . these configurations have very large bending energy even away from the singular curve and do not have the rotational @xmath17-fold symmetry observed in experiments @xcite . a natural question is the connection between this existence result and the non - existence result of holmgren @xcite stating that given a local " immersion of an open set in @xmath3 into @xmath0 , there is a maximal domain of finite diameter on which this immersion can be extended to a smooth global immersion with one of the principal curvatures diverging as we approach the boundary . in sec . [ sec : minimax ] , we obtain a quantitative version of this result , by showing numerically that for isometric immersions of a hyperbolic disk of radius @xmath12 the ( nondimensional ) maximum principal curvature of the immersion is bounded from below by a bound which grows exponentially in the ( nondimensional ) radius . we also identify the surfaces which realize this bound , and they turn out to be disks cut from hyperboloids of revolution with constant negative curvature . in sec . [ sec : small_slopes ] , we use the small slopes approximation to find approximate isometric immersions with a periodic profile . the small - slopes approximation has been widely used to model torn sheets and growing tissues @xcite . in this setting the problem of finding immersions of @xmath5 reduces to solving a hyperbolic monge - ampere equation . the simplest solutions are a family of quadratic surfaces with two asymptotic lines intersecting at the origin . we then periodically extend the surface bounded between the two lines to construct periodic shapes that have the same scaling of amplitude versus period observed experimentally @xcite . the two wave configuration is the global minimum of the bending energy over the class of isometric immersions and we prove that with decreasing thickness the configurations should converge to this shape . in sec . [ sec : amsler ] we extend the surfaces constructed from solutions to the monge - ampere equation by studying hyperbolic surfaces with two asymptotic lines . these surfaces are called amsler surfaces , named after henri amsler who studied them in his paper on the nature of singularities of hyperbolic surfaces @xcite . amsler surfaces are generated by a similarity transformation of the sine - gordon equation which transforms it to a painlev iii equation in trigonometric form @xcite . from these surfaces , we can create periodic shapes as we did with the small slopes approximations that are energetically preferred over corresponding disks cut from the pseudosphere and hyperboloids . the bending energy of these surfaces is concentrated in small regions near the singular curves and the lines we are taking the periodic extension . furthermore for each @xmath18 , there is a radius @xmath19 such that the @xmath17-periodic amsler surfaces only exist for a radius @xmath20 . this gives a natural geometric mechanism for the refinement of the wavelength of the buckling pattern with increasing radius of the disk . finally , we present a concluding discussion in sec . [ sec : discuss ] . we model the stress free configuration of a thin elastic disk of thickness @xmath21 and radius @xmath12 that has undergone differential growth as an abstract riemannian manifold @xmath22 with metric @xmath5 . more specifically , if we choose a coordinate system for @xmath22 so that @xmath23 lies at the center of the disk then @xmath24 , where @xmath25 is the distance function induced by @xmath5 . a configuration of the sheet in three dimensions is then given by a mapping @xmath26 , where @xmath27 , the image of the mapping , is taken to be the mid - surface of a sheet . the equilibrium configurations of the sheet are extrema for the elastic energy functional : @xmath28=\int_{d}\\|d\mathbf{x}^t\cdot d\mathbf{x}-\mathbf{g}\\|^2\,da+t^2\int_{d}\\|4h^2-k\\|^2\,da,\ ] ] where @xmath29 is the mean curvature of the surface @xmath13 , @xmath30 is the gaussian curvature , @xmath31 is the jacobian matrix of partial derivatives of @xmath32 , and @xmath33 is the area element on @xmath22 @xcite . this functional is an asymptotic expression for a full three dimensional elastic energy functional per unit thickness @xcite . the first integral measures the stretching energy and vanishes when @xmath32 is an isometric immersion while the second integral measures bending energy and vanishes when the image of @xmath32 is a plane . this functional is a fvk type since it is the sum of stretching and bending energies such that ratio of bending to stretching energy scales like @xmath34 @xcite . if @xmath5 is not euclidean , i.e. if the gaussian curvature corresponding to @xmath5 is non - zero , then both of these terms can not simultaneously vanish in three dimensions and consequently the equilibrium configuration will not be able to reduce all of its elastic energy even in the absence of external forces . in this paper we will focus on ( [ energy ] ) when @xmath5 generates a constant negative gaussian curvature @xmath35 . @xmath6 is the _ target gaussian curvature _ , and we can always nondimensionalize such that @xmath36 , _ i.e. _ , we chose the units for length as @xmath37 . if the surface is ( initially ) parameterized by @xmath38 then @xmath39 , @xmath40 , are the dimensionless parameterization variables and @xmath41 is the dimensionless radius . therefore , if we can find a configuration @xmath32 such that @xmath42 then , the gaussian curvature of the immersion @xmath43 and ( [ energy ] ) reduces to the following equivalent elastic energy functional @xmath44=\int_d(k_1 ^ 2+k_2 ^ 2)\,da,\ ] ] where @xmath45 and @xmath46 are the principal curvatures of @xmath13 . since @xmath47 , this energy is equivalent to the willmore energy @xcite . we will refer to @xmath48 as the ( normalized ) _ bending energy _ of the configuration @xmath32 , although , strictly speaking , the bending energy is the quantity @xmath49 and not the quantity @xmath48 itself . since we are integrating over a geodesic circle it is convenient to work in geodesic polar coordinates , which are the natural analogs of the radial and polar angle coordinates on a riemannian manifold @xcite . in these coordinates @xmath13 is parameterized by the coordinates @xmath50 which map to a point @xmath51 satisfying @xmath52 and @xmath53 lies on a geodesic whose initial velocity vector makes an angle @xmath54 with respect to a basis of the tangent plane at @xmath55 . in this coordinate system the metric @xmath5 has components @xmath56 and equation ( [ bending - energy - general ] ) takes the form : @xmath57=\int_0^{2\pi}\int_0^r \sinh(r)(k_1 ^ 2+k_2 ^ 2)\,dr\,d\psi.\ ] ] lewicka and pakzad @xcite have shown that , for immersions @xmath32 in the space @xmath58 , @xmath59 = \begin{cases } \mathcal{b}[\mathbf{x } ] , & d\mathbf{x}^t\cdot d\mathbf{x } = \mathbf{g } \\ + \infty , & \mbox{otherwise } \end{cases}\ ] ] where @xmath60 implies that @xmath32 is an isometric immersion . therefore , in the limit as the thickness vanishes , equation ( [ bending - energy - polar ] ) is the appropriate functional with the set of admissible functions @xmath61 being sufficiently regular isometric immersions of the disk of radius @xmath12 in @xmath3 . in particular , we need @xmath62 < \infty$ ] , so the principal curvatures are @xmath1 functions on the disk @xmath22 , and we need to consider @xmath2 immersions @xmath63 . in this section we show how isometric immersions of @xmath22 can be cut out of a surface @xmath9 with constant negative gaussian curvature . let @xmath64 be a parametrization of @xmath9 and for @xmath10 consider the set @xmath65 . by minding s theorem @xcite as long as @xmath13 is smooth it can form an isometric immersion of @xmath22 . to construct @xmath13 we need to calculate the geodesics on @xmath9 and also to determine @xmath62 $ ] we need to calculate the principal curvatures on @xmath9 . to calculate these quantities recall the following facts and definitions from the theory of hyperbolic surfaces ( see pgs . 683 - 692 in @xcite ) . an asymptotic curve is a curve in @xmath9 whose normal curvature vanishes everywhere . moreover , since the principal curvatures @xmath45 and @xmath46 of a hyperbolic surface have opposite signs it follows that through each point @xmath66 there are exactly two asymptotic curves that intersect at @xmath55 . consequently , we can locally parameterize any hyperbolic surface by its asymptotic lines . such a parameterization is a c - net and has a metric @xmath67 with the following components @xmath68 where @xmath15 is the angle between the asymptotic curves at each point in @xmath9 . therefore , all of the intrinsic properties of the surface are expressible in terms of @xmath15 , and for this reason @xmath15 is called the generating angle . the condition that there must be two distinct asymptotic lines at each point in @xmath13 implies that @xmath15 must satisfy the constraint @xmath69 furthermore , it follows from brioschi s formula @xcite that @xmath15 must satisfy the sine - gordon equation @xmath70 and thus solutions to the sine - gordon satisfying ( [ angle - constraint ] ) generate hyperbolic surfaces . in fact hilbert proved that there is no smooth immersion of @xmath3 by showing that there is no smooth solution to ( [ sine - gordon ] ) that satisfies ( [ angle - constraint ] ) in the entire @xmath71 plane . the points on the surface where @xmath72 are precisely the boundaries where the immersion fails to be smooth . to determine the geodesics on @xmath9 we need to calculate the christoffel symbols . they are intrinsic quantities determined by the metric and hence by @xmath15 : @xmath73 which gives us the following geodesic equations : @xmath74 the extrinsic quantities of @xmath9 are also expressible in terms of @xmath15 . the components of the second fundamental form @xmath75 are given by @xmath76 and the principal curvatures are @xmath77 consequently , since @xmath78 is an isometric immersion of @xmath22 we have that equation ( [ bending - energy - polar ] ) can be expressed entirely in terms of @xmath15 : @xmath79=\int_0^{2\pi}\int_0^r \sinh(r)(\tan^2(\phi/2)+\cot^2(\phi/2))\,dr\,d\psi.\ ] ] we conclude this section by pointing out that a closely related parameterization of @xmath9 , that is sometimes easier to make calculations with , is a parameterization by curves of vanishing curvature or a c - net of the second kind @xcite . if @xmath32 is a c - net , then a c - net of the second kind @xmath80 is given by @xmath81 . the components of the metric @xmath67 corresponding to @xmath80 are @xmath82 since @xmath83 , the parameter curves of this parameterization are orthogonal and in fact they bisect the asymptotic curves . therefore , @xmath84 and thus we can determine the generating angle by such a parameterization . to illustrate the process of calculating the elastic energy of an isometric immersion of @xmath22 we consider a commonly known hyperbolic surface , the pseudosphere . the pseudosphere is given by the following c - net of the second kind @xcite @xmath85 the mapping is singular on the curve @xmath86 , which is where the surface fails to be a smooth immersion , and consequently we will only consider the upper half " @xmath87 of the pseudusphere . furthermore , since the sheets we are modeling have the topology of a plane and not a tubular surface we will think of the pseudosphere as a multi - sheeted surface . directly calculating , the metric coefficients are given by the following simple expressions : @xmath88 consequently , by equation ( [ tchebyshef - secondkind - metric ] ) the generating angle has the form @xmath89 and thus by ( [ principal - curvatures ] ) the principal curvatures are @xmath90 therefore , by equation ( [ bending - energy ] ) the elastic energy of a disk @xmath13 centered at @xmath91 and lying in the pseudosphere is simply @xmath92=\int_{0}^{2\pi}\int_0^{r}\sinh(r ) \left(\sinh^2(\eta(r,\psi))+\frac{1}{\sinh^2(\eta(r,\psi))}\right)\,dr\,d\psi.\ ] ] now , to determine how @xmath93 depends on @xmath94 and @xmath54 and the shape of a disk @xmath13 we need to calculate the arclength of geodesics on the pseudosphere . the arclength of a curve @xmath95 starting at @xmath91 and terminating at @xmath96 is given by the functional @xmath97=\int_{\eta_0}^{\eta_f}\sqrt{\tanh^2(\eta)+\sinh^2(\eta)\left(\frac{d\xi}{d\eta}\right)^2}\,d\eta\ ] ] and the geodesics that extremize this functional are given implicitly by @xmath98 where @xmath99 and @xmath22 are determined by the condition that the geodesic passes through @xmath91 and @xmath96 . define the function @xmath100 to be arclength of the geodesic starting at @xmath91 and terminating at @xmath101 . since the geodesics fail to be a function of @xmath93 at the critical value @xmath102 the arclength of a geodesic will have to be computed over both branches of a square root . assuming that @xmath103 , if @xmath104 then @xmath100 can be computed by switching @xmath105 and @xmath93 , we have by equation ( [ arclength ] ) that @xmath106 by making the substitution @xmath107 , and taking into account the case @xmath104 , the integrals can be explicitly evaluated . the contour @xmath108 forms the boundary of the disk @xmath13 , ( see figure [ ps : pseudospherecut ] ) . furthermore , the geodesics on the pseudosphere itself are given by @xmath109 . the tangent vector @xmath110 evaluated at @xmath86 gives us the tangent vector of this curve which we can use to calculate the angle a geodesic makes with respect to the basis @xmath111 of the tangent plane at the center of the disk by taking inner products . by making the substitution @xmath112 , this calculation gives us that the angle @xmath54 satisfies @xmath113 consequently , by specifying a value for @xmath54 we can calculate @xmath99 and therefore use equation ( [ l(m , n ) ] ) to determine @xmath93 for a particular value of @xmath94 and @xmath54 . this allows us to color the disk by bending energy density ( see figure [ ps : psdiscs ] ) and numerically integrate ( [ ps - energy ] ) over the disk to determine how the bending energy scales with @xmath12 for various values of @xmath114 ( see figure [ ps : energyplot ] ) . now , a disk centered at @xmath91 can not be made arbitrarily large since it will eventually meet the singular curve @xmath86 . we can prove that the largest disk on the pseudosphere with this center has a radius of @xmath115 . to prove this note that the curve @xmath116 is a geodesic , it trivially solves the euler - lagrange equation , and therefore @xmath117 now , let @xmath95 be another geodesic that terminates at a point @xmath118 on the boundary . then , @xmath119 consequently , the contour @xmath120 meets the curve @xmath86 at only one point and thus @xmath121 is the radius of the largest possible disk centered at @xmath91 . therefore , we can see that for arbitrary @xmath12 we can find a value @xmath105 such that @xmath22 is isometric to @xmath13 , and thus by varying the center of the disk we can create arbitrarily large stretching free configurations . this proves the following proposition let @xmath22 be a disk of radius @xmath12 in the hyperbolic plane . there exists a smooth isometric immersion @xmath122 such that @xmath13 is a subset of the pseudosphere . in this section we explore numerically how the principal curvatures of smooth isometric immersions scale with the size of the disk . in particular , we show that the maximum principal curvature on the disk grows exponentially in @xmath123 , where @xmath6 is the target gaussian curvature . let @xmath55 denote the center of the disk , and let us choose the asymptotic coordinates @xmath124 and @xmath125 such that @xmath55 is the point @xmath126 . if @xmath127 , it follows easily that there exists @xmath128 in @xmath22 with coordinates @xmath129 since @xmath130 . let @xmath53 , @xmath94 , @xmath131 and @xmath21 denote the vertices of a square " denoted @xmath132 $ ] in asymptotic coordinates given by the intersections of the asymptotic lines @xmath133 for some given @xmath134 . we redefine the variables @xmath135 and then drop the primes to transform equation ( [ sine - gordon ] ) into @xmath136 a smooth isometric immersion corresponds to a smooth function @xmath137 satisfying the sine - gordon equation ( [ eq : sg ] ) . we want to consider the variational problem for minimizing the @xmath138 norms of the principal curvatures . the natural mode of converge is then @xmath138 convergence and this does not preserve the class of smooth functions . we thus enlarge the admissible set . let @xmath139 denote the set of all measurable functions on the unit square with values in @xmath140 $ ] let @xmath141 be the set of measurable functions @xmath142 ^ 2 \rightarrow [ 0,\pi]$ ] that are distributional solutions of the sine - gordon equation , _ i.e _ @xmath143 ^ 2 } \phi g_{uv } du dv = \lambda \int_{[0,1]^2 } \sin(\phi ) g du dv , \qquad \forall g \in c_c^\infty((0,1)^2).\ ] ] clearly @xmath144 ^ 2)$ ] for all @xmath145 . if @xmath146 is a sequence in @xmath61 and @xmath147 in @xmath138 , the mean value theorem implies that @xmath148 , so it follows that @xmath149 . consequently , @xmath61 is a closed subset of @xmath150 ^ 2)$ ] for all @xmath145 . ( compactness ) since @xmath139 consists of bounded measurable functions , for every sequence @xmath151 , there exists a subsequence @xmath152 and a limit point @xmath153 such that @xmath154 weakly in all @xmath150)^2 $ ] for @xmath155 and @xmath156 ( converges weak-@xmath157 ) in @xmath158 ^ 2)$ ] . note that the limit @xmath159 is independent of @xmath55 . in fact , on @xmath139 , the weak @xmath138 topologies for @xmath155 and the weak-@xmath157 @xmath4 topoloy are all the same @xcite . [ rem : cmpct ] we are interested in a lower bound for the principal curvatures of a _ chebyshev patch _ , _ i.e. _ , a portion of a surface of constant negative curvature covered by a single coordinate system given by a chebyshev net . we thus define @xmath160 \times[0,1 ] } \max\left[\cot\left(\frac{\phi}{2}\right),\tan\left(\frac{\phi}{2}\right)\right].\ ] ] while we can estimate @xmath161 by direct numerical computation , we use a slightly different approach that yields some more information . since @xmath162 , we get @xmath163 . this motivates the definition @xmath164 where @xmath165 denotes the @xmath138 norm . this corresponds to the variational problem for functionals @xmath166 $ ] given by @xmath167 note that @xmath161 and @xmath168 are for a unit square in asymptotic coordinates , and this is not the same thing as a geodesic disk , although the two can be related . for all @xmath169 , the map @xmath170 is differentiable at @xmath171 . also , for any function @xmath172\times[0,1 ] \rightarrow \mathbb{r}$ ] , the @xmath55-norm @xmath173 is a nondecreasing function of @xmath55 . it is easy to see that , as a consequence , @xmath174 is non - decreasing in @xmath55 . a natural question is whether there exists a minimizer for @xmath175 . in this context we have let @xmath146 be a minimizing sequence for @xmath175 , _ i.e. _ , @xmath176 and @xmath177 . if @xmath146 ( or a subsequence ) converges pointwise to @xmath159 , @xmath159 is a minimizer for @xmath175 . [ prop : existence ] if @xmath146 ( or a further subsequence ) converges pointwise , it follows from dominated convergence that @xmath178 . since @xmath61 is closed in @xmath138 , it follows that @xmath179 . also , by fatou s lemma , we have @xmath180 showing that the minimum is attained . the function @xmath181 is convex on @xmath182 . therefore , the map @xmath183 ^ 2 } \sin(\phi ) h du dv,\quad h \in c_c^\infty((0,1)^2 ) , \ , h \geq 0,\ ] ] is weakly lower semi - continuous . if @xmath176 is any sequence and @xmath184 weakly in some @xmath185 or converges weak-@xmath157 in @xmath4 , @xmath186 it follows that @xmath143 ^ 2 } \phi^ * h_{uv } du dv - \lambda \int_{[0,1]^2 } \sin(\phi^ * ) h du dv \leq 0 , \qquad \forall h \in c_c^\infty((0,1)^2 ) , \ , h \geq 0.\ ] ] so , in general , @xmath187 , and thus , the functional @xmath175 are not lower semi - continuous with respect to weak or weak-@xmath157 convergence . thus , it is not enough to have weak convergence of the minimizing sequence for the existence of a minimizer . the pointwise convergence in the proposition is also necessary for using the direct method to show the existence of a minimizer for @xmath175 . the previous remark suggests considering the problem of minimizing the relaxed energy @xmath188 , namely the largest weakly lower semi - continuous function less than or equal to @xmath175 on @xmath139 @xcite . let @xmath48 denote the weak closure of @xmath61 in some @xmath189 ( the precise value of @xmath128 is irrelevant as indicated in remark [ rem : cmpct ] ) . the relaxation of @xmath175 is defined by @xmath190 where the infimum is over all sequences which converge weakly to @xmath15 . clearly , @xmath191 and @xmath188 is weakly lower semi - continuous by construction . it thus follows from remark [ rem : cmpct ] that we can prove the existence of a minimizer for @xmath188 by the direct method in the calculus of variations . the question of the existence of a minimizer for @xmath175 can thus be posed in terms of the nature of the minimizer of @xmath188 which always exists . if there exists @xmath179 which is a minimizer for @xmath192 , then @xmath159 is also a minimizer for @xmath175 . conversely , if all the minimizers of @xmath188 are in @xmath193 , it follows that @xmath175 does not have a minimizer . studying this question involves computing the relaxation @xmath188 which in turn needs a careful analysis of the potential oscillations in weakly convergent sequences of solutions of the sine - gordon equation . we are investigating these issues in ongoing work . we can discretize the sine - gordon equation on a @xmath194 grid covering the unit square , and minimize the @xmath195 norms of @xmath196 over all solutions of this discrete sine - gordon equation . this is a finite dimensional problem with a coercive energy , so there always exists a minimizer @xmath197 on the @xmath194 grid . [ numerics : lpnorm ] displays the numerically obtained results for this minimization . observe the lack of high frequency oscillations in the numerically obtained minimizers . in fact , increasing the mesh size suggests that as @xmath198 , the numerically obtained minimizers converge pointwise on @xmath199 ^ 2 $ ] . if this were indeed the case , proposition [ prop : existence ] implies the existence of minimizers for the functionals @xmath175 on the admissible set @xmath61 . we will henceforth assume that the infimum is attained , _ i.e. _ , for any given @xmath55 and @xmath200 , there is an admissible function @xmath201 such that @xmath202 , it follows that @xmath203 thus , numerically determining the function @xmath204 which minimizes the @xmath55-norm of @xmath205 in the admissible set , gives both lower and upper bounds for @xmath206 , and the difference between these bounds gives an estimate for the error in a numerical determination of @xmath206 . once we determine @xmath207 , we can compute @xmath161 as the following argument shows : @xmath208 where @xmath209 is a monotone function with a monotone inverse . consequently , @xmath210 and the value of @xmath161 is attained on @xmath211 , the extremizer for @xmath206 . the same argument can be applied to any function @xmath212 with the property that there is a continuous monotone function @xmath213 with a continuous monotone inverse such that @xmath214 . in particular , if @xmath215 , there is such a function @xmath213 , and it follows that @xmath216 \nonumber \\ & = \mbox{arg}\min_{\phi \in \mathcal{a } } \max_{x \in d } \left[\phi - \frac{\pi}{2}\right]^2 . \label{eq : equivalence}\end{aligned}\ ] ] we can solve any of these minimax problems to determine the extremizing function @xmath211 . for our numerics , we use the third formulation of the problem , _ i.e. _ , finding the minimax for the square of the deviation of @xmath15 from @xmath217 . also , we note that , @xmath206 directly estimates the maximum bending energy density @xmath218 and @xmath219 estimates the total bending energy . figure [ numerics : lpnorm ] shows the numerically computed values of @xmath174 and @xmath220 as a function of @xmath221 , @xmath222 , for @xmath223 . figure [ numerics : phi ] and figure [ numerics : cot ] shows the numerically determined extremizers @xmath204 as well as the quantity @xmath224 for @xmath223 and @xmath225 . the figures for other values of @xmath200 are similar in character . the key observation from these figures is that as @xmath55 gets large @xmath226 numerically approaches a limit that is purely a function of the combination @xmath227 ( see figure [ numerics : collapse ] ) . this motivates the ansatz @xmath228 . [ fig : numerics ] substituting this ansatz in equation ( [ eq : sg ] ) yields @xmath229 , which is the hamiltonian motion of a unit mass in a potential @xmath230 . the trajectories in phase space @xmath231 are given by conservation of the energy , @xmath232 ( see figure [ fig : phase_portrait ] ) . note that there are three distinct types of surfaces of revolution of constant negative curvature @xcite and for each of these surfaces @xmath15 is a function of @xmath227 only . therefore , the three distinct trajectories in the phase plane correspond to the three types of surfaces of revolution . the pseudosphere corresponds to the separatrix , hyperboloid surfaces correspond to closed orbits , and conical surfaces correspond to unbounded orbits . [ fig : phase_portrait ] . the three different types of orbits correspond to the three different types of surfaces of revolution of constant negative curvature . the piece of the orbit outlined in bold is the trajectory @xmath233\rightarrow ( 0,\pi)$ ] that minimizes the value of @xmath234 over all trajectories in the phase plane defined over the same time domain . ] we seek solutions on the unit square , @xmath235 , that are bounded away from @xmath236 on the interval @xmath237 $ ] . if we want trajectories that minimize the maximum value of @xmath234 , it is clear from the symmetry of @xmath238 we are interested in the closed orbits . furthermore , we seek a piece of these trajectories that satisfies the further symmetry requirement @xmath239 since if these conditions are not met we could select a different piece of the trajectory that could further minimize @xmath205 . this trajectory is depicted in bold in figure [ fig : phase_portrait ] . the value @xmath240 of a minimizing trajectory can be computed as @xmath241 ^ 2/2 + \lambda \cos(\psi(1 ) ) = \lambda \cos(\epsilon)$ ] , and the relation between @xmath242 and @xmath200 is determined by requiring that the time it takes to get from @xmath243 to @xmath244 is 1 , _ i.e. _ , @xmath245 we can solve this equation numerically , and thereby determine @xmath246 , provided that our ansatz is valid . generating an asymptotic series for ( [ eq : tof ] ) , we have for small @xmath242 that @xmath247 rearranging , we obtain @xmath248 + o[e^{-\sqrt{\lambda}}]\ ] ] this implies @xmath249 + o(1)$ ] . finally , if we set @xmath250 we have in terms of @xmath123 that @xmath251 we conclude this section by comparing the bending energy of disks cut from hyperboloids of revolution with the pseudosphere . the family of hyperboloids are parameterized by the following family of c - nets of the second kind @xmath252 where @xmath253 , @xmath254 , @xmath255 denote the usual jacobi elliptic functions @xcite , @xmath240 is the elliptic function of the second kind @xcite , and @xmath256 @xcite . the metric and principal curvatures are given by @xmath257 @xmath258 the christoffel symbols for this parameterization are @xmath259 therefore , the geodesic equations are @xmath260\\ 0&=\frac{d^2\xi}{dt^2}+\frac{2b^4\text{sn}(\eta , b^2)\text{cn}(\eta , b^2)\text{dn}(\eta , b^2)}{1-b^4\text{sn}^2(\eta , b^2)}.\end{aligned}\ ] ] it is easy to see that the principal curvatures diverge when @xmath261 . consequently , the center of the disk is located at the coordinate @xmath262 , @xmath263 is the elliptic function of the first kind , and thus the relationship between @xmath242 and @xmath264 is given by @xmath265 solving this equation yields @xmath266 therefore , for a particular radius @xmath12 we can numerically solve equation ( [ eq : tof ] ) to obtain @xmath264 . once , we determine @xmath264 we can then numerically integrate the bending energy @xmath267}=\int_{0}^{2\pi}\int_0^r\left(\frac{1-b^4\text{sn}^2(\eta , b^2)}{b^4\text{sn}^2(\eta , b^2)}+\frac{b^4\text{sn}^2(\eta , b^2)}{1-b^4\text{sn}^2(\eta , b^2)}\right)\sinh(r)\,drd\psi,\ ] ] following a similar procedure as we did with the pseudosphere . these results are summarized in figure [ fig : hyperboloid ] . [ fig : hyperboloid ] the configurations formed from the pseudosphere and hyperboloids can have very large bending energy concentrated at the center and top edge of the disk and these configurations do not have periodic profiles . in order to understand the experimental observations for negatively curved disks @xcite , we thus need to identify periodic configurations with a lower elastic energy . we begin by considering the small slopes approximation . in this approximation , valid for @xmath268 , the configuration is assumed to have a parameterization @xmath269 of the form : @xmath270 where @xmath271 , @xmath272 and @xmath38 are cartesian coordinates on @xmath22 @xcite . that is , as a mapping , @xmath273 , where @xmath274 is the standard immersion @xmath275 , @xmath276 is the out of plane " deformation , _ i.e. _ , @xmath276 maps into the orthogonal complement of @xmath277 and @xmath278 is the in - plane " displacement . the target metric @xmath5 takes the form @xmath279 that is , @xmath280 is an asymptotic expansion for the target metric where @xmath281 is the euclidean metric on @xmath282 . if we define the in - plane strain terms by @xmath283 then using the above asymptotic expansions in the elastic energy ( [ energy ] ) and keeping the lowest order terms , we get @xmath284 & = \epsilon^{-2 } \mathcal{e}[\mathbf{x } ] \nonumber \\ & = \int_{b(r)}\left(\\|d\chi^t + d \chi + d \omega^t \cdot d\omega-\mathbf{g}_1\\|^2+t^2\\|d^2\omega\\|^2\right)\,da \,+\mathcal{o}(\epsilon^2,t^0)\\ & = \int_{b(r)}\left(\gamma_{11}^2 + 2\gamma_{12}^2+\gamma_{22}^2\right)\,dudv+t^2\int_{b(r)}\\|d^2\omega\\|^2\,dudv \,+\mathcal{o}(\epsilon^2,t^0 ) , \label{reduced - energy}\end{aligned}\ ] ] where @xmath285 is a ( 2 dimensional euclidean ) disk of radius @xmath12 and @xmath33 is the euclidean area element . the stretching and normalized bending energies under this approximation take the forms @xmath286=\int_{b(r)}\left(\gamma_{11}^2 + 2\gamma_{12}^2+\gamma_{22}^2\right)\,dudv\ ] ] @xmath287 = \int_{b(r)}\\|d^2 \omega\\|^2\,du\,dv,\ ] ] with the appropriate elastic energy given by @xmath288=s[\mathbf{x}]+t^2\mathfrak{b}[\mathbf{x}].\ ] ] we are using the symbols @xmath9 and @xmath289 to indicate that these are the lowest order contributions to the stretching and bending energies . in this approximation , the mean and gaussian curvatures to the lowest orders in @xmath242 are given by @xmath290 , \quad \quad k = \epsilon^2 \,\,\mathrm{det}[d^2 \omega]\ ] ] so that a necessary and sufficient condition for the surface to have constant negative curvature @xmath291 is that the out of plane displacement @xmath276 should satisfy the following hyperbolic monge - ampere equation @xmath292 once a solution to ( [ monge - ampere ] ) is chosen we can match the metric @xmath67 to the target metric @xmath5 by solving for the in plane stretching terms @xmath293 and @xmath294 . a class of solutions to ( [ monge - ampere ] ) is the following one parameter family of quadratic surfaces @xmath295 parametrized by @xmath296 @xcite . these solutions are saddle shaped such that @xmath276 vanishes on a pair of lines passing through the origin of the @xmath71 plane intersecting at an angle @xmath297 . to create periodic structures we first consider the function @xmath298 which is simply a rotation of @xmath299 that aligns the asymptotic lines with the @xmath124 axis and the line @xmath300 . then , we take the odd periodic extension @xmath301 of @xmath302 with respect to @xmath303 about the two asymptotic lines . now , since the period of @xmath304 is @xmath305 , to guarantee that the shape is continuous at @xmath306 we must have that @xmath307 satisfies @xmath308 , where @xmath17 is the number of waves in the configurations profile ( see figure [ periodicprofiles ] ) . are saddles with asymptotic lines that vanish along the lines through the origin described by the polar angles @xmath297 colored in red . periodic solutions to the monge - ampere equation with @xmath309 periods can be constructed by taking odd periodic extensions of the solutions @xmath310 about the line @xmath311 . these solutions are @xmath312 and are smooth everywhere except on the asymptotic lines . ] since @xmath299 is smooth and vanishes along the two lines we are taking the extension about , it is a basic result of fourier analysis that @xmath301 has continuous first partial derivatives but a jump in the second partial derivative with respect to @xmath303 . consequently , these lines are singularities in the sense that the surface is not smooth across these lines but the bending energy remains continuous and is finite . furthermore , since the metric only involves first derivatives of the map @xmath32 the stretching energy will still vanish on these lines . therefore , these surfaces are valid isometric " immersions for the small slopes approximation i.e. , they are @xmath2 isometric immersions . the elastic energy of a disk with radius @xmath12 and @xmath17 periods is simply : @xmath313 for any solution of ( [ monge - ampere ] ) , the product of the eigenvalues of the hessian @xmath314 is @xmath315 . consequently , @xmath316 with equality only if @xmath317 . with @xmath318 ( _ i.e. _ @xmath319 ) in the periodic profiles from above , the hessian of the surface has eigenvalues @xmath320 . consequently , the 2-wave saddle @xmath321 gives a global minimum for the bending energy over all the solutions of ( [ monge - ampere ] ) . using this as a test function for the elastic energy in ( [ reduced - energy ] ) , we get an upper bound for the energy of the minimizer which scales like @xmath34 . it is a standard argument to show that for all @xmath322 there exist minimizers for the reduced elastic energy @xmath323 @xcite . further , if the appropriate @xmath324-limit of @xmath323 exists then these minimizing configurations and their derivatives will converge as @xmath325 to a limit configuration which satisfies satisfies ( [ monge - ampere ] ) and minimizes the bending energy ( [ smallslopes - energy ] ) @xcite . we provide a more detailed argument proving these statements below . recall that on a metric space @xmath326 a sequence of functionals @xmath327 @xmath324-converge as @xmath325 to a functional @xmath328 , written @xmath329 , if for all @xmath330 we have that 1 . ( * liminf inequality * ) for every sequence @xmath331 converging to @xmath332 @xmath333\leq \liminf_jf_j[x_j];\ ] ] 2 . ( * recovery sequence * ) there exists a sequence @xmath331 converging to @xmath332 such that @xmath334\geq \limsup_jf_t[x].\ ] ] let @xmath335 be the sequence of functionals on @xmath336 defined by @xmath337=f_t[(f , g,\omega)]=\frac{1}{t^2}e[\mathbf{x}],\ ] ] where we are naturally identifying a configuration @xmath32 with an element @xmath338 . then @xmath339=f_0[(f , g,\omega)]=\begin{cases } \mathfrak{b}[\omega ] & \text { if } \gamma_{11}=\gamma_{12}=\gamma_{22}=0\\ \infty & \text{otherwise } \end{cases}\ ] ] @xmath340 let @xmath341 such that @xmath342 . 1 . it is a basic property of weak convergence that @xmath343 if @xmath344 in @xmath345 then @xmath346 and @xmath347 such that @xmath348 . it follows immediately from item 1 that @xmath349\leq \liminf_{t\rightarrow 0}\mathfrak{b}[\mathbf{x}_t].\ ] ] 3 . since @xmath285 has finite area it follows that @xmath350 for @xmath351 and thus by rellich s compactness theorem we have that @xmath352 converges strongly to @xmath353 in the @xmath354 norm for @xmath355 . therefore for @xmath355 we have by the reverse triangle inequality that @xmath356 and thus @xmath357 . therefore we have the following limits and inequalities : 1 . @xmath358 2 . @xmath359\,da\\ & \ , & + \lim_{t\rightarrow 0 } \int_{b(r)}\frac{\partial f_t}{\partial u}\left(\frac{\partial \omega}{\partial u}\right)^2\,da\\ & = & \int_{b(r)}\frac{\partial f}{\partial u}\left(\frac{\partial \omega}{\partial u}\right)^2\,da,\end{aligned}\ ] ] 3 . @xmath360 4 . @xmath361 5 . @xmath362 + therefore by items ( i - v ) it follows that @xmath363 identical arguments can be used on the other strain terms to show that @xmath364\leq\liminf_{t\rightarrow 0}\mathcal{s}[\mathbf{x}_t].\ ] ] if @xmath365=0 $ ] then by item 2 we have that @xmath366=\liminf_{t\rightarrow 0}\frac{1}{t^2}\mathcal{s}[\mathbf{x}_t]+\liminf_{t\rightarrow 0}\mathfrak{b}[\mathbf{x}_t]\geq \mathfrak{b}[\mathbf{x}],\ ] ] while if @xmath365\neq 0 $ ] then by item 3 @xmath366\geq\liminf_{t\rightarrow 0}\frac{1}{t^2}\mathcal{s}[\mathbf{x}_t]+\liminf_{t\rightarrow 0}\mathfrak{b}[\mathbf{x}_t]\geq \liminf_{t\rightarrow 0}\frac{1}{t^2}\liminf_{t\rightarrow 0}\mathcal{s}[\mathbf{x}_t]=\infty.\ ] ] therefore , @xmath367\geq f_0[\mathbf{x}]$ ] proving property 1 of the definition . finally , if @xmath368 then by selecting the sequence @xmath369 we prove property 2 . let @xmath370 be the sequence in @xmath371 defined by @xmath372 . then , @xmath373 exists and @xmath374 . fix @xmath375 such that @xmath376 and let @xmath377 such that @xmath378 $ ] and @xmath379 $ ] . then , @xmath380\leq f_t[\mathbf{x}_s]=\frac{1}{t^2}s[\mathbf{x}_s]+\mathfrak{b}[\mathbf{x}_s ] < \frac{1}{s^2}s[\mathbf{x}_s]+\frak{b}[\mathbf{x}_s]=f_s[\mathbf{x}_s].\ ] ] therefore , the sequence @xmath372 is monotone increasing as @xmath381 and by ( [ linear - energy ] ) , with @xmath382 , satisfies @xmath383 . therefore by the bolzano weierstrass theorem @xmath370 converges . define @xmath384 by @xmath385 the above definition fixes the origin of a configuration and aligns the normal at the origin with the @xmath386-axis of @xmath0 . by a rigid translation and rotation any element @xmath387 can naturally be identified with an element of @xmath388 . if @xmath389 is a sequence such that @xmath390 $ ] then @xmath391\ ] ] and every limit of convergent subsequence of @xmath392 is a minimum point for @xmath393 . let @xmath394 such that for all @xmath395 , @xmath396\leq 2\pi r$ ] . then there exists @xmath397 such that @xmath398 . therefore by poincar s inequality there exists @xmath399 such that @xmath400 and thus by the banach - alaoglu theorem the set @xmath401\leq 2\pi r\}$ ] is weakly compact . 1 . let @xmath392 be a sequence such that @xmath390 $ ] . by ( [ linear - energy ] ) it follows that @xmath402\leq 2\pi r$ ] and thus by compactness of @xmath30 there exists a subsequence @xmath403 such that @xmath404 . therefore , by ( [ gamma - lsc ] ) and the above lemma we have that @xmath405\leq \liminf_{k\rightarrow 0 } f_k[\mathbf{x}_{t_k}]=\lim_{k\rightarrow 0}\min_{\mathcal{a}_0}f_{t_k}=\lim_{t\rightarrow 0 } \min_{\mathcal{a}_0}f_t.\ ] ] 2 . fix @xmath406 and let @xmath368 such that @xmath407\leq \min_{\mathcal{a}_0}f_0+\delta$ ] . then if @xmath408 is a sequence satisfying ( [ recovery ] ) then @xmath409\geq \limsup_{t\rightarrow 0}f_t[\mathbf{y_t}]\geq \limsup_{t\rightarrow 0}\min_{\mathcal{a}_0}f_t=\lim_{t\rightarrow 0}\min_{\mathcal{a}_0}f_t.\ ] ] by the arbitrariness of @xmath410 if follows that @xmath411 therefore , by items 1 and 2 it follows that @xmath412 furthermore , if @xmath392 is sequence as defined in item 1 and @xmath403 is a subsequence such that @xmath413 then @xmath414\leq \lim_{t\rightarrow 0 } \min_{\mathcal{a}_0}f_t\ ] ] which proves that @xmath415.\ ] ] consequently , the small slopes approximation will predict that , for small thickness , the configuration of the sheet will converge to a quadratic saddle @xmath416 . this is in disagreement with the experimental observation that the disks can obtain an arbitrary number of waves @xcite . however , if we consider the amplitude @xmath417 of an @xmath17 wave configuration , we get the scaling law @xmath418 which does agree with the experiment @xcite . a key difference between exact isometries and the small slopes approximation is captured by the following observation . the curvatures of the surfaces given by ( [ quad - surface ] ) are bounded by @xmath419 independent of @xmath12 , the radius of the disk . this is in contrast with our result ( [ eq : sclng ] ) which shows that for isometric immersions , the curvature has to grow exponentially in the radius @xmath12 . a qualitative expression of this distinction is that the analogs of hilbert s and efimov s theorems do not hold for the small slopes approximation . consequently , in the next section we look to extend our construction by looking for periodic surfaces that are exactly locally isometric to @xmath3 and for small @xmath12 are well approximated by the small slopes solution . to mimic the solutions to the small slope approximation we want to construct a hyperbolic surface that has two asymptotic curves that are straight lines and intersect at the origin of @xmath13 . hyperbolic surfaces satisfying this property exist , they are called amsler surfaces @xcite , and form a one parameter family of surfaces @xmath420 that are uniquely determined by the angle @xmath303 between the asymptotic lines @xcite . as in the construction of surfaces in the small slopes approximation , if the angle @xmath303 between the asymptotic lines satisfies @xmath421 we can take the odd periodic extension of the piece of the amsler surface bounded between the asymptotic lines and form a periodic profile with @xmath17 waves . we call the surfaces constructed in this manner periodic amsler surfaces . the key to generating periodic amsler surfaces is through the sine - gordon equation ( [ sine - gordon ] ) and the similarity transformation @xmath422 which transforms ( [ sine - gordon ] ) into a painlev iii equation in trigonometric form : @xmath423 where @xmath424 denotes differentiation with respect to @xmath386 @xcite . by imposing the initial conditions @xmath425 and @xmath426 solutions to ( [ painleve - equation ] ) , denoted by @xmath427 , generate surfaces such that the @xmath124 and @xmath125 axis correspond to the asymptotic lines of the surface . therefore , the piece of the surface we are extending periodically is parameterized in the first quadrant of the @xmath71 plane and all of its geometric quantities are determined by @xmath428 . and @xmath429 waves respectively . these surfaces are not drawn to scale , but are close to the maximum radii presented in figure [ amslerenergy : bendenergy(r ) ] . the coloring of the disks corresponds to contours of arclength data and indicate how the different geodesic circles cut from these surfaces would appear . these surfaces are in fact not true amsler surfaces but are discrete amsler surfaces creating using the algorithm presented in @xcite . ] the geodesic equations ( [ geodesic - equations ] ) under this similarity transformation take the form : @xmath430 with initial conditions @xmath431 , @xmath432 and @xmath433 for @xmath434 $ ] . notice that these equations have a singularity at the point @xmath435 . this is precisely where the immersion becomes singular and thus the singular curve is given in the @xmath71 plane by @xmath436 once the singular curve is determined we can calculate the maximum radius of a periodic amsler surface with @xmath17 waves by determining the shortest geodesic that intersects the singular curve . we can obtain a scaling for @xmath437 by multiplying equation ( [ painleve - equation ] ) by @xmath438 to get @xmath439 integrating , we get that @xmath440 therefore , @xmath441 we can use this scaling to prove that we can create arbitrary large disks from these periodic amsler surfaces . first noting that the same argument used to obtain ( [ pi - scaling ] ) implies that @xmath442 . now , let @xmath443 be a geodesic lying on a periodic amsler surface with generating angle @xmath444 . suppose @xmath445 starts at the origin and travels to the curve @xmath446 terminating at the point @xmath447 and without loss of generality assume that @xmath448 is a function of @xmath449 . then , the arclength of @xmath445 satisfies : @xmath450 thus , since @xmath451 as @xmath452 we have proved the following proposition . let @xmath22 be a disk of radius @xmath12 in the hyperbolic plane . there exists a @xmath2 isometric immersion @xmath122 such that @xmath13 is a subset of a periodic amsler surface . figure [ scaling ] is a plot of several solutions to equation ( [ painleve - equation ] ) and includes the scaling of the radius of the largest disk with @xmath17 waves that can be embedded in @xmath0 . we can see qualitatively that the maximum radius grows logarithmically with @xmath17 , a result that is confirmed by equation ( [ pi - scaling ] ) . the geodesics for the particular case when @xmath382 are plotted in figure [ amsler - geodesics : geodesics ] along with the geodesic circles for @xmath453 in figure [ amsler - geodesics : circles ] . versus @xmath94 satisfying the initial conditions @xmath454 for @xmath455 . the dashed horizontal line corresponds to where @xmath456 . at this point a principal curvature diverges and by choosing a higher value of @xmath17 a larger disk can be created . the inset plot illustrates how the maximum radius scales with @xmath17 . we can see that the radius grows very slowly with @xmath17 and it looks to be growing approximately at a logarithmic rate . ] to compute the elastic energy using equation ( [ bending - energy ] ) we need to determine the dependence of @xmath124 and @xmath125 on @xmath94 and @xmath54 . let @xmath457 and @xmath458 denote unit vectors in the @xmath71 plane that are aligned with the @xmath124 and @xmath125 axis respectively . if we let @xmath32 denote the parameterization of the amsler surface , then the pushforward or differential of @xmath32 , denoted @xmath459 , is a linear map . furthermore , since @xmath32 is an isometry and the images of @xmath457 and @xmath458 under @xmath459 make an angle of @xmath460 we have without loss of generality that @xmath461 where @xmath462 and @xmath463 are the standard basis elements for the tangent plane at @xmath464 . consequently , if we let @xmath465 be a geodesic defined in the @xmath71 plane satisfying @xmath466 then by linearity we have that @xmath467 therefore , we can conclude that the polar angle @xmath54 for this geodesic is given by @xmath468 now , we can numerically integrate the elastic bending energy for the periodic amsler surfaces . first , by specifying @xmath54 and numerically solving ( [ polar - angle ] ) for @xmath469 we can generate initial conditions for equations ( [ geodesics ] ) which can then be numerically solved to determine @xmath448 and @xmath449 . then , by fixing @xmath94 and numerically solving the arclength equation @xmath470 for @xmath471 we can calculate the values @xmath472 and @xmath473 which correspond to the coordinates @xmath474 . finally by setting up a mesh on the rectangle @xmath475\times [ 0,r]$ ] and using the above process to determine @xmath124 and @xmath125 we can numerically integrate equation ( [ bending - energy ] ) . these results are summarized in figure [ amslerenergy ] . free non - euclidean thin elastic sheets arise in a variety of physical @xcite and biological @xcite systems . the morphology of these sheets is usually modeled as the equilibrium configurations for an appropriate elastic energy . one approach is to model the sheet as an abstract manifold with a prescribed _ target metric _ @xmath5 , which is then used to define strains and hence an elastic energy for a configuration . in this paper we considered the geometric problem of immersing disks with _ constant negative curvature _ into @xmath0 . there are two settings for this problem which are relevant . one is the problem of exact " isometric immersions @xmath476 where @xmath63 is the configuration of the center surface . the other problem is a small - slopes approximation with an in and out of plane decomposition , @xmath477 with the isometric immersions given by @xmath478 in the small slopes approximation , we proved that global minimizers of the bending energy over solutions of ( [ eq : lnrzd ] ) are given by a quadratic saddle @xmath321 . we thus expect that , with decreasing thickness , the configuration of the sheet will converge to a two wave solution . this is not what is observed experimentally , where the number of waves increases with decreasing thickness @xcite . consequently , the physically realized configurations _ are not _ given by the global minimizers of the model energy . note however , that the periodic , non - smooth profiles that we construct in sec . [ sec : small_slopes ] do agree qualitatively , and in scaling behavior , with the experimental observations . this is a puzzle , since these periodic , non - smooth profiles _ are not _ global minimizers of the energy , and we need a different selection mechanism if these profiles do indeed represent the experimental configurations . one way to improve the modeling of the system is to consider the full " fppl von krmn energy ( [ energy ] ) and the associated isometric immersions given by ( [ eq : exact ] ) . for complete surfaces of constant negative curvature ( or surfaces whose curvature is bounded above by a negative constant ) , hilbert s theorem @xcite ( resp . efimov s theorem @xcite ) show that there are no analytic ( resp . @xmath8 ) solutions of ( [ eq : exact ] ) . these results are sometimes used as a basis for physical intuition that there are no global " isometric immersions of constant negative curvature surfaces , and this leads to the observed refinement of wavelength with decreasing thickness @xcite . we investigated this problem and showed ( sec . [ sec : pseudosphere ] and [ sec : minimax ] ) there exists smooth isometric immersions @xmath479 for geodesic disks of arbitrarily large radius @xmath12 . moreover , in elucidating the connection between this existence result and the non - existence results above , we provided numerical evidence that the maximum principal curvature of such immersions is bounded below by a bound which grows exponentially in @xmath123 . the existence of smooth , and thus also @xmath2 immersions of arbitrarily large disks have consequences for the modeling of free non - euclidean sheets . in particular , the minimizers of the elastic energy ( [ energy ] ) must converge as @xmath480 to a minimizer of ( [ bending - energy - general ] ) @xcite , and the energy of the minimizer @xmath481 is bounded from above by @xmath482 \leq t^2 \inf_{\mathbf{y}\in \mathcal{a}}\mathcal{b}[\mathbf{y}]\ ] ] where @xmath61 is the set of all the @xmath2 solutions of ( [ eq : exact ] ) . this also implies that the bending energy of the minimizing configurations should satisfy the inequality @xmath483 \leq \mathcal{b}_0 \equiv \inf_{\mathbf{y}\in \mathcal{a}}\mathcal{b}[\mathbf{y}]\ ] ] for all @xmath484 , _ independent of _ @xmath21 . this is an apparent contradiction with experimental observations where the bending energy of the minimizer scales with the thickness @xmath21 and is well fit by a power law , @xmath485 , which diverges as @xmath325 @xcite . this suggests that further improvements of the modeling might be necessary for a quantitative description of the experimental phenomena . the smooth immersions @xmath479 ( secs . [ sec : pseudosphere ] and [ sec : minimax ] ) do not have the rotational @xmath17-fold symmetry ( @xmath486 for the geodesic polar angle @xmath54 ) of the underlying energy functional , which is also seen in the experimentally observed configurations . in sec . [ sec : amsler ] we generalize the construction of the non - smooth , @xmath17-wave small - slopes solutions to obtain isometric immersions ( solutions of ( [ eq : exact ] ) ) which are also non - smooth and have the same symmetry / morphology as the experimentally observed configurations . these periodic amsler surfaces _ have lower bending energy _ than the smooth immersions given by subsets of the pseudosphere and hyperboloids with constant negative curvature . for each @xmath487 , there is a radius @xmath488 such that the @xmath17-periodic amsler surfaces only exist for a radius @xmath489 . this gives a natural mechanism for the refinement of the wavelength of the buckling pattern with increasing radius of the disk . however , it does not , at least directly , explain the observed refinement with decreasing thickness . we have considered the purely geometric model which arises as a limit @xmath480 . a natural question is the applicability if our results for physical sheets where @xmath322 . perhaps the refinement of the wavelength with decreasing @xmath21 is an effect which can only be captured by studying the full fppl von krmn energy ( [ energy ] ) . for @xmath322 , we expect that the minimizers of the energy will have boundary layers near the edge of the disk and also near the singular asymptotic lines , so that the disks satisfy stress and moment balance everywhere . the boundary layers near the outer edge of the disk may then set the wavelength throughout the disk . a similar phenomenon in stretched elastic sheets was studied in this manner in @xcite . the buckling patterns in non - euclidean sheets are the result of dynamical ( albeit slow ) processes , and it may be more appropriate to model this type of differential growth dynamically , perhaps as a gradient flow for the elastic energy . the pattern is then selected for dynamical reasons ( fastest growth rate , etc . ) and not by global minimization of an energy functional this might explain why local but not global extrema for the energy functional seem to describe the observed patterns . however , neither the boundary layers , nor the gradient flow dynamics , can explain how the refinement with decreasing thickness can lead to an apparent divergence of the the bending energy @xmath62 \sim t^{-1}$ ] . we have shown rigorously , for the elastic energy ( [ energy ] ) , in the limit @xmath480 , we should have @xmath62 \rightarrow \mathcal{b}_0 < \infty$ ] . this suggests the need for better models that are carefully validated by comparison with experiments in order to obtain a quantitative description of non - euclidean free sheets . 1 . in sec . [ sec : minimax ] , we use an _ ansatz _ motivated by numerical experiments to show that the solutions to the minimax problem @xmath490 are given by disks cut out of constant curvature hyperboloids of revolution , where the admissible set @xmath61 consists of all @xmath8 or smoother isometric embeddings . we are working on a rigorous ( ansatz free ) proof of this result . 2 . our lower bound for the maximum principal curvature ( and hence the maximum bending energy density ) in sec . [ sec : minimax ] can be interpreted as a lower bound on the @xmath491 norm of the bending energy density @xmath492 over the admissible set , which consists of isometric immersions . the limiting energy and configuration of the non - euclidean sheet as @xmath480 is given by the lower bound for the @xmath493 norm of the bending energy density . motivated by this observation , we are investigating lower bounds for the @xmath138 norms @xmath155 to connect the known case @xmath494 with the case of interest @xmath495 . chebyshev nets give a natural discretization of surfaces with constant negative curvature that respects both the intrinsic and the extrinsic geometry of the surface . this idea has already been used for constructing discrete isometric immersions ( @xmath30-surfaces ) for such surfaces @xcite . this can be interpreted as a @xmath480 limit for the variational problem ( [ energy ] ) . we are developing numerical methods for the variational problem given by ( [ energy ] ) with @xmath484 that exploits the chebyshev net structure . our results strongly use the fact that the target metric on the surface has constant negative curvature , which then allows us to naturally associate a chebyshev net with each isometric immersion through an asymptotic parameterization @xcite . however , we believe that our results , in particular the lower bound for the maximum principal curvature ( [ eq : sclng ] ) , and the construction of @xmath496 and piecewise smooth isometric immersions can also be extended to other target metrics whose curvature is bounded above by a negative constant . it is natural to study these questions , in particular the difference between smooth and @xmath496 immersions using the ideas in @xcite . 5 . for the periodic amsler surfaces in sec . [ sec : amsler ] , the global structure , _ i.e_. the number @xmath17 of waves is determined by the local structure at the origin . the origin is a _ bifurcation point _ @xcite for the chebyshev net induced by the embedding in the sense that , every point in the complement of the origin has precisely two asymptotic directions , while the origin has @xmath17 asymptotic directions with @xmath497 . the authors of @xcite remark that for surfaces whose total positive curvature is bounded by @xmath498 , it is always possible to find a global chebyshev net , except it can have multiple bifurcation points . this suggests the following natural questions : are there non - smooth immersions ( @xmath496 or even @xmath2 immersions ) with multiple bifurcation points ? + the geometry of such surfaces will then serve as a model for the observed morphology in many non - euclidean sheets including torn plastic @xcite and lettuce leaves @xcite which do not have a globally defined number of waves , but rather have local buckling behavior which increases the number of waves as we approach the boundary . * acknowledgements : * the authors wish to thank eran sharon , efi efrati and yael klein for many useful discussions , and for sharing pre - publication experimental data with us . we would also to thank marta lewicka and reza pakzad for fruitful discussions . finally we would like to thank the anonymous reviewer for many useful comments that improved the quality of this paper . this work was supported by the us - israel bsf grant 2008432 and nsf grant dms0807501 . 45 natexlab#1#1[2]#2 , , , , , , ( ) . , , , , , ( ) . , , , , ( ) . , , , , , edition , . . , , , , ( ) . , , , ( ) . , , , ( ) . , , , , , ( ) . , , , ( ) . , , , ( ) . , , , ( ) . , , , , studies in advanced mathematics , , edition , . , , , ( ) . , , , , ( ) . , , , ( ) . , , , , ( ) . , , , , ( ) . , , ( ) . , , ( ) . , , ( ) . , , . , , ( ) . , , ( ) . , , , in : , , , pp . . , , , in : , volume , , , pp . . , , ( ) . , , , ( ) . , , , ( ) . , , , volume of _ _ , , , . , , ellis horwood series : mathematics and its applications , , , . , , in : , , , , pp . . , , in : ( ed . ) , , volume , , , p. . , , , volume of _ _ , , . , , , , ( ) . , , , , ( ) .	we investigate isometric immersions of disks with constant negative curvature into @xmath0 , and the minimizers for the bending energy , _ i.e. _ the @xmath1 norm of the principal curvatures over the class of @xmath2 isometric immersions . we show the existence of smooth immersions of arbitrarily large geodesic balls in @xmath3 into @xmath0 . in elucidating the connection between these immersions and the non - existence / singularity results of hilbert and amsler , we obtain a lower bound for the @xmath4 norm of the principal curvatures for such smooth isometric immersions . we also construct piecewise smooth isometric immersions that have a periodic profile , are globally @xmath2 , and numerically have lower bending energy than their smooth counterparts . the number of periods in these configurations is set by the condition that the principal curvatures of the surface remain finite and grow approximately exponentially with the radius of the disc . we discuss the implications of our results on recent experiments on the mechanics of non - euclidean plates . nonlinear elasticity of thin objects , geometry of hyperbolic surfaces , pattern formation , morphogenesis in soft tissue .
as more extrasolar planets are discovered , we are increasingly pressed to describe how planets can form in such a variety of environments . until just recently , observational selection biases have resulted in the fact that all observed extrasolar planets have been found to orbit within a few au of their star @xcite . since it seems unlikely that these planets could have formed _ in situ _ @xcite , planet migration is usually invoked @xcite . unfortunately , this means that little is known about where and hence how these planets originally formed . in contrast , the technique of direct - imaging has presented us with a new set of extrasolar planets that lie far from their star @xcite , along with a potential protoplanet @xcite . like previous techniques , direct imaging preferentially detects giant planets of several jupiter masses . furthermore , planet migration need not be invoked to explain how these planets could form at their observed locations . one possible mechanism for giant planet formation is core accretion followed by rapid gas accretion @xcite . however , this mechanism has difficulty forming giant planets at large radii . the primary reason for this is that the initial core accretion time scales as @xmath1 , where @xmath2 is the orbital radius of the planet @xcite . thus , while it may take @xmath3 myr to form a gas giant at 5 au via core accretion , it would take @xmath3 gyr for the same process at 50 au far longer than the observed lifetimes of protoplanetary disks @xcite . another mechanism for giant planet formation is disk fragmentation as a consequence of the gravitational instability . provided that the disk surface density is sufficiently large , this mechanism can form giant planetary embryos on time scales of a few orbital periods . however , if the surface density is too large , the disk is unable to cool sufficiently fast for fragmentation to take place at all @xcite . the combination of these requirements implies gravitational instability can only form massive planets at large radii . in this letter , we consider the planet @xcite , the triple - planet system @xcite , and the potential protoplanet orbiting @xcite . each of these systems possesses at least one planet with orbital characteristics favored by the disk fragmentation mechanism . by determining the range of surface densities required to form a giant planet with the same semi - major axis as these observed planets , we can infer the range of disk masses needed for the fragmentation mechanism to have operated in these systems . the stability of a thin , massive disk is controlled by the @xcite @xmath4 parameter @xmath5 where @xmath6 is the isothermal sound speed , @xmath7 is the orbital angular frequency ( assuming a keplerian disk ) , and @xmath8 is the surface density . the disk becomes gravitationally unstable for @xmath9 . however , even if a disk is gravitationally unstable , it can only fragment if it possesses a sufficiently short cooling time @xcite . specifically , fragmentation will only occur if @xmath10 where @xmath11 is the local cooling time for a small , point - source perturbation , and @xmath12 is a factor of order unity that can depend on the history of the disk @xcite . we adopt @xmath13 and @xmath14 for our fragmentation criteria . typically , the effects of cooling have been studied using time - dependent hydrodynamic simulations . inevitably , these numerical approaches have to employ significant simplification of the radiation field for the sake of computation time ( e.g. , optically thin cooling or flux limited diffusion ) . many of the simulations show that fragmentation does occur given sufficiently high surface densities @xcite . in contrast , @xcite used an analytic , order - of - magnitude calculation to show that cooling times derived from the equations of radiative transfer were much longer , and that fragmentation thus did not work , except at radii @xmath15 au . here , we adopt an approach inspired by rafikov , but with a more complete calculation of the radiative transfer . in brief , we find cooling times that are , in most cases , over an order of magnitude shorter than those given by rafikov ( see nero & bjorkman , in prep . for a more complete discussion ) . as a consequence , we find that fragmentation over a larger range of the outer disk is possible , depending on the details of the system . we emphasize that the cooling time we calculate here is for a _ perturbation _ , and is not the same as the total disk cooling time employed by @xcite . while the later may be more convenient for numerical hydrodynamic simulations , the former is necessary to properly account for background heating by external illumination ( i.e. , the host star ) . the perturbation cooling time determines the onset and initial growth of the instability ( in the linear regime ) , while the total cooling time controls the ultimate ( typically non - linear ) completion of the instability . note , however , that when self - heating is small , the perturbation and total cooling times will be the same within a factor of order unity . the perturbation cooling time @xmath16 , where @xmath17 is energy per unit area added by the perturbation , and @xmath18 is the frequency - integrated eddington flux at the disk surface . we consider an annulus within the disk , which we approximate as a plane - parallel atmosphere with finite thickness . for simplicity , we assume that the perturbation is located at the disk mid - plane and that the disk cools equally from its top and bottom surfaces . under these assumptions , the perturbation cooling time is @xmath19 where @xmath20 is the adiabatic constant for the gas , @xmath21 is the mean opacity ( absorption plus scattering ) , @xmath22 is the optical depth coordinate , @xmath23 and @xmath24 are the depth - dependent planck function and its perturbation , and @xmath25 , @xmath26 , and @xmath27 are the planck function , the temperature , and the isothermal sound speed at the disk mid - plane , respectively . the limits of integration , @xmath28 and @xmath29 , are the optical depth coordinates at the `` top '' and `` bottom '' surface , respectively . note that we break convention here by placing @xmath30 at the disk mid - plane , rather than at the top surface . locations in the disk below the disk mid - plane have negative optical depth coordinates , while those above have positive . also , note that we have assumed that @xmath21 is approximately constant over the vertical extent of the disk , since the disk is nearly isothermal in the vertical direction . while this is not true for the surface layers , the error is minimal since most of the disk mass and thus internal energy is located in the disk interior . similarly , the vertically isothermal assumption does not apply when accretion is the dominant source of heating ( since then there would be a significant vertical temperature gradient ) ; however , at large radii accretion luminosity is usually not the dominant heating mechanism for the disk . we assume that the relevant physics ( i.e. , reprocessing / absorption of external radiation along with viscous energy generation ) can be preserved by splitting the intensity into two frequency components : _ diffuse _ , which corresponds to photons that have been reprocessed and emitted by the disk , and _ external _ , which corresponds to unabsorbed photons emitted from the central star and potentially scattered in the disk . assuming gray opacity ( i.e. , the appropriate mean for each spectrum ) , the frequency integrated moments of the transfer equations are@xmath31 where @xmath32 is the mean intensity , @xmath33 is the eddington flux , and @xmath34 and @xmath35 are the total and absorptive opacity , respectively . we have written all optical depth coordinates in terms of @xmath36 . the eddington factors , @xmath37 and @xmath38 , are assumed to be constant with depth . note that we have included no thermal emission for the _ external _ frequency because the disk is typically much cooler than the star . to find the _ external _ radiation , we combine eqs . ( [ eq:1st - ext ] ) and ( [ eq:2nd - ext ] ) , which have the solution@xmath39 where @xmath40 $ ] and @xmath41 is the net _ external _ surface flux . there are three sources of energy for the _ diffuse _ radiation : 1 ) absorption of _ external _ radiation @xmath42 , 2 ) accretion luminosity @xmath43 with surface flux @xmath44 , and 3 ) the point - source perturbation at the mid - plane @xmath45 . thus , the flux transported by the disk is @xmath46 . from eq . ( [ eq:2nd - diff ] ) , we can now obtain the _ diffuse _ mean intensity @xmath47 where @xmath48 is the second eddington factor . finally , the temperature may now be found from radiative equilibrium , eq . ( [ eq:1st - diff ] ) , which gives @xmath49 as well as@xmath50 given the temperature and its perturbation , we calculate the cooling time from eq . ( [ eq : tcool ] ) . in terms of the surface density , @xmath51 , the cooling time @xmath52 where the exponent @xmath53 depends on both the disk optical depth and the ratio of heating from the star to self - heating from accretion ( see nero & bjorkman , in prep . ) , but is generally in the range 02 . in the negligible self - heating regime ( @xmath54 , @xmath55 ) , eq . ( [ eq : tcool ] ) reduces to @xmath56 in the optically thick limit ( @xmath57 @xmath58 ) , eq . ( [ eq : tcool - thick+thin ] ) further simplifies to @xmath59 note that this expression is directly comparable to the optically thick limit of eq . ( 4 ) in @xcite , with the exception that our numerical factor ( 3/128 ) is over an order of magnitude smaller than his ( 1/4 ) , thus leading to significantly reduced cooling times . this difference from rafikov s result is much larger than the difference between total and perturbation cooling times . it instead arises from his assumption that several factors were of order unity , which is true individually ; however , the cumulative effect of his assumptions results in his making a large overestimate . our analytic expression for the cooling time , eq . ( [ eq : tcool - thick ] ) , assumes no accretion luminosity and large optical depth . while it is probably safe to assume that self - heating is small at large disk radii , it is not as certain that the optically thick limit will hold . for this reason , we use the full solution of eq . ( [ eq : tcool ] ) when calculating @xmath11 , rather than assuming that this limit holds . nonetheless , eq . ( [ eq : tcool - thick ] ) makes for a good estimate in most cases where @xmath8 is large enough to satisfy @xmath13 in the outer disk . when determining if disk fragmentation is a viable mechanism for forming giant planets , another point to consider is the issue of producing the proper planetary mass of a few @xmath60 . while a full treatment of this problem is beyond the scope of this letter , we provide a toy model to argue that this is likely to be the case . if @xmath9 at some radius @xmath2 , the disk becomes gravitationally unstable . supposing @xmath61 spiral arms form , each arm has local surface density@xmath62 where @xmath8 is the original surface density of the previously axisymmetric disk , @xmath63 is the current width of the spiral arms , and @xmath64 is a constant that depends on the winding angle . for simplicity , we assume that most of the disk mass is confined to the spiral arms , while the space between the arms is effectively empty . in addition , we assume @xmath65 , corresponding to moderate winding . if @xmath66 , then the spiral arms are pressure supported and are stable against fragmentation . however , if @xmath67 , then the arms are instead supported by centrifugal forces . as @xmath63 continues to decrease , they will fragment radially once the centrifugal support is lost . balancing self - gravity against the centrifugal support , @xmath68 , fragmentation occurs when @xmath69 where @xmath70 is the stellar mass . the fragment mass is @xmath71 , so assuming @xmath72 , we find a characteristic fragment mass @xmath73 which is consistent with our requirement to produce jupiter - mass planets . the condition that both the toomre @xmath4 and the cooling time be sufficiently small ( eqs . [ [ eq : q ] ] and [ [ eq : tcool - omega ] ] ) can be used to place limits on what disk surface densities lead to fragmentation . to be gravitationally unstable , the toomre @xmath4 condition requires that the surface density be larger than a minimum , @xmath74 , while the cooling condition imposes a maximum surface density , @xmath75 . therefore , local disk fragmentation is only possible if @xmath76 . it of course follows that for fragmentation to be possible at all , @xmath74 must be less than @xmath75 . this limits the range of radii were fragmentation is even a possibility . surface density limits for disk fragmentation . the solid line denotes @xmath75 , which is the maximum surface density for the cooling time constraint . the dashed line denotes the minimum surface density for fragmentation , @xmath74 , which is the locus toomre @xmath77 . disk fragmentation is only allowed in the region @xmath78 , which is shaded gray . the locations of known planets are plotted as vertical dotted lines . ] lcccr fomalhaut b & 2.0 & 1.8 & 8500 & @xmath79 + hr 8799 b , c , d & 1.5 & 1.8 & 8200 & 68 , 38 , 24 + hl tau b & 0.3 & 0.6 & 3700 & 65 + lcccccc fomalhaut b & 3594 & 0.130.34 & 0.170.45 & 0.260.68 & 0.511.36 & 521 + hr 8799 b & 6289 & 0.100.15 & 0.140.19 & 0.200.29 & 0.410.58 & 47 + hl tau b & 2143 & 0.030.06 & 0.040.09 & 0.060.13 & 0.130.26 & 24 + in fig . [ fig:2much2little ] , we plot @xmath74 and @xmath75 for the systems fomalhaut , hr 8799 , and hl tau , using the parameters listed in table [ tab : parameters ] . in all cases we used an accretion rate of @xmath80 , a mean molecular weight for the disk of @xmath81 , and an adiabatic gas constant of @xmath82 . the temperature is determined from eq . ( [ eq : planck - fn ] ) , using a flared disk model with a power law scale height @xmath83 , which determines the angle of incidence of the _ external _ radiation . we use the dust opacity of @xcite and the rosseland mean gas opacity of @xcite . we note that the stellar parameters in table [ tab : parameters ] are not necessarily appropriate if planet formation occurs during the class 0/i phase when the star is significantly more luminous . however , our results are relatively insensitive to this effect , since @xmath84 is only weakly dependent on the stellar luminosity . each system presented here has a planet that might have formed via disk fragmentation , assuming that the local surface density had a value between @xmath74 and @xmath75 for at least a few orbital periods . by assuming a power law for the surface density , @xmath85 , we can calculate a range of disk masses ( interior to @xmath86 ) that satisfies this condition . a survey of 24 circumstellar disks by @xcite found @xmath87 0.01.0 with an average of @xmath88 , while the hydrodynamical simulations of @xcite found @xmath87 1.02.0 with an average around @xmath89 . disk mass limits @xmath90 for @xmath91 , 0.5 , 1.0 and 1.5 , along with the more fundamental surface density limits are given in table [ tab : mass - limits ] . we also provide the characteristic fragment mass ( approximate planet mass ) @xmath92 from eq . ( [ eq : mass ] ) that we would expect from the disk fragmentation mechanism . also note that , in order to be conservative , we are using the smallest radius found by @xcite for fomalhaut b. using one of their better fits ( e.g. 115 au ) will decrease our lower disk mass limit by a few percent , increase our upper disk mass limit by @xmath93 ( which would make fragmentation slightly easier ) , and increase the characteristic fragment mass by @xmath94 . we have refined the calculations of @xcite and found cooling times over an order of magnitude shorter . we have used these cooling times , along with the observed stellar parameters of fomalhaut , hr 8799 , and hl tau , to test the viability of the disk fragmentation mechanism . we found that in each of these systems , at least one planet could have formed _ in situ _ as the result of fragmentation , assuming the disk mass interior to those planets fell within a particular range as indicated in table [ tab : mass - limits ] . while the ranges in table [ tab : mass - limits ] only span a factor of a few , this is not by itself a significant limitation . even if the local surface density is above the upper instability limit , fragmentation may still occur since the surface density must eventually drop through the unstable regime as the disk evolves and dissipates . the caveat is that the surface density needs to evolve on a timescale longer than an orbital period so that there can be sufficient time to fragment . our minimum disk masses for fomalhaut b , hr 8799 * * * * b , and hl tau b are about an order of magnitude larger than those inferred from observations @xcite . note , however , that this is a problem for all planet formation models in general . even core accretion models require an enhanced surface density ( although to a somewhat lesser extent ) @xcite . one possible mechanism for increasing the surface density is mass loading from an infalling envelope @xcite . conversely , current estimates of disk masses may be too low because they depend on : 1 ) the extrapolation of surface densities in the outermost regions of the disk to the inner disk , and 2 ) the rather uncertain dust opacity . for example , larger dust grains would require larger disk masses to fit the observed seds @xcite . as further evidence for underestimated disk masses , numerical hydrodynamical simulations by @xcite found disk masses much higher than those of @xcite . in particular , stars like fomalhaut and hr 8799 can support disks as large as 0.5 @xmath95 , while hl tau could have a disk as massive as 0.1 @xmath95 , all of which are within our limits for disk fragmentation . we caution , however , that our choice of opacity model can have a major effect on our results . for example , decreasing the dust opacity raises the temperature and decreases the cooling time in the outer disk , resulting in disk fragmentation at smaller radii . on the other hand , increasing the opacity would have the opposite effect . regardless of the above considerations , hr 8799 c and d are too close to their parent star to have formed _ in situ _ via fragmentation under the conditions modeled here . appealing to chronically overestimated dust opacity can only get us so far . dropping the opacity by an order of magnitude brings hr 8799 c into the fragmentation zone , but still leaves hr 8799 d out . likewise , twiddling other parameters can also move the fragmentation radius inward , but reaching the required 24 au with reasonable parameters does not seem possible . we therefore conclude that hr 8799 c and d likely did not form _ in situ _ as the result of disk fragmentation ( of course , they could have formed at larger radii were the disk is more likely to fragment and migrated inward ) . for those planets that could form by disk fragmentation , we find characteristic fragment masses ( approximate planet masses ) of a few @xmath60 for the lower end of unstable disk surface density . our estimates are mostly consistent with expectations , although @xcite found fomalhaut b to have an upper mass limit of 3 @xmath60 , which is 60% lower than our lowest estimate of 5 @xmath60 . nonetheless , considering the crudeness of eq . ( [ eq : mass ] ) , we are not convinced that this discrepancy rules out fomalhaut * * * * b from having formed as a result of disk fragmentation . in conclusion , given our current uncertainty of typical disk masses and dust opacities , we can not rule out planet formation from disk fragmentation . of the three currently known systems that might posses planets formed in this manner , all are able to produce at least their outermost giant planet via fragmentation , given a large enough disk mass at some point in their history . this research was supported by nsf grant ast-0307686 to the university of toledo . we would like to thank ken rice and lee hartmann for their valuable insights regarding this work . we also wish to thank the anonymous referee for very helpful suggestions that significantly improved the manuscript .	disk fragmentation resulting from the gravitational instability has been proposed as an efficient mechanism for forming giant planets . we use the planet fomalhaut b , the triple - planetary system hr 8799 , and the potential protoplanet associated with hl tau to test the viability of this mechanism . we choose the above systems since they harbor planets with masses and orbital characteristics favored by the fragmentation mechanism . we do not claim that these planets _ must _ have formed as the result of fragmentation , rather the reverse : if planets can form from disk fragmentation , then these systems are consistent with what we should expect to see . we use the orbital characteristics of these recently discovered planets , along with a new technique to more accurately determine the disk cooling times , to place both lower and upper limits on the disk surface density and thus mass required to form these objects by disk fragmentation . our cooling times are over an order of magnitude shorter than those of @xcite , which makes disk fragmentation more feasible for these objects . we find that the required mass interior to the planet s orbital radius is @xmath0 for fomalhaut b , the protoplanet orbiting hl tau , and the outermost planet of hr 8799 . the two inner planets of hr 8799 probably could not have formed _ in situ _ by disk fragmentation .
are structures ( sheets or filaments of vorticity ) a vital ingredient of intermittency in 3d - incompressible turbulence ? to date , this important question remains open @xcite , and an answer starting from first principle , i.e. , from a controlled approximation to the navier - stokes equations , seems over the horizon . the new understanding of the anomalous scaling in the kraichnan s model of passive advection @xcite , based on the identification of zero modes in the homogeneous hopf equations for equal - time correlators , has in particular strengthened the belief that field - theoretical methods would eventually be able to capture the full statistics of turbulent flows without an explicit account of structures . interestingly enough , the relative interplay between coherent ordered structures and incoherent turbulent fluctuations turns out to be a subtle matter already in the restricted framework of the so - called shell models of turbulence @xcite . it was noticed very soon @xcite that elementary bricks of intermittency in those deterministic 1d - cascade models could be pulses or bursts of activity growing in an almost self - similar way as they move from large to small scales . however , genuine dynamically stable self - similar solutions of the equations of motion in the inertial range display an unique scaling exponent ( to be denoted below as @xmath2 ) , provided they are localized in @xmath3-space ( which , in the shell model approach , reduces to a discrete set of wave numbers @xmath4 , where the shell index @xmath5 goes from 1 to @xmath6 ) . furthermore , the exponent @xmath2 , giving the logarithmic slope of the velocity gradient spectrum left in the trail of the pulse , happens to be rather close to the kolmogorov value @xmath7 ( @xmath8 ) in the case of the gledzer - okhitani - yamada ( goy ) model , in the range of parameters where it reproduces at best the multiscaling properties of real turbulent flows . in ref.@xcite , the role played by the interaction of pulses with the rest of the flow in producing more singular events was unravelled , and a two - fluid picture was introduced , where coherent structures form in and propagate into a featureless random background . our goal in this paper is to elevate this still rather qualitative proposal to the rank of a semi - quantitative theory and to test its predictive power about intermittency in the goy model . we shall assume that turbulent fluctuations on the shells downstream the pulse , i.e. , small scales , act on the coherent part of the flow as a random , white - in - time , gaussian forcing and ask whether the inviscid stochastic extension of the goy model obtained in this way is able to reproduce the statistics of strong deviations of the full turbulent system in the inertial range . there is _ a priori _ quite a lot of freedom in the parameterisation of the forcing . therefore , in order to keep things as simple as possible , we bind ourselves to use a single adjustable parameter ( hereafter noted @xmath0 ) , which measures the level of noise . we consider the semiclassical limit @xmath9 of these systems and study the statistics of singular structures appearing in this regime . semiclassical ( or instanton ) techniques are well suited to capture large and rare excursions of fluctuating fields @xcite . as such , they have gained recently a renewal of interest in the field of turbulence and have already led to noteworthy results in the context of burger s turbulence @xcite , and of the kraichnan s model of passive scalar advection @xcite . one usually starts from a path integral representation of high order structure functions and uses a saddle point approximation to determine the coupled field - force configurations contributing mostly to those quantities . the nature of the statistical object to be computed imposes precise boundary conditions on the physical field and the random force ( respectively at large and small scales , where the cascade processes start and end ) . instantons , which in the inertial range often reduce to a self - similar collapse along some spatial dimensions , are eventually selected by a delicate matching procedure at the two boundaries . in shell models we are dealing with an intrinsically discrete lattice of logarithmic scales . as a consequence , the analytic computation of instantons is completely out of reach in the inertial range , not to speak about the matching on both sides of the cascade . to circumvent this difficulty , we shall focus on the probability distribution function ( pdf ) of scaling exponents after @xmath5 cascade steps , @xmath10 , and argue that , in the semiclassical limit @xmath9 , this pdf builds up from the neighborhood of a single self - similar instanton ( of scaling exponent @xmath1 ) that dynamic stability considerations will help us to construct numerically . in order to get non trivial physics , it turns out to be necessary to perform the semiclassical expansion of @xmath11 ( the rate of rarefaction of singularities of scaling exponent @xmath1 in the multifractal picture ) up to next to leading order in powers of @xmath0 . this can be achieved via a summation over quadratic fluctuations around the instantons , once the proper set of boundary conditions for the corresponding trajectories in configuration space has been defined . we shall show in details how to carry out this program and end up with a prediction for @xmath10 lending itself to a straight confrontation with the pdf of effective scaling exponents of coherents events that can be extracted from simulations of the goy model at very high reynolds numbers . although our interest lies primarily in gaining a better understanding of intermittency in the framework of shell models of turbulence , the emphasis will be put in this paper on the technical aspects of the method that we had to develop for computing instantons . we believe that this method is general enough to find applications in other contexts or physical problems , like for instance the motion of complex objects or excitations on 1d - lattices in the presence of a co - moving random environment . the paper is organized as follows . in section [ def ] , we define the stochastic extensions of the goy model that we shall study . in section [ msrformalism ] the equations of motion for instantons will be derived using the well - known martin siggia rose path integral representation of probability distribution functions for stochastic dynamical systems . section [ extsol ] is devoted to the computation of self - similar extremal trajectories , with the theoretical considerations underlying the solution explained in subsection [ extsoltheory ] and its practical implementing , together with the results , exposed in subsection [ extsolimplementing ] . the important effect of quadratic fluctuations and the rather heavy formal work behind their computation are discussed in section [ fluct ] . the comparison of the results issuing from the instanton approach with numerical data on the statistics of coherent structures in the genuine goy model is given in section [ discussion ] . we conclude in section [ conclusion ] , in particular as to the relevance of a two - fluid description of intermittency in shell models of turbulence . equations of motion for the goy model in the inertial range read : @xmath12 where the complex variable @xmath13 should be understood as the fourier component of the gradient velocity field at wavenumber @xmath14 and the integer @xmath5 runs from @xmath15 to @xmath16 . throughout this paper , usual values of parameters @xmath17 and @xmath18 will be assumed . it is convenient to cast ( [ goymod ] ) in a vectorial form @xmath19 , \label{goyvect}\ ] ] where the infinite - dimensional vector @xmath20 is built up from the @xmath21 s , while the @xmath22 component of the nonlinear kernel @xmath23 $ ] is given by the right hand side of ( [ goymod ] ) . it is worth noting at this point that @xmath24 plays dimensionally the role of enstrophy in real flow and that the inverse square root of this quantity sets the order of magnitude of the smallest time scale on the shell lattice . + since quadratic nonlinearities lead generically to finite time singularities , it is very useful to introduce a desingularizing time variable @xmath25 related to the physical time @xmath26 by the differential law @xmath27 this turns ( [ goyvect ] ) into @xmath28 } { ( { { \bf b } } ^{}\cdot { { \bf b } } ) ^{1/2 } } , \label{goytau}\ ] ] where both sides of the equation have the same scaling dimension in the field @xmath20 , which shows that an infinite `` time '' is now required to form a singularity by travelling across the whole shell axis . > from previous work @xcite , we know that every initial condition of finite enstrophy , when evolving under dynamics ( [ goytau ] ) , eventually organizes itself in a soliton - like pulse , moving from large to small scales at a constant speed with an exponential growth of its amplitude . the asymptotic state is unique , up to trivial phase symmetries of the goy model @xcite , time translations or multiplicative rescaling of the field @xmath20 , which all leave the equation of motion ( [ goytau ] ) invariant . we may restrict our attention without loss of generality to the case where the phase pattern along the shell axis does not break into a three - sublattice structure . the asymptotic floquet state , to be noted henceforth @xmath29 , is then purely real and such that : @xmath30 the period @xmath31 is the `` time '' needed for the center of the pulse to go from shell @xmath5 to shell @xmath32 , while the ( positive ) lyapunov exponent @xmath33 controls its growth . both quantities @xmath31 and @xmath33 , are dynamically selected in an unique way . the scaling exponent @xmath2 associated to the pulse ( fixing in particular the logarithmic slope of the spectrum left in its trail ) can be extracted from the identity @xmath34 . its value turns out to be 0.72 in the case of the goy model for the choice of parameters stated before . + we turn now to the stochastic models , that we are interested in solving by the instanton method . their physical motivation has been explained in ref . @xcite : we assume that pulses parameterize adequately singular ( and temporally coherent ) structures in shell models but that the deterministic dynamics ( [ goyvect ] ) should be enlarged towards a stochastic one , in order to describe the interaction of a given pulse with incoherent fluctuations produced by the relaxation of the trails left by its predecessors . we are therefore led to consider the following extension of the original inviscid goy model : @xmath35 + \sqrt{\gamma } ( { { \bf b } } ^{}\cdot { { \bf b } } ) ^{3/4 } b [ { { \bf c } } ] { \mbox{\boldmath $ \eta$}}\,,\ ] ] where @xmath36 is a gaussian noise , delta - correlated in time and shell index , whose correlations read : @xmath37 the various factors coming in front of @xmath36 in ( [ def : syststoch ] ) have the following meaning : the number @xmath0 fixes the relative strength of incoherent fluctuations with respect to coherent ones and we shall be interested in the semiclassical limit of small @xmath0 amenable to semi - analytic treatment . as will be clearer in a while , the overall scale factor @xmath38 is there to keep noise relevant all along the cascade , thereby preserving scale invariance . finally the matrix @xmath39 $ ] , of zero scaling dimension in the field @xmath20 since it depends only on the unit vector @xmath40 , may be used either to introduce spatial correlations of noise ( along the shell axis ) or to localize its action with respect to the instantaneous position of the pulse . although the formalism to be developed in this paper can deal with the most general situation , we restricted ourselves in practical investigations to diagonal matrices @xmath39 $ ] , just playing with the degree of localization of noise . results will be presented for three rather emblematic choices of @xmath41 : ( i ) @xmath42 , which describes a completely delocalized noise ; ( ii ) @xmath43 which keeps some flavour of the original goy dynamics and makes noise active just at the leading edge of the pulse ; and finally ( iii ) @xmath44 , which removes the action of noise further away from the center of the pulse . we must emphasize that these particular choices were not dictated by rigorous considerations on the underlying dynamics of the goy model , but rather used to scan the variety of behaviours which may be expected from such stochastic dynamical systems . it should be noted that the strucutre of the matrix @xmath41 is not constrained by any conservation law , since the coherent part of the flow does not form anymore a closed system , even in the inertial range , in our two - fluid description . finally , to simplify the following analysis , we are going to restrict the fields @xmath20 and @xmath36 in ( [ def : syststoch ] ) and ( [ def : correl1 ] ) to being real - valued vectors and neglect the effect of imaginary fluctuations . this is certainly not a serious restriction as for the instantons themselves , which are expected to be , like the self - similar deterministic solution described above , purely real , up to trivial phase symmetries of the goy model . it can also be remarked that the model ( ii ) ( which will be found later on to give the more convincing results ) does not require a complex noise , since the phase degrees of freedom have already been incorporated in the definition of the matrix b in that case . + while the deterministic dynamics ( [ goyvect ] ) selects a single self - similar solution blowing up in finite time with scaling exponent @xmath2 , the presence of noise in ( [ def : syststoch ] ) allows for a continuum of scaling exponents , even in the manifold of normalizable fields @xmath20 . in the small noise ( or semiclassical ) limit @xmath9 , the probability density of developing an effective growth exponent @xmath1 after @xmath45 cascade steps will take the form : @xmath46,\ ] ] where @xmath47 is the action per unit cascade step of the self - similar extremalsolution of scaling exponent @xmath1 of optimal bare gaussian weight ( or instanton ) , and @xmath48 measures , to lowest order in @xmath0 , how the basin of attraction of the instanton in phase space evolves with the number of cascade steps . note that the argument of the exponential in ( [ pred_pn ] ) , @xmath49 , is nothing but the cramr function introduced in the theory of large deviations , which governs the rate of rarefaction of singularities in the multifractal picture @xcite . we will show in this paper how to compute in a clean way both quantities @xmath50 and @xmath48 . before doing this , we must carefully handle problems related to the time discretization of the stochastic equation ( [ def : syststoch ] ) since a consistent treatment of them is necessary to get the right expression of the first correction @xmath51 . we shall adopt the view that the initial stochastic equation ( [ def : syststoch ] ) is to be understood in the stratonovich sense @xcite . however , in the path integral formulation of stochastic dynamical systems that we shall heavily use in the following , it is much simpler to work with the ito prescription which , in the limit of small time steps , amounts integrating ( [ def : syststoch ] ) within a basic euler scheme with all @xmath20-dependent quantities in the r.h.s . estimated at the prepoint . when switching to the ito discretization recipe , the stochastic equation has to be changed into : @xmath52 + \sqrt{\gamma } ( { { \bf b } } \cdot { { \bf b } } ) ^{3/4 } b [ { { \bf c } } ] { \mbox{\boldmath $ \eta$}}\,,\ ] ] where the new kernel @xmath53 $ ] differs from @xmath23 $ ] by the addition of the so - called ito drift - term . we give , for the sake of completeness , the resulting expression of the @xmath54 component of @xmath55 : @xmath56=n_{n } [ { { \bf b } } ] + \frac{1}{2}\gamma \ , \frac{\partial } { \partial b_{k } } [ ( { { \bf b } } \cdot { { \bf b } } ) ^{3/4}b_{nj } ] ( { { \bf b } } \cdot { { \bf b } } ) ^{3/4}b_{kj}\,.\ ] ] at this point , we may write down the discrete analog of ( [ def : tau ] ) as @xmath57 ( where @xmath58 is the time index ) and redefine the noise as @xmath59 . this leads to the following stochastic extension of ( [ goytau ] ) which will be the starting point of our formal analysis : @xmath60 } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2 } } + ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2 } b [ { { \bf c } } ] { \mbox{\boldmath $ \xi$}}\,,\ ] ] with @xmath61 statistics of classical fields in the presence of random forces can be examined with the help of field theoretical techniques formulated in @xcite . in particular , the probability to go from point @xmath62 at time @xmath63 to point @xmath64 at time @xmath65 may be written as a path integral : @xmath66\,,\ ] ] where @xmath67 $ ] is an effective action to be defined below , @xmath68 an auxiliary field conjugated to the physical one @xmath20 and @xmath69 stands for : @xmath70 in the last equation , the time interval @xmath65 was divided into @xmath71 subintervals of length @xmath72 ( with @xmath73 and @xmath74 ) and the number of shells was set to a finite value @xmath75 , in order to give a clear meaning to the measure . for the problem of interest ( [ ito_tau ] ) , the effective action @xmath76 takes the form : @xmath77=\sum_{i=0}^{i = n-1 } i { { \bf p } } _ { i}.\left ( { { \bf b } } _ { i+1}- { { \bf b } } _ { i}-\delta \tau \frac { { { \bf n } } _ { \gamma } [ { { \bf b } } _ { i } ] } { ( { { \bf b } } _ { i}\cdot { { \bf b } } _ { i})^{1/2}}\right)+\frac{\gamma } { 2 } ( { { \bf b } } _ { i}\cdot { { \bf b } } _ { i})\ , { { \bf p } } _ { i}\cdot b [ { { \bf b } } _ { i}]\,^{t}\ ! b [ { { \bf b } } _ { i } ] { { \bf p } } _ { i}\,,\ ] ] or , in the continuum limit : @xmath78=\int _ { 0}^{\tau_{f}}\,d\tau \ , i { { \bf p } } .\left(\frac{d { { \bf b } } } { d\tau } -\frac { { { \bf n } } _ { \gamma } [ { { \bf b } } ] } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2}}\right)+\frac{\gamma } { 2 } ( { { \bf b } } \cdot { { \bf b } } ) \ , { { \bf p } } \cdot b [ { { \bf b } } ] \,^{t}\ ! b [ { { \bf b } } ] { { \bf p } } \,.\ ] ] the last term in ( [ msraction2 ] ) , quadratic in @xmath68 , appears as a result of averaging over the gaussian noise @xmath79 , while the first one , linear in @xmath68 , would still be there in the absence of noise as a formal way of enforcing the deterministic equation of motion of @xmath20 . @xmath67 $ ] will be referred to in the following as the martin - siggia - rose ( msr ) action . + rescaling the auxiliary field @xmath68 as @xmath80 puts an overall large factor @xmath81 in front of the effective action and opens the way to a saddle point approximation to the path integral ( [ pathint ] ) . extremization of the action with respect to the configurations of both fields @xmath20 and @xmath68 between times @xmath15 and @xmath82 , for fixed endpoints , leads in a straightforward way to the following set of coupled equations defining extremal trajectories : @xmath83 } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2 } } + ( { { \bf b } } \cdot { { \bf b } } ) b\,^{t}\ ! b { \mbox{\boldmath $ \theta $ } } \,,\\ \frac{d{\mbox{\boldmath $ \theta $ } } } { d\tau } & = & - ^{t}\!{\cal m}\,{\mbox{\boldmath $ \theta $ } } -\frac{1}{2}{\partial } _ { { { \bf b } } } [ ( { { \bf b } } \cdot { { \bf b } } ) \ , { \mbox{\boldmath $ \theta $ } } \cdot b\,^{t}\!b{\mbox{\boldmath $ \theta $ } } ] \ , . \label{extremalesmsr2}\end{aligned}\ ] ] in the above equation , we set @xmath84 and @xmath85 is the jacobian matrix of the kernel @xmath86 } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2}}$ ] : @xmath87 } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2}}- \frac { { { \bf n } } _ { \gamma } [ { { \bf b } } ] \otimes { { \bf b } } } { ( { { \bf b } } \cdot { { \bf b } } ) ^{3/2}}}$ ] . as usual , equations ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) inherit a canonical structure @xmath88 where the hamiltonian @xmath89 reads @xmath90 } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2}}+\frac{1}{2 } ( { { \bf b } } \cdot { { \bf b } } ) ( { \mbox{\boldmath $ \theta$}}\cdot b\,^{t}\!b{\mbox{\boldmath $ \theta$}})\,.\ ] ] since @xmath89 is not explicitly time - dependent , we conclude that its value , to be called the pseudo - energy in the sequel , is conserved along any extremal trajectory . the action @xmath91 $ ] may be rewritten in terms of @xmath89 as @xmath92=\int _ { 0}^{\tau_{f}}d\tau \left({\mbox{\boldmath $ \theta$}}. \frac{d { { \bf b } } } { d\tau } -{\cal h}\right)\,,\ ] ] from which it is seen that the further requirement that the trajectory be extremal with respect to time reparametrization leads to the condition of vanishing pseudo - energy @xmath93 . noting that each term of @xmath89 in ( [ hamiltonian ] ) has the same scaling dimension in @xmath20 and @xmath94 ( either 1 or 2 ) , one finds that @xmath95 which shows that the overlap @xmath96 between the physical and auxiliary fields is also conserved , together with @xmath89 , along an extremal trajectory . this property reflects the scaling invariance of the stochastic cascade processes we have in mind . we should at this point insist on the fact that , in contrast to instantons in the framework of equilibrium statistical mechanics or quantum mechanics , equations for extremal trajectories in stochastic dynamical systems describe the real motion of the physical field in a particular `` optimal '' realization of the noise . the comparison of equations ( [ ito_tau ] ) and ( [ extremalesmsr1 ] ) shows indeed that the following relation holds between @xmath97 and @xmath94 : @xmath98 like their deterministic parent ( [ goytau ] ) , the equations of motion ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) sustain formally traveling wave - like solutions , such that @xmath99 whose scaling exponent @xmath100 is expected to be now related to the overlap @xmath101 ( with @xmath102 for @xmath103 , in the absence of noise ) . however there is little hope to find these solutions by a direct forward in time integration of ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) , as could be done successfully for equation ( [ goytau ] ) . this is because the auxiliary field @xmath94 intrinsically propagates `` backward '' in time , as is clear from the discretized version of ( [ extremalesmsr2 ] ) ( deduced from the extremization of ( [ msraction1 ] ) ) . in the present problem , we have observed numerically that regular floquet states emerge as dynamical attractors of ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) only for rather high values of @xmath104 ( otherwise the system evolves in a chaotic manner ) . they form a branch of solutions definitely distinct from the one to be obtained in the next section and correspond presumably to local maxima of the action rather than the local minima of interest to us . the previous considerations suggest that equations ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) should not be treated on the same footing . the careful examination of physical properties that instantons should possess will give us keys for computing them . assume for a while that a solution has been found , obeying to ( [ floquet1 ] ) and ( [ floquet2 ] ) . we note @xmath29 and @xmath105 the corresponding configurations of @xmath20 and @xmath97 . the linearization of the equation of motion ( [ extremalesmsr1 ] ) at fixed noise leads to the following evolution of fluctuations @xmath106 of @xmath20 around @xmath107 : @xmath108 the periodicity properties of the linear operator @xmath109 ensure that the fluctuations of @xmath20 may be decomposed on a complete set of eigendirections @xmath110 evolving according to ( [ linearisee1 ] ) and such that @xmath111 where @xmath112 denotes translation by one unit in the right direction along the shell lattice . in practice we shall have to work with a finite number of shells @xmath75 and , in order to get rid of boundary effects , it will be necessary to fully periodize the shell lattice : the index @xmath58 then runs between 1 and @xmath75 and the translation operator is easy to represent as a matrix . formally , the @xmath113 s can be determined at time @xmath63 by diagonalizing the floquet operator : @xmath114 where @xmath115 is a chronologically time ordered product ( initial time on the right ) . one observes that @xmath107 satisfies ( [ rightvect1 ] ) with a time averaged lyapunov exponent @xmath116 . + we claim now that every initial condition @xmath117 evolving in the configuration of noise @xmath118 should be attracted towards the instantonic trajectory . if it were not true , some perturbations would be able to grow in the comoving frame of the pulse , thereby generating scaling exponents larger than @xmath1 at no cost in the action , in contradiction with the hypothesis that the optimal realization of a singularity of exponent @xmath1 has been found . this strong criterion of dynamic stability is another way of stating that the cramr function should be insensitive to the details of the production of pulses in the forcing range . it implies that @xmath119 is an upper bound for the real part of the @xmath120 s . arranging the eigendirections @xmath113 in order of decreasing @xmath121 , we are therefore led to identify @xmath29 with @xmath122 . in the case of zero noise where we recover the deterministic solution of section [ def ] ( with @xmath123 in both ( [ extremalesmsr1 ] ) and ( [ linearisee1 ] ) ) , the time derivative @xmath124 is also solution of ( [ linearisee1 ] ) with the same lyapunov exponent as @xmath107 , @xmath116 ( @xmath125 in this case ) . in this limit we would naturally define @xmath126 as @xmath124 . this property is lost in the more general situation of a non vanishing noise , because @xmath127 is not time - invariant . what remains true however is the fact that @xmath107 and @xmath124 still span the set of `` coherent '' fluctuations which do not affect the shape of the pulse but modify its height and position . + by turning now our attention to the linear dynamics dual to ( [ linearisee1 ] ) we shall come close to the equation ( [ extremalesmsr2 ] ) . let us indeed consider the equation of motion @xmath128 where in order to limit the proliferation of symbols , we keep the same notation @xmath129 for the new auxiliary field , although it is only in particular circumstances , to be clarified below , related to the @xmath129 of equations ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) . from ( [ linearisee1 ] ) we get : @xmath130 the dual dynamics enables one to construct a basis of left eigenvectors @xmath131 ( with @xmath132 ) satisfying @xmath133 as well as the following orthogonality conditions with the members of the first basis : @xmath134 at every time . the vectors @xmath135 are determined by diagonalizing the adjoint floquet operator @xmath136 and enforcing the normalization condition ( [ ortho ] ) at time @xmath63 ( @xmath137 is now an anti - chronologically time ordered product ) . we may note at this point that the first left eigenvector @xmath138 is , in the generic case of non - zero noise , the only one to display the scaling behaviour anticipated for @xmath139 according to ( [ floquet2 ] ) , since its lyapunov exponent equals @xmath140 . we conclude that the auxiliary equation ( [ extremalesmsr2 ] ) in the restricted manifold of self - similar solutions is tantamount to the relation : @xmath141 where the multiplicative constant @xmath104 is nothing but the overlap @xmath142 ( @xmath143 ) , which was shown before to be indeed a conserved quantity . this claim is further confirmed by rewriting the original equation ( [ extremalesmsr2 ] ) as @xmath144{\mbox{\boldmath $ \theta$}}\\ & = & -^{t}\!{\cal m}\,{\mbox{\boldmath $ \theta$ } } -({\mbox{\boldmath $ \theta$}}\cdot b\,^{t}\!b{\mbox{\boldmath $ \theta$ } } ) { { \bf b } } - ( { { \bf b } } \cdot { { \bf b } } ) \,{\mbox{\boldmath $ \theta$}}.(\partial _ { { { \bf b } } } \,b)\,^{t}\!b{\mbox{\boldmath $ \theta$}}\,.\end{aligned}\ ] ] putting back 0 superscripts and reintroducing @xmath127 by using ( [ lienxitheta ] ) , we arrive at @xmath145 which shows that @xmath139 obeys the dual dynamics defined by equation ( [ dual1 ] ) . + having interpreted ( [ extremalesmsr2 ] ) as a condition of self - consistency for the conjugate momentum @xmath94 expressed by ( [ selfcons1 ] ) , we could contemplate the following newton - like procedure for catching numerically self - similar instantons . first make a guess for @xmath97 in the form of a traveling wave ( @xmath146 ) , integrate ( [ extremalesmsr2 ] ) forward in time in order to determine the asymptotic traveling state reached by @xmath20 in the prescribed configuration of the noise . then compute @xmath138 from the diagonalization of @xmath147 ( or from running ( [ dual1 ] ) backward in time in order to let emerge the eigendirection of lowest growth rate ) , employ ( [ selfcons1 ] ) for producing a new configuration of @xmath94 ( and thereby @xmath97 ) and iterate this loop many times at a fixed value of the overlap @xmath104 until convergence is achieved . however two major difficulties call for an improvement of the method : they both have to do with the stability of the trajectory upon time reparametrization . first we do not know the speed ( or the inverse period @xmath148 ) of the final traveling wave which must carry together @xmath20 and @xmath97 . therefore , when performing the first step of the iterative loop , we must allow continuous time reparametrization of our ansatz for the noise in order to fine tune the speeds of the two pulses formed by @xmath20 and @xmath97 and let both terms in the right hand side of ( [ extremalesmsr1 ] ) be always relevant . it will be explained in the next subsection how this goal can be achieved in practice . the second difficulty is much more serious than the preceding one and in the way of getting around it resides perhaps the most tricky part of this work . the point is that traveling wave solutions to ( [ extremalesmsr1 ] ) and ( [ extremalesmsr2 ] ) may perfectly have a non zero pseudo - energy @xmath89 , while we are looking for the particular ones with @xmath93 . we shall be able to fulfill asymptotically the two conditions @xmath149 and @xmath93 , if and only if our iterative guess for @xmath94 is constructed within a two - dimensional space rather than a unidimensional one as in the naive proposal made above . for this purpose , we are going to embed the linearized dynamics ( [ linearisee1 ] ) into a new one which admits the time translation mode @xmath150 as a true eigenstate , of the same growth factor @xmath119 as @xmath107 , restoring thereby the symmetries present in the absence of noise . we shall do that in the most economical way , from both formal and numerical points of view , by substituting @xmath151 to @xmath152 , i.e. , the eigendirection along which the fluctuations of @xmath20 around @xmath107 are the less stable . + the left eigenvector @xmath153 is first rescaled as @xmath154 ( which makes sense as long as @xmath155 , a condition always found to be satisfied in practice ) , so that @xmath156 has a unit overlap with @xmath157 , while being orthogonal to all other right eigenvectors @xmath113 with @xmath158 ( which will be noted @xmath159 from now on ) . one then considers the modified linear dynamics : @xmath160 it is easily checked that @xmath151 obeys ( [ linearisee2 ] ) , since equation ( [ extremalesmsr1 ] ) yields upon time derivation : @xmath161 it is also trivially seen that the other vectors @xmath162 for @xmath158 keep the same evolution under ( [ linearisee2 ] ) as under ( [ linearisee1 ] ) . the dual dynamics reads now : @xmath163 it leads to a new family of left eigenvectors @xmath164 , dual to the direct basis , whose second member @xmath156 has been defined above and the others relate to their original counterparts @xmath165 as @xmath166 although the @xmath164 s were introduced as a rather formal trick , it should be emphasized that @xmath167 and @xmath156 have an appealing physical meaning . parameterizing a perturbed trajectory for @xmath20 as @xmath168 , where the `` incoherent '' part of fluctuations @xmath169 is bound to be a linear superposition of the less dangerous modes @xmath159 for @xmath170 , one has indeed , to linear order in @xmath106 , @xmath171 these two relations will be useful in the computation of quadratic fluctuations to be presented in section [ fluct ] . they show that by projecting out the multidimensional fluctuation field @xmath106 onto the two vectors @xmath167 and @xmath156 , one has access to the most relevant part of it affecting respectively the amplitude and the time delay of the pulse constituting the instanton . in terms of @xmath167 and @xmath156 , the self - consistency condition ( [ selfcons1 ] ) for @xmath139 together with the requirement of zero pseudo - energy @xmath89 take the following form : @xmath172 where @xmath173 since from ( [ selfcons2 ] ) , @xmath174 , and from the equation of motion ( [ extremalesmsr2 ] ) , @xmath175 , the relation ( [ def_mu2 ] ) is just a way of restating @xmath93 . that ( [ selfcons1 ] ) implies ( [ selfcons2 ] ) results from the general link between @xmath138 and @xmath167 ( see ( [ lienphipsi ] ) ) . the reverse is true only under the supplementary condition of constant @xmath89 or @xmath176 , which is guaranteed by ( [ def_mu2 ] ) . it is proven by checking that in that case @xmath177 , as given by ( [ selfcons2 ] ) , obeys , as it should , ( [ dual1 ] ) : @xmath178 the great advantage of ( [ selfcons2 ] ) and ( [ def_mu2 ] ) with respect to ( [ selfcons1 ] ) is that this couple of equations lends itself to iterative procedures leading inexorably to a fixed point of zero pseudo - energy , a task seemingly out of reach before . there is some unavoidable arbitrariness in the construction proposed here , concerning in particular the definition of the vector @xmath156 , about which the reader may feel a little uncomfortable . we suspect that these unwanted features do not affect the final results since the original equations to be solved as well as the corresponding conserved quantities all have a clear mathematical definition where @xmath167 and @xmath156 merge together into @xmath138 . the action density @xmath179 could be computed successfully for the three stochastic models defined in section [ def ] using the iterative scheme outlined before . the shell lattice was first mapped onto a circle of @xmath75 sites , with @xmath75 typically ranging between 20 and 30 . finite size effects turn out to be completely negligible at such lengths of the chain , due to the strongly localized structure of the instantons . to start the computation , we make a guess for both the unit field @xmath180 and the noise @xmath181 called henceforth @xmath182 and @xmath183 . they are such that @xmath184 and @xmath185 furthermore , the noise is normalized in such a way that the overlap @xmath186{\mbox{\boldmath $ \xi$}}$ ] , takes on a prescribed value @xmath187 held as a control parameter during all the steps of the computation . a possible and convenient choice would be for instance @xmath188 , where @xmath189 is the deterministic solution of scaling exponent @xmath2 , and @xmath190 { \mbox{\boldmath $ \phi$}}^{0}_{1l}(\tau ) $ ] , where @xmath191 is the left eigenvector dual to @xmath29 , i.e. , in the absence of noise . [ fig.1 ] shows how both vectors @xmath192 and @xmath193 look like at a given time . their shapes will in fact little evolve as we let the scaling exponent @xmath1 depart from @xmath2 . + in order to allow time reparametrization of the trajectory , we first get an estimate of the instantaneous position of the pulse along the shell axis in our trial configuration by computing the following quantity @xmath194\left(c^{in}_{n}(\tau ) \right)^{2}.\ ] ] the notation @xmath195 $ ] recalls that , due to cyclic boundary conditions , the shell index @xmath5 is now only defined modulo @xmath75 and that in practice a continuous determination of this integer should be adopted close to the center of the pulse which contributes mostly to the right hand side of ( [ def_centermass ] ) . one has by construction @xmath196 . having recorded @xmath197 and @xmath198 during a whole period @xmath199 , we integrate forward in time the nonlinear evolution equation for @xmath200 deduced from ( [ ito_tau ] ) by projecting out the longitudinal part of its right hand side : @xmath201(\tau ) + b [ { { \bf c } } ] ( \tau ) { \mbox{\boldmath $ \xi$}}^{in}(\tau ' ) \right\}_{\perp},\ ] ] note that the subindex @xmath0 disappeared from the nonlinear kernel @xmath202 because the ito - drift term being linear in @xmath0 does not matter in the computation of the action to leading order ( we shall see in the next section how to handle it to next to leading order ) . the most salient feature of ( [ dync ] ) is that the noise configuration is evaluated in relation not to the time @xmath25 but rather to the actual instantaneous position of the pulse . this means that the time @xmath203 is automatically delayed or advanced with respect to @xmath25 , according to the recipe @xmath204.\ ] ] after integrating ( [ dync ] ) long enough , a new traveling wave state @xmath205 , @xmath206 will usually emerge , of period @xmath207 possibly different from @xmath199 and averaged growth factor @xmath208+b [ { { \bf c } } ^{out}]{\mbox{\boldmath $ \xi$}}^{out}\right ) . { { \bf c } } ^{out}\,d\tau .\ ] ] the vectors @xmath209 , @xmath210 are then identified as the two eigenvectors of @xmath211 ( defined in ( [ leftop ] ) ) of smallest ( real negative ) lyapunov coefficient and the corresponding @xmath212 , @xmath213 constructed as linear combinations of them obeying for all times the following relations @xmath214 finally , the trial noise configuration is renewed as @xmath215{\mbox{\boldmath $ \theta$}}^{in}$ ] with @xmath216 where @xmath217 is determined upon imposing the condition of zero pseudo - energy on the trial solution ( @xmath218 , @xmath219 ) @xmath220+\frac{1}{2 } ( { { \bf b } } ^{out}. { { \bf b } } ^{out}){\mbox{\boldmath $ \theta$}}^{in}\cdot b [ { { \bf c } } ^{out}]\,{^{t}\!b } [ { { \bf c } } ^{out } ] { \mbox{\boldmath $ \theta$}}^{in}=0\ ] ] after setting @xmath221 , we are ready to repeat the operations described above as many times as needed until a fixed point of the transformation ( such that @xmath222 and @xmath223 ) is reached , which solves the problem . the good stability properties of the algorithm , as well as its iterative character , authorize a rather unsophisticated handling of issues raised by the time discretization . as required by the ito- convention , the equation of motion ( [ dync ] ) was integrated using a first - order euler scheme with a time step @xmath224 about 350 times smaller than the period . the time @xmath225 was approximated as the multiple of @xmath226 making the relation ( [ lientemps ] ) best satisfied . similarly no higher order interpolation scheme was devised for estimating with accuracy the output period @xmath207 : it was again simply approximated as the multiple of @xmath226 making periodicity conditions ( [ periodicity1 ] ) best satisfied for the output . however the time step @xmath226 was changed at each iteration of the loop so to maintain the time resolution @xmath71 constant . the efficiency of the method was greatly improved , when seeking solutions of exponents @xmath1 far from @xmath227 , by increasing @xmath1 gradually ( through an increase of the control parameter @xmath187 ) using as first guess solutions of lower but close scaling exponent @xmath228 . in this way convergence toward satisfactory self - similar solutions ( of exponent @xmath1 varying by less than @xmath229 under iteration and pseudo - energy @xmath230 ) was attained in no more than 20 iterations . + we turn now to the presentation of our results . [ fig.2 ] shows the action density @xmath179 for the three models ( i ) , ( ii ) and ( iii ) defined in section [ def ] . values of @xmath0 were adjusted in order to provide the same curvature of @xmath231 at the bottom of the curves , reached evidently at @xmath102 . we see that the variations of the zeroth - order action get sharper on the @xmath232 side ( the only one displayed in fig . [ fig.2 ] ) , as one goes from model ( i ) to model ( iii ) . figures [ fig.3 ] , [ fig.4 ] and [ fig.5 ] , referring respectively to models ( i ) , ( ii ) and ( iii ) , show the normalized coherent field @xmath200 and the random force @xmath233 at increasing values of @xmath1 ( 0.75 , 0.85 and 0.95 ) . in all cases the random force is found to be negative at the leading edge of the pulse , in agreement with the physical picture advocated in @xcite : growth can be enhanced only by frustrating the energy transfer processes . for model ( iii ) , the coherent field itself gets negative at the forefront : noise in that case just helps to prepare the system in an initial condition consisting of a pulse and a negative well in front of it , which then collide . an interesting upshot of our computations is that models like ( ii ) or ( iii ) involving only a local coupling between @xmath20 and @xmath97 escape the disaster met in the framework of model ( i ) , namely a cross - over toward an asymptotic linear growth of @xmath179 with @xmath1 , already perceptible in fig . [ fig.2 ] . such a behaviour forbids the existence of velocity moments at arbitrary orders and is , thus , clearly undesirable . it turns out that the whole shape of @xmath179 for model ( i ) can be pretty well understood from an adiabatic approximation , which is carried out in the appendix [ appendixi ] , where solutions of arbitrary scaling exponents are constructed using adequate time reparametrizations and dilations of the deterministic solution of scaling exponent @xmath2 . the validity of this approximation for the model ( i ) is somehow obvious from fig . [ fig.3 ] , where it can be checked that instantons keep indeed almost the same shape , even for quite sizable variations of @xmath1 . its failure in models ( ii ) and ( iii ) is probably due to too strong deformations of the solutions as @xmath1 increases , again suggested by figs . [ fig.4 ] and [ fig.5 ] . the full non - linear treatment of the problem proposed in this paper was however necessary to reach this quite fortunate conclusion . in order to compute the first order ( in @xmath234 ) correction to the action per unit cascade step ( @xmath51 in the expression ( [ pred_pn ] ) ) of the density of probability @xmath10 , we have to expand the msr action up to quadratic order in fluctuations @xmath235 around the extremal trajectory @xmath29 of scaling exponent @xmath1 and then sum over them in a way which will be explained below . since typical fluctuating paths are not differentiable but rather behave as wiener paths with derivatives of the order @xmath236 , we shall stick to time - discretized expressions in all the following manipulations of the path integral . for the sake of clarity , the superscript 0 referring to the extremal trajectory in the previous section will be taken away , whereas @xmath237 , @xmath238 will be short - hand notations for @xmath239 , @xmath240 , where @xmath241 is the ( small ) time step used in the discretization . + we start from ( [ pathint ] ) and the representation ( [ msraction1 ] ) of the effective action @xmath242 $ ] , expand it to quadratic order in both fluctuations @xmath243 and @xmath235 and then integrate out the fluctuations of the auxiliary field . to begin , the time @xmath65 during which we let the system evolve will be equal to the time @xmath244 needed by the ideal instanton to perform @xmath5 steps along the shell axis . the ideal initial and final configurations @xmath245 and @xmath74 describe then a pulse centered successively around the shells of index @xmath15 and @xmath5 . to quadratic order in deviations from the extremal trajectory , the probability of joining the perturbed endpoints @xmath246 and @xmath247 in the time @xmath65 take the following expression : @xmath248- \delta ^{2}s[{\mbox{\boldmath $ \delta$}{\bf b}}]\right)\,,\ ] ] where the measure of integration is defined as @xmath249\,,\ ] ] the linear variation @xmath250 reduces to the boundary term @xmath251=\frac{1}{\gamma } \ { { \mbox{\boldmath $ \theta$}}_{n-1}\cdot{\mbox{\boldmath $ \delta$}{\bf b}}_{n } -{\mbox{\boldmath $ \theta$}}_{0}\cdot{\mbox{\boldmath $ \delta$}{\bf b}}_{0}\}\,,\ ] ] and the quadratic one reads @xmath252=\frac{\delta \tau } { 2\gamma } \sum_{i=0}^{n-1 } \left [ ( { { \bf b } } _ { i}\cdot { { \bf b } } _ { i})^{-1/2}\,b_{i}^{-1}\ , ( \frac{{\mbox{\boldmath $ \delta$}{\bf b}}_{i+1}-{\mbox{\boldmath $ \delta$}{\bf b}}_{i } } { \delta \tau } -{\cal a}_{i}{\mbox{\boldmath $ \delta$}{\bf b}}_{i})\right]^{2 } + { \mbox{\boldmath $ \delta$}{\bf b}}_{i}.\,{\cal v}_{i}\,{\mbox{\boldmath $ \delta$}{\bf b}}_{i}\,.\ ] ] the drift- and potential- terms in eq.([action_fluct1 ] ) are found to be given by the following relations @xmath253 and @xmath254 where @xmath255 is , as in previous sections , the jacobian matrix of the nonlinear kernel @xmath256 } { ( { { \bf b } } \cdot { { \bf b } } ) ^{1/2}}$ ] . + our task is to perform explicitly the integration over fluctuations @xmath257 at intermediate steps in the path integral ( [ prob_fluct ] ) . first it will be convenient to get rid of the anisotropy of the `` mass '' tensor acting in the kinetic term of @xmath258 @xcite by switching to normalized fluctuating fields defined as @xmath259 in this way the measure in the integral transforms into @xmath260 with @xmath261 in performing this change of variables in @xmath258 , we must pay attention to the fact that @xmath262 , as well as @xmath263 , are potentially of order @xmath264 . one finds that , up to o(@xmath265 corrections ( negligible in the continuum limit ) , @xmath258 becomes : @xmath266=\frac{\delta \tau } { 2\gamma } \sum_{i=0}^{n-1 } \left[q_{i}^{1/2}\,(\frac { { { \bf h } } _ { i+1}- { { \bf h } } _ { i } } { \delta \tau } + { \cal d}_{i}\frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2 } -{\cal a}'_{i } { { \bf h } } _ { i})\right]^{2 } + { { \bf h } } _ { i}.\,{\cal v}'_{i}\ , { { \bf h } } _ { i}\,,\ ] ] where @xmath267\ , \left[1+\left(\frac { { { \bf b } } _ { i+1}\cdot { { \bf b } } _ { i+1 } } { { { \bf b } } _ { i}\cdot { { \bf b } } _ { i}}\right)^{1/2}\,b_{i}^{-1}\,b_{i+1}\right]\,,\ ] ] @xmath268 @xmath269 and @xmath270^{-1}\,\left [ ( { { \bf b } } _ { i+1}. { { \bf b } } _ { i+1})^{1/2}b_{i+1}- ( { { \bf b } } _ { i}\cdot { { \bf b } } _ { i})^{1/2}b_{i}\right]\,.\ ] ] a few simplifications are now in order . on one hand we expand @xmath271 as @xmath272 , where @xmath273 is of order @xmath274 . tracking @xmath275 terms in @xmath258 , we may replace @xmath271 by 1 everywhere in ( [ action_fluct2 ] ) , except for the kinetic term @xmath276 , which according to standard computation rules @xcite in path integrals can be reduced to @xmath277 . on the other hand , exact ways of tracing out quadratic forms like @xmath258 are available only once they are written in terms of the mid - point field @xmath278 rather than @xmath279 ( which amounts going back at this stage to the stratonovich prescription ) . again to order @xmath274 , the substitution of @xmath278 to @xmath279 in ( [ action_fluct2 ] ) is harmless except in the cross - product term @xmath280 which gets into @xmath281 . + putting things together , we arrive at the following expression for the transition probability in the neighborhood of the instanton : @xmath282\,,\ ] ] where we defined @xmath283 and the reduced path integral @xmath284 $ ] as @xmath285=\int _ { { { \bf h } } _ { 0}= { { \bf h } } _ { in}}^ { { { \bf h } } _ { n}= { { \bf h } } _ { f } } { \cal d } { { \bf h } } \ , \exp -\delta^{2}s [ { { \bf h } } ] \,.\ ] ] with @xmath286=\frac{\delta \tau } { 2\gamma } \sum_{i=0}^{n-1 } \left[\frac { { { \bf h } } _ { i+1}- { { \bf h } } _ { i } } { \delta \tau } -{\cal b}_{i}\frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}\right]^{2 } + \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}.\,{\cal v}'_{i}\ , \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}\,,\ ] ] and @xmath287 one easily shows that in the @xmath288 limit @xmath289=d\,a\tau_{f}\,,\ ] ] since the periodicity of the instanton implies that @xmath290 , while @xmath291-\frac { { { \bf n } } [ { { \bf b } } ] \cdot { { \bf b } } } { ( { { \bf b } } \cdot { { \bf b } } ) ^{3/2}}\}\,d\tau\,.\ ] ] however the ensuing expression for @xmath292 is not yet complete to order @xmath293 . this is because we discarded the @xmath294 ito drift - term in our computation of extremal trajectories . the small deviations induced by this term may be neglected in @xmath295 and @xmath296 $ ] which are already first order corrections in a @xmath234-expansion but they must be taken care of in the zeroth order term . rather than @xmath297 , it should read @xmath298 $ ] , where the index @xmath234 denotes a quantity or field evaluated in the presence of the ito - term specified in ( [ itokernel ] ) . to first order in @xmath234 , we can take advantage of the extremum property of @xmath299 and write down @xmath300-s [ { { \bf b } } ^{0}]&= & s_{\gamma } [ { { \bf b } } _ { \gamma}^{0}]-s_{\gamma } [ { { \bf b } } ^{0}]+ s_{\gamma } [ { { \bf b } } ^{0}]-s [ { { \bf b } } ^{0}]\\ & \approx & s_{\gamma } [ { { \bf b } } ^{0}]-s [ { { \bf b } } ^{0}]= \displaystyle{\frac{1}{\gamma}\int _ { 0}^{\tau_{f}}{\mbox{\boldmath $ \theta$}}^{0}. \frac { { { \bf n } } ( { { \bf b } } ^{0}]- { { \bf n } } _ { \gamma } [ { { \bf b } } ^{0 } ] } { ( { { \bf b } } ^{0}\cdot { { \bf b } } ^{0})^{1/2}}\,d\tau } , \end{aligned}\ ] ] where the last identity just comes from the expression ( [ msraction2 ] ) of the action . rearranging things under the assumption ( satisfied by each of the particular models that we considered ) that the entry @xmath301 of the matrix @xmath41 involves homogeneous monomials of degree @xmath302 built up only from components @xmath303 with @xmath304 ( one has @xmath305 for model ( i ) and @xmath306 for models ( ii ) and ( iii ) defined in section [ def ] ) , one finds that @xmath292 in ( [ prob_fluct2 ] ) should finally be understood as : @xmath307 this preliminary work being done , the discussion will now concentrate on the reduced path integral @xmath308 $ ] . we restrict first our attention to the case of fixed endpoints @xmath309 . the instantons found in the previous sections are physically satisfactory only if the quadratic functional @xmath310 $ ] is positive for all @xmath311 such that @xmath312 ( we shall see later on that this is not a sufficient condition in the present problem ) . according to standard results of functional analysis @xcite , positiveness of @xmath310 $ ] is tantamount to the absence of points conjugate to the origin during the whole time interval @xmath313 $ ] . recall that the definition of conjugate points goes as follows : @xmath75 being the dimension of the space ( here the number of shells ) we construct @xmath75 initial conditions @xmath314 such that @xmath315 where @xmath316 is a shell index running as @xmath317 between 1 and @xmath75 and let them evolve under the euler - lagrange equations derived from @xmath310 $ ] . the time @xmath318 is said to be conjugate to the origin if the system formed by the @xmath75 vectors @xmath319 gets degenerate there . one can build a matrix @xmath320 such that @xmath321 , in terms of which the initial conditions read @xmath322 while @xmath323 ( for @xmath324 ) is obtained from @xmath325 and @xmath320 through a matrix euler - lagrange equation , derived from ( [ action_fluct3 ] ) and conveniently cast into the following form : @xmath326 where @xmath327 can be seen as a matrix momentum . in the absence of conjugate points , @xmath328 never vanishes except at the origin and one gets the following simple expression for the reduced path integral , provided the @xmath288 limit is ultimately taken : @xmath329 = \left(\frac{1}{2\pi \gamma}\right)^{d/2}\ , \frac{1}{\sqrt{\mbox{det}u(\tau_{f})}}\,.\ ] ] details on this result , which may be found in many textbooks on path integrals @xcite , are provided in the appendix [ appendixii ] . a very nice feature of the proof of the connection of the positiveness of @xmath310 $ ] with the absence of conjugate points is that it provides also an efficient way for computing the reduced path integral with @xmath330 arbitrary ( but still small naturally ) . indeed the main idea consists in adding to @xmath310 $ ] a boundary term of the form @xmath331 where @xmath332 is a symmetric matrix at all times and then selecting the right @xmath332 such that @xmath333 becomes a perfect square . in order to achieve this task , @xmath334 must be a solution of a matrix ricatti equation ( see again the appendix b for details ) which , upon the substitution of a new unknown matrix @xmath320 defined implicitely by the relation ( for @xmath335 ) : @xmath336 is found to be nothing but the matrix euler equation ( [ matrixel ] ) . it follows that the result ( [ z_fixed ] ) may be extended to the case of an arbitrary final configuration @xmath330 as @xmath337 = \left(\frac{1}{2\pi \gamma}\right)^{d/2}\ , \frac{1}{\sqrt{\mbox{det}u(\tau_{f})}}\ , \exp -\frac{1}{2\gamma } { { \bf h } } _ { f}\cdot w(\tau_{f } ) { { \bf h } } _ { f}\,,\ ] ] where @xmath338 and @xmath339 are to be computed by letting both matrices evolve from @xmath63 to @xmath340 according to ( [ matrixel ] ) and ( [ lienwu ] ) . there are some subtleties about the choice of initial conditions and proper counting of the number of unknowns whose discussion we prefer to relegate in the appendix [ appendixii ] . we are now in a good position to compute the next to leading order term @xmath51 in the expansion of @xmath341 in powers of @xmath234 . we shall obtain an estimate for @xmath342 by summing over all the trajectories which lead to the same growth of the pulse as the ideal instanton @xmath29 after @xmath5 cascade steps . in the small @xmath234 limit , all statistically relevant trajectories remain close to @xmath29 and we may define unambiguously their `` arrival '' time at the shell of index @xmath5 as the first time @xmath343 such that @xmath344 where @xmath345 reduces to a linear combination of stable `` irrelevant '' modes @xmath346 for @xmath170 . up to multiplicative factors growing at most algebraically with @xmath5 , we can then write @xmath10 as the following integral over the arrival time and the position of the endpoint @xmath347 the density of probability in the right hand side of this expression is known from ( [ prob_fluct2 ] ) and ( [ z_mobile ] ) , where @xmath348 should be taken equal to @xmath349 and @xmath350 equal to @xmath351 . calling @xmath352 the time delay , we get the following relation between @xmath235 and @xmath345 , valid up to @xmath353 corrections : @xmath354 note that @xmath241 scales typically like @xmath355 in the semi - classical limit . we may thus , to leading order , replace @xmath349 by @xmath356 and identify @xmath235 with @xmath357 everywhere in the integrand of the right hand side of ( [ finalp_nz1 ] ) , except in @xmath358 ( the combination appearing in the exponential prefactor of ( [ prob_fluct2 ] ) ) which deserves more care . it follows from the definition of the action that ( again up to @xmath353 corrections ) @xmath359 and from ( [ lienfluct ] ) ( together with the condition @xmath360 ) that : @xmath361 when summing these two equations , linear terms in @xmath241 disappear as expected and we are left after some rearrangements with the remainder , of quadratic order in fluctuations : @xmath362 where we set @xmath363 , so that the time delay @xmath241 is simply expressed in terms of @xmath235 as @xmath364 . + the end of our theoretical considerations is reached with the following expression for @xmath342 : @xmath365 where @xmath366 ( so that @xmath367 ) , @xmath368 , with @xmath332 defined in the last subsection and @xmath369 we see that the condition of positiveness of the matrix @xmath370 must be supplemented by the condition of positiveness of the restriction @xmath371 of @xmath372 to the @xmath373-dimensional space orthogonal to @xmath374 , in order to make the instantons found in the preceding section physically meaningful . provided these two requirements are met , @xmath51 is obtained as @xmath375 which is the main result of this section . + tracing back all the steps leading to ( [ expr_s1z ] ) , one could object to our starting point ( [ finalp_nz1 ] ) the fact that fluctuations of the initial endpoint are not taken into account . this could be done at the expense of rather more cumbersome formulae for the reduced path integral @xmath376 $ ] when both @xmath377 and @xmath378 do not vanish . however , we believe on physical grounds that the blowing - up associated with the instanton washes out any influence of the fluctuations at large scales on the part of the action scaling linearly with the number of steps @xmath5 . therefore the expression ( [ expr_s1z ] ) should be exact . the most difficult part of the computation of @xmath51 lies in the evaluation of @xmath379 and @xmath380 ( as defined in the preceding subsection ) , which requires a good control of all the eigenvalues of these two matrices . numerical instabilities could be avoided for a time long enough to get a precise estimate of @xmath381 by using the exact expression of the matrix euler equation ( [ matrixel ] ) and the relation ( [ lienwu ] ) between @xmath332 and @xmath370 . in this problem , there are two goldstone modes associated with uniform rescaling and time translation of the instanton : as a consequence , the matrix @xmath382 ( resp . @xmath383 ) is expected to have two eigenvalues scaling like @xmath25 ( resp . like @xmath384 ) ( in order to obtain the simplest transcription of these symmetries , one has to go back to the original fluctuation field @xmath385 and the change of variable influences eigenvalues of @xmath370 and @xmath372 ) . when restricting @xmath372 to the @xmath373-dimensional space orthogonal to @xmath374 ( @xmath386 ) , one looses one of the eigenvalues scaling like @xmath384 , so that there remains an algebraic factor @xmath387 in the product @xmath388 , which we had to take away by hand in order to make more conspicuous the leading exponential growth of this quantity . to give an idea of the accuracy of our procedure and confirm the soundness of the intricate formula ( [ expr_s1z ] ) that was proposed for @xmath51 , we show in fig . [ fig.6 ] the behaviour of various relevant quantities in the case of model ( ii ) , and for a moderate scaling exponent @xmath389 . it is observed in the picture on the top that the logarithmic derivatives of @xmath390 and @xmath391 exhibit a linear behaviour with almost opposite slopes . it can be shown by considering simpler and exactly solvable models for quadratic fluctuations without inter - shell couplings , that this strange effect mostly reflects the stiff variations suffered by the noise variance on any shell , as the latter moves back from the leading edge of the instanton to its rear end . the argument is presented in the appendix [ appendixiii ] , since it may help the reader to get a feeling for the order of magnitudes at play in both @xmath370 and @xmath332 matrices . + it could be tempting at this level to approximate @xmath51 as @xmath392 this expression would come out if , without great physical justification , it were decided to sum over all positions of the final endpoint in the path integral formulation in order to estimate the volume of the basin of attraction of the instanton . we found however that the matrix momentum @xmath393 develops invariably a negative eigenvalue after some time , so that the projection onto the restricted phase space introduced in the previous subsection is a necessary step for restoring the statistical stability of the instanton . further , even before this instability occurs , there was found to be a residual @xmath394 term in @xmath395 which forbids any reliable estimate for @xmath51 to be deduced from ( [ simple_s1z ] ) . the picture at the bottom of fig . [ fig.6 ] shows by contrast that the more precise quantity @xmath396 quickly settles to a perfect exponential growth , once algebraic transients have been factored out . note that the positiveness of both @xmath370 and @xmath371 could be checked in any instance . + in all the models that we investigated , we found that the first order correction @xmath51 to the action takes an approximate parabolic shape of positive or negative concavity , centered around a value of @xmath1 different from @xmath2 . in the case of the model ( ii ) , the concavity of @xmath51 is opposite to the one of @xmath179 and the maximum of @xmath51 is reached at a scaling exponent @xmath397 , significantly lower than @xmath8 . this means that the minimum of the total action density @xmath398 ( for values of @xmath0 such that @xmath399 remains concave as it should ) is displaced toward the side of larger exponents , just as a result of fluctuations . the trend is just the opposite for the model ( iii ) , where @xmath51 presents this time the same concavity as @xmath179 and a minimum on the left side of @xmath2 . [ fig.7 ] shows the graph of @xmath400 that is obtained for the model ( ii ) and for the value @xmath401 , which we believe to be of some relevance for the goy model ( see section [ discussion ] ) . the systems we have analyzed here were introduced to describe inertial singular structures of the goy model . to check their physical relevance , we first have to define a class of events observed in full simulations of the goy model which are likely to be the best candidates for a description in terms of instantons . it is clear that relative maxima of the instantaneous energy flux @xmath402 ( with @xmath403 ) are useful observables for tracking the passage of coherent structures across the whole inertial range . but their total number is found to grow with @xmath5 as @xmath404 , due to the acceleration of time scales typical of the kolmogorov energy cascade . one may consider that they develop on tree - like patterns in the @xmath405-plane , which are renewed at each turn - over of the large scales ( totls ) . we say that such trees provide a realization of the propagation of a coherent event from shell @xmath406 to shell @xmath407 , whenever the @xmath408 nodes of the tree closest in time to their supposedly common ancestor on the shell of lower index @xmath406 appear in the order of increasing shell index . in order to discard too weak , and therefore irrelevant , events , we imposed that @xmath409 be greater than half the mean energy flux . figure [ fig.8 ] shows the logarithm of the histogram of the logarithmic amplitudes @xmath410 for all relative maxima on one hand and for the restricted class of coherent events defined above on the other hand , with @xmath411 ( far enough from the forcing range ) and @xmath412 ( well in the inertial range ) . the reynolds number of the simulation is @xmath413 . statistics have been run over @xmath414 totls , and in average there are three `` coherent lines '' for two totls . we note that the statistics of coherent events is very close to log - normal for @xmath415 . + the effective exponent @xmath1 of a coherent burst after @xmath408 cascade steps is obtained via the relation @xmath416 . if anomalous scaling is preserved in the @xmath417 limit , the pdf of scaling exponents @xmath1 should behave at large cascade lengths as @xmath418 , where generically @xmath399 will present a quadratic minimum at some @xmath419 , with an expansion around @xmath419 that we write as @xmath420 . the histogram of the variable @xmath421 should consequently evolve , as @xmath408 increases , towards a gaussian shape , whose center @xmath422 and variance @xmath423 grow linearly with @xmath5 and relate to @xmath419 and @xmath424 as @xmath425 the actual behaviours of @xmath426 and @xmath427 , obtained from a very high reynolds number simulation ( @xmath428 , which sets the dissipative scale at the shell index @xmath429 ) are plotted in fig . [ fig.9 ] . error bars were estimated by varying the range of the quadratic fit to the logarithm of the histogram of @xmath421 , as well as the domain of initial amplitudes @xmath430 used to construct this histogram . it appears that the two graphs are rather far from simple straight lines : this is especially true for the variance , whose graph from concave gets convex beyond the shell index 15 . from the investigation of lower reynolds numbers @xmath431 and @xmath432 we could deduce that this transition occurs at a shell index @xmath433 always of the order of @xmath434 and defines a clear - cut boundary between the inertial range and a surprisingly wide pre - viscous range . we believe that the direct action of viscosity on intense bursts starts to show up only at the shell index @xmath435 , beyond which the local slope of the @xmath422-graph ceases to vary and points to a value of the average growth exponent of coherent structures precisely equal to @xmath227 . the fact that @xmath436 lies rather far from @xmath437 means that the cut - off imposed by viscosity exerts a long - range influence on the statistics of the random environment seen by a coherent structure . a decent linear regime for both the drift and the variance is observed in the range @xmath438 , from which we get the two estimates @xmath439 and @xmath440 . assuming that the fitting range @xmath441 provides the best clue to the asymptotic scaling of the inertial range , one gets the second set of values @xmath442 and @xmath443 . it appears that the physics of the previscous range is quite well reproduced within our modelling ( ii ) of the incoherent background . by choosing @xmath444 ( a value hopefully small enough to fall within the range of validity of semiclassical approximations ) , one obtains , as fig . [ fig.7 ] shows , an almost perfect parabolic shape of @xmath445 , with a maximum reached at @xmath446 and a curvature @xmath447 . however the parameter @xmath234 can not be adjusted so as to account for the higher values of @xmath424 and @xmath419 characterizing the inertial range . it would seem that in this range of scales it gets necessary to assume some bias in the incoherent fluctuations boosting the increase of the renormalized value of @xmath419 , while keeping the noise width small . we have developed a general scheme for computing numerically self - similar instantons in scale invariant stochastic dynamical systems . as concerns the physics of the goy model , we believe that the bunch of results presented here give a strong support to the relevance of an approach focusing from the outset on structures in order to understand intermittency and treating the rest of the flow as a noise of weak amplitude . in particular the trend toward log - normal statistics of coherent structures is nicely recovered . the detailed study of various stochastic extensions of the goy model shows that the resulting pdf of scaling exponents of singular strucures is very sensitive to the hypothesis made on the coupling of noise to the velocity gradient fiels . + we hope that our approach will be useful for attacking the 3d - navier - stokes dynamics along similar lines , once an adequate decomposition of the flow into coherent and incoherent parts will have been introduced . an application of the method to the kraignan s model of passive scalar advection formulated on a lattice of shells has already been attempted @xcite . it has given encouraging results with regard to the validity of a semi - classical analysis , even in situations where a small parameter ( like @xmath0 in the present problem ) is missing . we thank l. biferale , g. falkovich , v. hakim , v. lebedev , p. muratore , d. vandembroucq and r. zeitak for useful discussions or suggestions at various stages of this work . j .- g. has been partly supported by the cnrs / crtbt and by grants from the israeli science foundation and the minerva foundation ( germany ) . siggia , j. fluid . mech . * 107 * , 375 ( 1981 ) ; z .- s . she , e. jackson , s. orszag , proc . soc . lond . * 434 * , 101 ( 1991 ) ; j. jimnez , a. wray , p.g . saffman , r. rogallo , j. fluid . mech . * 255 * , 65 ( 1993 ) ; f. belin , j. maurer , p. tabeling , h. willaime , j.phys.ii france * 6 * , 573 ( 1996 ) . reviewed by k. gawedzki in _ advances in turbulence vii _ ( editor : u. frisch , kluwer academic publishers , 1998 ) , 493 - 502 . for an introduction to shell models , see t. bohr , m.h . jensen , g. paladin , a. vulpiani , _ dynamical systems approach to turbulence and extended systems _ , cambridge university press , cambridge ( 1998 ) . siggia , phys . a * 17 * , 1166 ( 1978 ) ; t.nakano , prog . * 79 * , 569 ( 1988 ) ; g. parisi , rome preprint ( 1990 ) . gilson , t. dombre , phys . . lett . * 79 * , 5002 ( 1997 ) . freidlin and a.d . wentzell , _ random perturbations of dynamical systems _ , springer , new - york ( 1984 ) for the mathematical foundations of such computations . l. n. lipatov , jetp lett . * 24 * , 157 ( 1976 ) , sov . jetp * 44 * , 1055 ( 1976 ) , jetp lett . * 25 * , 104 ( 1977 ) , sov . jetp * 45 * , 216 ( 1977 ) . g. falkovich , i. kolokolov , v. lebedev , a.migdal , phys . e. * 54 * 4896 ( 1996 ) . v. gurarie , a. migdal , phys . e , * 54 * 4908 ( 1996 ) . e. balkovsky , g. falkovich , i. kolokolov , v. lebedev , phys . lett . * 78 * 1452 - 1455 ( 1997 ) . m.chertkov , phys . e , * 55 * , 2722 ( 1997 ) . e. balkovsky , v. lebedev , phys . e. * 58 * , 5776 ( 1998 ) . t. dombre , j .- l . gilson , physica d * 111 * , 265 ( 1998 ) . r. benzi , l. biferale , g. parisi , physica d * 65 * , 352 ( 1993 ) . u. frisch , _ turbulence : the legacy of a.n . kolmogorov _ , cambridge university press , cambridge ( 1995 ) . gardiner , `` handbook of stochastic methods for physics , chemistry , and the natural sciences '' , berlin , springer verlag , series in synergetics , ( 1982 ) . martin , e. siggia , h. rose , phys . rev . a * 8 * 423 ( 1973 ) ; c. de dominicis , j. phys . ( paris ) * 37 * , 247 ( 1976 ) ; h. janssen , z. phys . b * 23 * , 377 ( 1976 ) . f. langouche , d. roekaerts , e. tirapegui , `` functional integration and semiclassical expansions '' , d.reidel publishing co ( dodrecht , holland ) ( 1982 ) . fomin , s.v . guelfand , `` calculus of variations '' , prentice hall ( 1963 ) . w. dittrich , m. reuter , `` classical and quantum dynamics '' , berlin springer verlag , 2nd edition ( 1994 ) . actually , this change of variable is not strictly necessary . it may be shown to roughly `` square root '' the range of values spanned by the coefficients of matrices @xmath370 and @xmath332 introduced in subsection [ fluctformal ] . the ensuing reduction of numerical stiffness allows computations at larger sizes of the shell lattice . however , the switch from @xmath235 to @xmath448 gets totally impracticable when the matrix @xmath41 can not be easily inverted . schulman , `` techniques and applications of path integration '' , john wiley and sons , ( 1981 ) . l. biferale , i. daumont , t. dombre , a. lanotte , `` about coherent structures in random shell models for passive scalar advection '' , to be published in phys . e ( rc ) ( 1999 ) . we carry out in this appendix the adiabatic approximation alluded to in section [ extsolimplementing ] . we look for self - similar instantons within the restricted manifold of configurations of the type @xmath449 where @xmath29 is the deterministic solution of scaling exponent @xmath2 , @xmath450 can be thought of as a `` proper '' time referring to the actual position of the pulse . the two variables @xmath451 and @xmath452 parameterize then local changes of speed and amplitudes of the pulse , which keeps the same shape as in the absence of noise . note that if ( [ trialinst ] ) is to represent a self - similar instanton of scaling exponent @xmath453 , @xmath454 must obey the constraint @xmath455 we plug now the ansatz ( [ trialinst ] ) in the equation of motion ( [ ito_tau ] ) and project it onto the two directions @xmath456 and @xmath457 . we get by doing so @xmath458 if the other dimensions of the configuration space are neglected , ( [ reducedstoch1 ] ) and ( [ reducedstoch2 ] ) form a closed two - dimensional system , which may be rewritten as @xmath459 where correlation functions of @xmath460 and @xmath461 read @xmath462 with @xmath463 the gaussian action density associated to one cascade step within this restricted stochastic system is given by @xmath464 once expressed in terms of the diffusing variables @xmath452 and @xmath465,it becomes @xmath466+\frac{d\tau}{d\tilde{\tau } } ( v^{-1})_{22}\nonumber \\ & & -\left[\dot{x}\,(v^{-1})_{12}+(v^{-1})_{22}\right]\}.\end{aligned}\ ] ] the extremization of @xmath467 with respect to @xmath468 leads to @xmath469^{1/2},\ ] ] and to an effective action for the remaining variable @xmath470 @xmath471&=&\int_{0}^{t_{0}}d\tilde{\tau } \ , \{\left[(\dot{x}^{2}(v^{-1})_{11}+2\dot{x}\,(v^{-1})_{12}+ ( v^{-1})_{22})(v^{-1})_{22}\right]^{1/2}\nonumber \\ & & -\left[\dot{x}\,(v^{-1})_{12}+(v^{-1})_{22}\right]\}.\end{aligned}\ ] ] assuming the coefficients @xmath472 , @xmath473 and @xmath474 to be almost constant inside the time interval @xmath31 , one deduces an analytic expression for @xmath179 from @xmath475 by just replacing in the integral @xmath476 by @xmath477 ( which follows from ( [ constraintx ] ) ) . one gets in particular for large enough @xmath478 , @xmath479 i.e , a linear behaviour as observed for the true solution of model ( i ) . we derive the formal expression of the reduced path integral @xmath480 $ ] given in eq . ( [ z_mobile ] ) of the section [ fluct ] of the text . the proof is presented in many textbooks on path integrals but , as far as we know , always using a continuous definition of time . this leaves some ambiguity in the right equations that matrices @xmath370 and @xmath332 should obey , once time is discretized for computing purposes . we found that this issue is crucial mostly for evaluating @xmath332 and preserving its symmetry properties . this is why we feel it useful to show how every step of the proof given in the continuum limit receives an exact transcription in the discrete time case . + we start from the quadratic functional : @xmath481=\frac{\delta \tau } { 2\gamma } \sum_{i=0}^{i = n-1 } \left\{\left[\frac { { { \bf h } } _ { i+1}- { { \bf h } } _ { i } } { \delta \tau } -{\cal b}_{i}\frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}\right]^{2 } + \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}.\,{\cal v}'_{i}\ , \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}\right\}\,,\ ] ] upon the addition of the boundary term @xmath482 , it becomes without any approximation : @xmath483&=&\frac{\delta \tau } { 2\gamma } \sum_{i=0}^{i = n-1}\ { \left[\tilde{q}_{i}^{1/2}\left(\frac { { { \bf h } } _ { i+1}- { { \bf h } } _ { i } } { \delta \tau}-{\tilde{q}_{i}}^{-1}({\cal b}_{i}-\frac{w_{i}+w_{i+1}}{2})\ , \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}\right)\right]^{2 } \nonumber \\ & & \qquad + \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}.\,\tilde{\cal v}_{i}\ , \frac { { { \bf h } } _ { i+1}+ { { \bf h } } _ { i}}{2}\}\,,\end{aligned}\ ] ] where @xmath484 and @xmath485 we see that @xmath486 $ ] reduces to the time integral of a single ( positive ) square , if and only if @xmath334 is such that for all @xmath335 @xmath487 we note that ( [ ricatti ] ) forces @xmath334 to remain symmetric for all time , provided that @xmath488 ( arbitrary at this stage ) is chosen to be such . in order to solve the above ricatti matrix equation , one makes the following change of matrix variable @xmath489 > from the expression ( [ defqtilde ] ) of @xmath490 , it is easily shown that ( [ defu ] ) is equivalent to the equation ( [ lienwu ] ) quoted in the text . furthermore , by multiplying both sides of ( [ ricatti ] ) by @xmath491 on the right , one gets for @xmath492 @xmath493 by half - summing the two relations yielded by ( [ secondlienwu ] ) at subsequent values @xmath494 and @xmath58 of the temporal index ( with then @xmath495 ) , and using ( [ lienwu ] ) after noticing that @xmath496 , one can eliminate @xmath332 and check that @xmath370 obeys the matrix euler equation ( [ matrixel ] ) , as promised in the text . + so far , we have proven that , as long as the matrix @xmath370 may be inverted , the positiveness of @xmath258 is guaranteed , since in that case the matrix @xmath332 exists at all times ( from ( [ defu ] ) ) and allows one to transform the initial quadratic form into the time integral of a single square . we show now how these matrices lead to a compact expression of @xmath497 $ ] . we first note that ( [ defu ] ) and ( [ secondlienwu ] ) provide @xmath498 relations for @xmath499 unknowns @xmath500 , @xmath501 . this gives much freedom in the choice of @xmath488 and @xmath502 . in the particular case of @xmath503 , it is convenient to set @xmath504 and @xmath505 ( which should be understood as the limit as @xmath506 of @xmath507 and @xmath508 , so that @xmath488 is indeed symmetric ) . a quick inspection of ( [ defu ] ) and ( [ secondlienwu ] ) reveals that @xmath320 and @xmath334 behave then respectively as @xmath509 and @xmath510 to leading order in @xmath274 for @xmath511 . one has for instance the exact result @xmath512 . recall that the quantity we wish to estimate reads now : @xmath513 = { \e}^{-\frac{1}{2\gamma } { { \bf h } } _ { f}.w_{n } { { \bf h } } _ { f } } \int { \cal d } { { \bf h } } \,\delta^{d } ( { { \bf h } } _ { n}- { { \bf h } } _ { f } ) { \e}^ { -\frac{1}{2}\sum_{i=0}^ { n-1}{\mbox{\boldmath $ \psi$}}_{i+1}.\tilde{q}_{i}{\mbox{\boldmath $ \psi$}}_{i+1}}\,,\ ] ] where we defined the new field @xmath514 ( for @xmath335 ) as @xmath515 and the measure of integration @xmath516 as @xmath517 with the help of ( [ defu ] ) , the transformation ( [ defpsi ] ) may be rewritten as @xmath518 this relation is easily inverted by setting @xmath519 , which gives @xmath520 so that we deduce ( under the assumption @xmath503 ) , for @xmath335 , @xmath521\,.\ ] ] since @xmath522 is linearly related to the @xmath523 s of index @xmath524 lower than @xmath58 , only the diagonal blocks @xmath525 enter the jacobian of the transformation and one has : @xmath526 to enforce the boundary condition @xmath527 at time @xmath348 in terms of the new variables @xmath528 , we introduce the usual integral representation of the @xmath529-function : @xmath530}\,.\ ] ] after performing the gaussian integration over the @xmath528 s , one arrives at @xmath531 = \prod_{i=0}^{n-1}\left\{\frac{\det ( u_{i}+u_{i+1})}{\det 2u_{i } } \times \frac{1}{\sqrt{\det \tilde{q}_{i}}}\right\}\ , { \e}^{-\frac{1}{2\gamma } { { \bf h } } _ { f}.w_{n } { { \bf h } } _ { f}}\times \int\,\frac{d{\mbox{\boldmath $ \alpha$}}}{(2\pi ) ^{d}}{\e}^{-i{\mbox{\boldmath $ \alpha$}}\cdot { { \bf h } } _ { f } } { \e}^{-\frac{\delta \tau \gamma}{2}{\mbox{\boldmath $ \alpha$}}.g{\mbox{\boldmath $ \alpha$}}}\,,\ ] ] where @xmath532\,^{t}u_{n}\,.\ ] ] this awkward non local operator @xmath533 is greatly simplified when the singular initial condition already mentioned : @xmath507 , @xmath534 with @xmath506 is adopted . in that case , @xmath533 is completely dominated by the first term of the series in the r.h.s . of ( [ defg ] ) which diverges as @xmath535 and , to leading order in @xmath536 , one has @xmath537 the summation over @xmath538 can then be done and , since @xmath539 vanishes in the @xmath506 limit , one gets @xmath540 = \left(\frac{1}{2\pi \gamma \delta \tau}\right)^{d/2 } \frac{\det ( u_{0}+u_{1})}{\det u_{n } } \prod_{i=1}^{n-1}\left\{\frac{\det ( u_{i}+u_{i+1})}{\det 2u_{i } } \times \frac{1}{\sqrt{\det \tilde{q}_{i}}}\right\}\ , { \e}^{-\frac{1}{2\gamma } { { \bf h } } _ { f}.w_{n } { { \bf h } } _ { f}}\,.\ ] ] note that all the manipulations presented in this appendix were devoid of any approximation . it is finally a straightforward matter ( details will be skipped here ) , to check that in the @xmath288 limit , the infinite product in front of the exponential in ( [ z_mobile4 ] ) reduces to @xmath541 , making thereby ( [ z_mobile4 ] ) identical to the result ( [ z_mobile ] ) quoted in the text . we consider the following quadratic action : @xmath542=\frac{1}{2}\sum_{n=0}^{d-1}\,\int_{0}^{\tau } d\tau \,\frac{\dot{x}_{n}^{2}}{a_{n}(\tau ) } , \ ] ] where , in order to mimic the stochastic models studied in this paper , the variance @xmath543 evolves on each shell as @xmath544 i.e. , as the variance of noise , @xmath545 , for the model ( ii ) . in ( [ c2 ] ) , @xmath29 may be thought of as the deterministic self - similar solution , and @xmath543 can therefore be cast into the form @xmath546 where the function @xmath547 satisfies @xmath548 because of the periodicity of the shell lattice . let us assume that the instanton is centered around the shell of index @xmath549 at time @xmath63 . at its leading edge ( @xmath550 ) , @xmath551 decreases very abruptly as @xmath552 , with r=@xmath553 and @xmath554 a constant of order 1 . this essential singularity comes from the necessity of balancing @xmath555 with the dominant term @xmath556 of the non - linear kernel of the goy model in this range of scales . in the trail of the instanton ( @xmath557 ) , one has a much smoother behaviour @xmath558 . when the shell lattice is periodized , the leading edge and the tail of the instanton have to be glued together and the locus of matching , as well as the residual amplitude of @xmath551 at that place , will be imposed by the side supporting the slowest variations of @xmath551 . we conclude that in a cyclic chain containing @xmath75 shells , most of them reside in the exponential tail of the instanton , so that we may write ( again under the hypothesis of an instanton initially centered around the origin @xmath549 and with a shell index @xmath5 defined between 0 and @xmath559 ) @xmath560 thus , the range of values spanned by the function @xmath561 is very large and scales with the total number of shells like @xmath562 . for shells on the exponential ramp , @xmath543 first decreases exponentially in time like @xmath563 ( because @xmath551 decreases like @xmath564 in this region ) and goes by a sharp maximum of order @xmath565 at the time @xmath566 when the center of the instanton reaches the corresponding shell . then it starts again to decrease exponentially . + having understood these basic dynamical features , we can compute the matrices @xmath370 and @xmath332 introduced in subsection [ fluctformal ] for the quadratic action given by ( [ c1 ] ) . to make contact with the normalized field @xmath448 used in the real problem , we switch from the variable @xmath567 to the variable @xmath568 . this transforms the original action into @xmath569=\frac{1}{2}\sum_{n=0}^{d-1}\,\int_{0}^{\tau } d\tau \ , ( \dot{y}_{n}+\frac{1}{2}\frac{\dot{a}_{n}}{a_{n}}y_{n})^{2}.\ ] ] since there is no inter - shell coupling , the matrices @xmath370 and @xmath332 are diagonal in the shell index . the extremization of @xmath570 $ ] with respect to @xmath571 leads to the couple of first - order differential equations @xmath572 one has simply @xmath573 and @xmath574 . the solution of eqs . ( [ c7 ] ) and ( [ c8 ] ) under the initial conditions @xmath575 and @xmath576 is @xmath577 as long as @xmath578 , the instanton has not passed through the shell of index @xmath5 and the integral in the right hand side of ( [ c10 ] ) is dominated by the neighborhood of the lower bound @xmath63 . we deduce that for indices @xmath579 @xmath580 by contrast , when @xmath25 gets larger than @xmath343 ( by some units of time @xmath31 ) , the integral in the right hand side of ( [ c10 ] ) is dominated by the neighborhood of the time @xmath343 where @xmath581 takes its maximal value . we get @xmath582 where @xmath583 is a number of order 1 . it follows that for indices @xmath584 , @xmath585 where we used the estimate ( [ c5 ] ) for @xmath586 . since @xmath587 for @xmath588 and @xmath589 for @xmath590 , we conclude that @xmath591 increases exponentially like @xmath592 . but this property is not shared by @xmath390 or @xmath391 considered individually . indeed , since the number of shells crossed by the instanton increases linearly in time , eqs . ( [ c11 ] ) and ( [ c13 ] ) show that @xmath593 while from eqs . ( [ c12 ] ) and ( [ c14 ] ) , @xmath594 this argument captures apparently a good part of the physics of fluctuations around a moving self - similar system , though badly treating hybridization effects between neighbouring shells . it also explains how large ( resp . small ) numbers are generated in the spectrum of the matrix @xmath370 ( resp . @xmath332 ) and why in practice one does not have much freedom in the choice of the total number of shells @xmath75 .	it has been shown recently that intermittency of the gledzer ohkitani yamada ( goy ) shell model of turbulence has to be related to singular structures whose dynamics in the inertial range includes interactions with a background of fluctuations . in this paper we propose a statistical theory of these objects by modelling the incoherent background as a gaussian white - noise forcing of small strength @xmath0 . a general scheme is developed for constructing instantons in spatially discrete dynamical systems and the cramr function governing the probability distribution of effective singularities of exponent @xmath1 is computed up to first order in a semiclassical expansion in powers of @xmath0 . the resulting predictions are compared with the statistics of coherent structures deduced from full simulations of the goy model at very high reynolds numbers .
the _ howard s algorithm _ ( also called _ policy iteration algorithm _ ) is a classical method for solving a discrete hamilton - jacobi equation . this technique , developed by bellman and howard @xcite , is of large use in applications , thanks to its good proprieties of efficiency and simplicity . it was clear from the beginning that in presence of a space of controls with infinite elements , the convergence of the algorithm is comparable to newton s method . this was shown under progressively more general assumptions @xcite until to @xcite , where using the concept of _ slant differentiability _ introduced in @xcite , the technique can be shown to be of semi - smooth newton s type , with all the good qualities in term of superlinear convergence and , in some cases of interest , even quadratic convergence . in this paper , we propose a parallel version of the policy iteration algorithm , discussing the advantages and the weak points of such proposal . in order to build such parallel algorithm , we will use a theoretical construction inspired by some recent results on domain decomposition ( for example @xcite ) . anyway , for our purposes , thanks to a greater regularity of the hamiltonian , the decomposition can be studied just using standard techniques . we will focus instead on convergence of the numerical iteration , discussing some sufficient conditions , the number of iteration necessary , the speed . parallel computing applied to hamilton jacobi equations is a subject of actual interest because of the strict limitation of classical techniques in real problems , where the memory storage restrictions and limits in the cpu speed , cause easily the infeasibility of the computation , even in cases relatively easy . with the purpose to build a parallel solver , the main problem to deal with is to manage the information passing through the threads . our analysis is not the first contribution on the topic , but it is an original study of the specific possibilities offered by the policy algorithm . in particular some non trivial questions are : is convergence always guaranteed ? in finite time ? with which rate ? which is the gain respect to ( the already efficient ) classical howard s algorithm ? in literature , at our knowledge , the first parallel algorithm proposed was by sun in 1993 @xcite on the numerical solution of the bellman equation related to an exit time problem for a diffusion process ( i.e. for second order elliptic problems ) ; an immediately successive work is @xcite by camilli , falcone , lanucara and seghini , here an operator of the semilagrangian kind is proposed and studied on the interfaces of splitting . more recently , the issue was discussed also by zhou and zhan @xcite where , passing to a quasi variational inequality formulation equivalent , there was possible a domain decomposition . our intention is to show a different way to approach the topic . decomposing the problem directly in its differential form , effectively , it is possible to give an easy and consistent interpretation to the condition to impose on the boundaries of the sub - domains . thereafter , passing to a discrete version of such decomposed problem it becomes relatively easy to show the convergence of the technique to the correct solution , avoiding the technical problems , elsewhere observed , about the manner to exchange information between the sub - domains . in our technique , as explained later , we will substitute it with the resolution of an auxiliary problem living in the interface of connection in the domain decomposition . in this way , data will be passed implicitly through the sub - problems . the paper is structured as follows : in section 2 we recall the classic howard s algorithm and the relation with the differential problem , focusing on the case of its control theory interpretation . in section 3 , after discussing briefly the strategy of decomposition , we present the algorithm , and we study the convergence . section 4 is dedicated to a presentation of the performances and to show the advantages with respect the non parallel version . we will end presenting some possible extensions of the technique to some problems of interest : reachability problems with obstacle avoidance , max - min problems . the problem considered is the following . let be @xmath0 bounded open domain of @xmath1 ( @xmath2 ) ; the steady , first order , _ hamilton - jacobi equation _ ( hj ) is : @xmath3 where , following its _ optimal control interpretation _ , @xmath4 is the _ discount factor _ , @xmath5 is the _ exit cost _ , and the _ hamiltonian _ @xmath6 is defined by : @xmath7 with @xmath8 ( _ dynamics _ ) and @xmath9 ( _ running cost _ ) . the choice of such hamiltonian is not restrictive but useful to simplify the presentation . as extension of the techniques we are going to present , it will be shown , in the dedicate section , as the same results can be obtained in presence of different kind of hamiltonians , as in obstacle problems or in differential games . under classical assumptions on the data ( for our purposes we can suppose @xmath10 and @xmath11 continuous , @xmath12 and @xmath13 lipschitz continuous for all @xmath14 and verified the _ soner s condition _ @xcite ) , it is known ( see also @xcite , @xcite ) that the equation admits a unique continuous solution @xmath15 in the _ viscosity solutions _ sense . the solution @xmath16 is the value function to the infinite horizon problem with exit cost , where @xmath17 is the _ first time of exit _ form @xmath0 : @xmath18 numerical schemes for approximation of such problem have been proposed from the early steps of the theory , let us mention the classical finite differences schemes @xcite , semilagrangian @xcite , discontinuous galerkin @xcite and many others . in this paper we will focus on a _ monotone _ , _ consistent _ and _ stable _ scheme ( class including the first two mentioned above ) , which will provide us the discrete problem where to apply the howard s algorithm . considered a discrete grid @xmath19 with @xmath20 points @xmath21 , @xmath22 on the domain @xmath23 , the finite @xmath20-dimensional approximation of @xmath16 , @xmath24 , will be the solution of the following discrete equation ( @xmath25 ) @xmath26 where @xmath27 , ( maximal diameter of the family of simplices @xmath28 built on @xmath19 ) is the discretization step , and related to a subset of the @xmath29 , there are included the dirichlet conditions following the obvious pattern @xmath30 we will assume on @xmath31 , some hypotheses sufficient to ensure the convergence of the discretization * _ monotony . _ for every choice of two vectors @xmath32 such that , @xmath33 ( component - wise ) then @xmath34 for all @xmath35 . * _ stability . _ if the data of the problem are finite , for every vector @xmath24 , there exists a @xmath36 such that @xmath24 , solution of , is bounded by @xmath37 i.e. @xmath38 independently from @xmath39 . * _ consistency . _ this hypothesis , not necessary in the analysis of the convergence of the scheme , is essential to guarantee that the numerical solution obtained approximates the continuous solution . it is assumed that @xmath40 for every @xmath41 , @xmath42 , with @xmath43 , @xmath44 , and @xmath45 . under these assumptions it has been discussed and proved @xcite that @xmath24 , solution of , converges to @xmath16 , viscosity solution of for @xmath46 . the special form of the hamiltonian @xmath47 gives us a correspondent special structure of the scheme @xmath31 , in particular , with a rearrangement of the terms , the discrete problem can be written as a resolution of a nonlinear system in the following form : @xmath48 where @xmath49 is a @xmath50 matrix and @xmath51 is a @xmath20 vector . the name @xmath51 is chosen to underline ( it will be important in the following ) that such vector there are contained information about the dirichlet conditions imposed on the boundaries . policy iteration algorithm _ ( or howard s algorithm ) consists in a two - steps iteration with an alternating improvement of the policy and the value function , as shown in table [ ha ] . it is by now known @xcite that under a monotonicity assumption on the matrices @xmath52 , ( we recall that a matrix is monotone if ans only if it is invertible and every element of its inverse are non negative ) , automatically derived from ( h1 * ) ( as shown below ) , the above algorithm is a non smooth newton method that converges superlinearly to the discrete solution of problem . the convergence of the algorithm is also discussed in the earlier work @xcite where the results are given in a more regular framework . additionally , if @xmath53 has a finite number of elements , and this is the standard case of a discretized space of the controls , then the algorithm converges in a finite number of iterations . let us state , for a fixed vector @xmath54 the subspace of controls @xmath55 [ p:1 ] let us assume the matrix @xmath52 is invertible . if ( h1 * ) holds true , then @xmath52 is monotone and not null for every @xmath56 with @xmath54 . for a positive vector @xmath24 , consider a vector @xmath57 such that @xmath58 componentwise , then for h1 * @xmath59 where @xmath60 , therefore @xmath61 suppose now that the @xmath62 column of @xmath63 has a negative entry : choosing @xmath64 ( @xmath65 @xmath62 column of the identity matrix ) multiplying the previous relation for @xmath63 we have a contradiction . then @xmath66 is monotone . ( 1,0)380 + howard s algorithm ( ha ) ( 1,0)380 + inputs : @xmath67 , @xmath68 . ( implicitly , the values of @xmath24 at the boundary points ) + initialize @xmath69 and @xmath70 + iterate @xmath71 : * find @xmath72 solution of @xmath73 . + if @xmath74 and @xmath75 , then stop . otherwise go to ( ii ) . . + set @xmath77 and go to ( i ) outputs : @xmath78 . ( 1,0)380 + it is useful to underline the conceptual distinction between the convergence of the algorithm and the convergence of the numerical approximation to the continuous function @xmath16 as discussed previously . in general , the howard s algorithm is an acceleration technique for the calculus of the approximate solution , the error with the analytic solution will be depending on the discretization scheme used . to conclude this introductory section let us make two monodimensional basic examples . [ ex1 ] an example for the matrix @xmath52 and the vector @xmath79 is the easy case of an upwind explicit euler scheme in dimension one @xmath80 where @xmath81 is a uniform discrete grid consisting in @xmath20 knots of distance @xmath39 . moreover , @xmath82 and @xmath83 . in this case the system is @xmath84 } { h\lambda } & -\frac{f^+_1}{h\lambda } & 0 & \cdots & 0 \\ \frac{f^-_2}{h\lambda } & 1+\frac{\left[f^+_2-f^-_2\right]}{h\lambda } & -\frac{f^+_2}{h\lambda } & \cdots & 0\\ 0 & \ddots & \ddots & \ddots & 0\\ 0 & \cdots & \cdots & \frac{f^-_n}{h\lambda } & 1+\frac{\left[f^+_n - f^-_n\right]}{h\lambda } \end{array } \right),\ ] ] and @xmath85 it is straightforward that the solution of howard s algorithm , verifying @xmath86 , is the solution of . [ ex2 ] if we consider the standard 1d semilagrangian scheme , the matrix @xmath52 and the vector @xmath79 are @xmath87 and @xmath88 where @xmath89 and the coefficients @xmath90 are the weights of a chosen interpolation @xmath91(x_i+h f(x_i,\alpha_j))=\sum_{i=0}^{n+1}b_i(\alpha_j)v_i$ ] . despite the good performances of the policy algorithm as a _ speeding up _ technique , in particular in presence of a convenient initialization ( as shown for example in @xcite ) an awkward limit appears naturally : the necessity to store data of very big size . just to give an idea of the dimensions of the data managed it is sufficient consider that for a 3d problem solved on a squared grid of side @xmath92 , for example , it would be necessary to manage a @xmath93 matrix , task which becomes soon infeasible , increasing @xmath92 . this give us an evident motivation to investigate the possibility to solve the problem in parallel , containing the complexity of the sub problems and the memory storage . the strict relation between various points of the domain displayed by equation , makes the problem to find a parallel version of the technique , not an easy task to accomplish . the main problem , in particular , will be about passing information between the threads , necessary without a prior knowledge of the characteristics of the problem . our idea is to combine the policy iteration algorithm with a domain decomposition principle for hj equations . using the theoretical framework of the resolution of partial differential equations on submanifolds , presented for example in @xcite , we consider a decomposition of @xmath0 on a collection of subdomains : @xmath94 where the interfaces @xmath95 , @xmath96 are some strata of dimension lower than @xmath97 defined as the intersection of two subdomains @xmath98 for @xmath99 . the notion of viscosity solution on the manifold , in this regular case , will be coherent with the definition elsewhere a upper semicontinuos function @xmath100 in @xmath101 is a subsolution on @xmath101 if for any @xmath102 , any @xmath103 sufficiently small and any maximum point @xmath104 of @xmath105 is verified @xmath106 where with @xmath107 we indicate the hamiltonian @xmath47 restricted on @xmath108 . + the definition of supersolution is made accordingly . it is useful to underline that , differently from multidomains problems ( like the already quoted @xcite ) there is no need to use a specific concept of solutions through the interfaces . thanks to the regularity of the hamiltonian , the simple definition of viscosity solution on an enlargement of @xmath101 ( called @xmath108 ) will be effective ; as described by the following result . [ t:1 ] let us consider a domain decomposition as stated in . the continuous function @xmath109 , verifying , for a @xmath103 , in the viscosity sense the system below @xmath110 is coincident with the viscosity solution @xmath111 of . it is necessary to prove the uniqueness of a continuous viscosity solution for . after that , just invoking the existence and uniqueness results for the solution @xmath16 ( solution of the original problem ) , and observing that it is also a continuous viscosity solution of the system , from coincidence on the boundary , we get thesis . to prove the uniqueness it is possible to use the classical argument of `` doubling of variables '' . we recall the main steps of the technique for the convenience of the reader . for two continuous viscosity solutions @xmath112 of using the auxiliary function @xmath113 which has a maximum point in @xmath114 , it is easy to see that @xmath115 now the limit @xmath116 is proved as usual deriving @xmath117 and using the properties of sub supersolution , ( for example , @xcite theo . ii.3.1 ) with the observation that no additional difficulty appears when a subsequence @xmath118 is definitely in @xmath101 because of the regularity of the hamiltonian through the same interface ; for the possibility to exchange the role between @xmath119 and @xmath120 ( both super and subsolutions ) we have uniqueness . in the following section we propose a parallel algorithm based on the numerical resolution of the decomposed system above . this technique consists of a two steps iteration : * use howard s algorithm to solve in parallel ( @xmath92 threads ) the nonlinear systems obtained after discretization of on the subdomains @xmath121 ( in this step the values of @xmath24 are fixed on the boundaries ) ; * update the values of @xmath24 on the interfaces of connection @xmath122 by using howard s algorithm on the nonlinear system obtained from the second equation of ( in this case the _ interior points _ of @xmath123 are constant ) . as it is shown later , this two - step iteration permits the transfer of information through the interfaces performed by the phase ( ii ) . this procedure , anyway , is not priceless , the number of the steps necessary for its resolution will be shown to be higher than the classic algorithm ; the advantage will be in the resolution of smaller problems and the possibility of a resolution in parallel . moreover , the coupling between phase ( i ) and ( ii ) produces a succession of results convergent in finite time , in the case of a finite space of controls . the good performances of the algorithm , benefits and weak points will be discussed in details in section [ s : test ] . to describe precisely the algorithm it is necessary to state the following . let us consider as before a uniform grid @xmath124 , the indices set @xmath125 , and a vector of all the controls on the knots @xmath126 . the domain @xmath0 is decomposed as @xmath127 , where , coherently with above @xmath128 ; this decomposition induces an similar structure in the indices set @xmath129 , where every point @xmath130 of index in @xmath131 is an `` _ _ interior point _ _ '' , in the sense that for every @xmath132 ( ball centred in @xmath130 of radius @xmath39 , defined as previously ) , @xmath133 , for every @xmath134 . the set @xmath135 is the set of all the `` _ _ border points _ _ '' , which means , for a @xmath136 we have that there exists at least two points @xmath21 , @xmath137 such that @xmath138 and @xmath139 with @xmath134 . we will build @xmath92 discrete subproblems on the subdomains @xmath121 using as described before a monotone , stable and consistent scheme . in this case a discretization of the hamiltonian provides , for every subdomain @xmath121 , related to points @xmath21 , @xmath140 , a matrix @xmath141 and a vector @xmath142 . we highlighted here , the dependance of @xmath143 from the border points which are , either , points where there are imposed the dirichlet conditions ( data of the problem ) or points on the interface @xmath101 which have to be estimed . assumed for simplicity that every @xmath131 has the same number of @xmath144 elements , called @xmath145 , we have @xmath146 , and @xmath147 , @xmath148 . in resolution over @xmath101 we will have a matrix @xmath149 and a relative vector @xmath150 , in the spaces , respectively , @xmath151 and @xmath152 . ( for the 1d case , e.g. , we can easily verify that @xmath153 ) . in this framework , the numerical problem after the discretization of equations is the following : find @xmath154 with @xmath155 for @xmath156 and @xmath157 , solution of the following system of nonlinear equations : @xmath158 the resolution of first and the second equation of will be called respectively _ parallel part _ and _ iterative part _ of the method . solving the parallel and the iterative part will be performed alternatively , as a double step solver . the iteration of the algorithm will generate a sequence @xmath159 solution of the two steps system @xmath160 where @xmath161 , @xmath162 are the matrices and vectors in @xmath163 , and @xmath164 , containing @xmath165 , @xmath166 and such to return as solution the argument of @xmath167 elsewhere . evidently , @xmath161 @xmath168 with @xmath169 are : equal to @xmath170 in the @xmath171 blocks , and equal to the rows @xmath172 of the identity matrix elsewhere , @xmath173 in the @xmath174 elements of the vector and @xmath175 elsewhere , ( we call these entries , in the following _ identical arguments _ ) ; the same , in the @xmath176 block , @xmath177 elements of the vector for @xmath178 . it is clear that , despite this formal presentation , made to simplify the notation in the following , each equation of , negletting the trivial relations , is a nonlinear system on the same dimension than . clearely , a solution of is the fixed point of . the convergence of the discrete problem above to the solution of equation , for a consistent , monotone and stable scheme was proved by souganidis in @xcite ) , other examples are @xcite . it is consequent then , the domain decomposition result stated before gives the theoretical justification to the transition . different issue will be to show the convergence of the method ; point discussed in the following . ( 1,0)380 + parallel howard s algorithm ( pha ) ( 1,0)380 + inputs : @xmath165 , @xmath179 for @xmath180 + initialize @xmath69 and @xmath181 . + iterate @xmath71 : 1 . _ ( parallel step ) _ for each @xmath182 + call ( ha ) with inputs @xmath183 and @xmath184 + get @xmath185 2 . _ ( sequential step ) _ + call ( ha ) with inputs @xmath186 and @xmath187 + get @xmath188 3 . compose the solution @xmath189 + if @xmath190 then _ exit _ , otherwise go to ( 1 ) . outputs : @xmath78 ( 1,0)380 + it is evident that such technique can be expressed as @xmath191 where , coherently with above @xmath192_j$ ] for @xmath140 . the hypotheses ( h1 * -h2 * ) will be naturally adapted to the new framework as below : * _ monotony . _ for every choice of two vectors @xmath32 such that , @xmath33 ( component - wise ) then @xmath193 for all @xmath194 , and @xmath180 . * _ stability . _ if the data of the problem are finite , for every vector @xmath24 , and every @xmath57 s.t . @xmath195 , there exists a @xmath36 such that @xmath24 , solution of @xmath196 with @xmath197 and @xmath198 , is bounded by @xmath37 independently from @xmath39 . this will be sufficient , thanks also to h3 , to ensure convergence of @xmath199 solution of @xmath200 to @xmath201 for @xmath43 . from the assumptions on the discretization scheme some specific properties of @xmath161 and @xmath168 can be derived [ pp ] let us assume @xmath202 . let state also * if @xmath203 then @xmath204 , for all @xmath180 , for all @xmath14 . then it holds true the following . 1 . if invertible , the matrices @xmath205 are monotone , not null for every @xmath198 , and for every @xmath206 with @xmath207 . if @xmath208 , we have that for all @xmath198 and for every @xmath209 , there exists a @xmath210 such that @xmath211 the same relation holds for @xmath212 . 3 . called @xmath213 the fixed point of , if we have @xmath214 ( resp . @xmath215 ) , then there exists a @xmath14 such that , for all @xmath216 , @xmath217 to prove @xmath218 let us just observing that the monotony of @xmath165 is sufficient end necessary for the monotony of @xmath161 , ( elsewhere @xmath161 is a diagonal block matrix with all the other blocks invertible ) , then the argument is the same of proposition [ p:1 ] , starting from two vectors @xmath219 with the only difference that we need assumption h4 to get @xmath220 or equivalently @xmath221 then the thesis . to prove 2 , it is sufficient to see @xmath222 , then for h2 the thesis . the proof of 3 is a direct consequence of monotony assumption h1 with the definition of @xmath213 as @xmath223 here we introduce a convergence result for the ( pha ) algorithm . [ t:1 ] assume that the function @xmath224 , with @xmath225 invertible , and @xmath226 are continuous on the variable @xmath227 for @xmath216 , @xmath53 is a compact set of @xmath1 , and @xmath228 hold . then there exists a unique @xmath213 in @xmath164 solution of . moreover , the sequence @xmath229 generated by the ( pha ) has the following properties : * every element of the sequence @xmath230 is bounded by a constant @xmath37 , i.e. @xmath231 . * if @xmath232 then @xmath233 for all @xmath71 , vice versa , if @xmath234 then @xmath235 . * @xmath236 when @xmath237 tends to @xmath238 . the existence of a solution comes directly from the monotonicity of the matrices @xmath239 , the existence of an inverse and then the existence of a solution of every system of . let us show that such solution is limited as limit of a sequence of vectors of bounded norm . observing that , @xmath240 without loss of generality we assume that @xmath241 . considering the problem @xmath242 we have for h2 that if @xmath243 is bounded then @xmath244 . adding that @xmath245 is chosen bounded , the thesis follows for induction . let us to pass now to prove the uniqueness : taken @xmath246 two solutions of , we define the vector @xmath247 equal to @xmath24 in the identical arguments of @xmath167 and equal to @xmath57 elsewhere , for a @xmath198 . we have that , for a control @xmath248 ( for proposition [ pp].3 ) , @xmath249 then @xmath250 and for monotonicity @xmath33 . exchanging the role of @xmath24 and @xmath57 , and for the arbitrary choice of @xmath251 ( in some arguments the relation above is trivial ) we get the thesis . \(i ) to prove that @xmath72 is an increasing sequence is sufficient to prove that taken @xmath252 solution of @xmath253 with ( the opposite case is analogue ) @xmath254 , for a choice of @xmath198 is such that @xmath255 . let us observe , for a choice of @xmath256 and using of prop . [ pp ] @xmath257 then @xmath258 then @xmath255 . + we need also to prove that @xmath259 : if it should not be true , then , with a similar argument than above @xmath260 then for h4 , @xmath261 which contradicts what stated previously . it is also possible to show that the method stops to the fixed point in a finite time . this is an excellent feature of the technique ; unfortunately , the estimate which is possible to guarantee is largely for excess and , although important from the theoretical point of view , not so effective to show the good qualities of the method . the performances will checked in the through some tests in the section [ s : test ] . if @xmath262 and convergence requests of theorem [ t:1 ] are verified , then @xmath263 converges to the solution in less than @xmath264 iterative steps . the proof is slightly similar to the classic howard s case ( cf . for example @xcite ) . let us consider the abstract formulation @xmath265 , where @xmath266 is determined by @xmath267 parameter in @xmath268 , and @xmath269 , where @xmath270 is determined by @xmath271 parameter in @xmath268 . then if we consider the iteration @xmath272 and we suppose ( theorem [ t:1 ] ) @xmath273 , @xmath274 ; than called @xmath275 the @xmath276 variables in @xmath268 associated to @xmath277 we know that there exist a @xmath144 and a @xmath278 where @xmath279 , such that @xmath280 , and again @xmath281 . afterwards @xmath277 is a fixed point of . + to restrict to our case is sufficient identify the process @xmath282 with the ( parallel ) resolution on the sub - domains and @xmath283 with the iteration on the interfaces between the sub domains . it is worth to notice that the above estimation is worse than the classical howard s case . in fact , the classical algorithm find the solution in @xmath264 , the @xmath263 will have the same number of iterative steps . this number has to be multiplied , called @xmath284 the maximum number of nodes in a sub - domain and @xmath285 the number of nodes belonging to the interface , for @xmath286 getting , at the end , a total number of simple steps equal to @xmath287 , much more than the classical case . in this analysis we do not consider anyway , the good point of the decomposition technique , the fact that any computational step is referred to a smaller and simpler problem , with the evident advantages in term of time elapsed in every thread and memory storage needed . the performances of the algorithm and its characteristics as speeding up technique will be tested in this section . we will use a standard academic example where , anyway , there are present all the main characteristics of our technique . [ [ d - problem ] ] 1d problem + + + + + + + + + + consider the monodimensional problem @xmath288 it is well known that this equation ( _ eikonal equation _ ) modelize the distance from the boundary of the domain , scaled by an exponential factor ( _ kruzkov transform _ , cf . @xcite ) . through a standard euler discretization is obtained the problem in the form . in table [ tt:2 ] is shown a comparison , in term of speed and efficacy , of our algorithm and the classic howard s one , in the case of a two thread resolution . it is possible appreciate as the parallel technique is not convenient in all the situations . this is due to the low number of parallel threads which are not sufficient to justify the construction . in the successive test , keeping fixed the parameter @xmath289 and tuning number of threads it is possible to notice how much influential is such variable in terms of efficacy and time necessary for the resolution . c\|2c\|4c & & + dx & time & it . & t. ( par . p. ) & it . p. ) & t. ( it . p. ) & total t. + * 0.1 & e-3 & 10 & 1e-4 & 4 & 1e-5 & 1e-3 + * 0.05 & 6e-3 & 20 & 8e-4 & 5 & e-5 & 3e-3 + * 0.025 & 0.09 & 40 & 7e-3 & 6 & 2e-5 & 0.04 + * 0.0125 & 0.32 & 80 & 0.048 & 8 & 1e-4 & 0.36 + * 0.00625 & 2.22 & 160 & 0.34 & 14 & 8e-4 & 3.26 + * * * * * c\|2c\|4c dx=0.0125 & & + threads & t. & it . & t. ( par . p. ) & it . ( par . ) & t. ( it . p. ) & total t. + * 2 & & & 0.48 & 4 & 1e-4 & 0.36 + * 4 & & & 8e-3 & 6 & 1e-4 & 0.086 + * 8 & 0.32 & 80 & 18e-4 & 7 & 6e-4 & 0.014 + * 16 & & & 7e-4 & 10 & 4e-4 & + * 32 & & & 2e-4 & 8 & 6e-3 & 0.011 + * * * * * in table [ tt:2 ] we compare the iterations and the time ( expressed in seconds as elsewhere in the paper ) necessary to reach the approximated solution , analysing in the various phases of the algorithm , time and iterations necessary to solve every sub - problem ( first two columns ) , time elapsed for the iterative part ( which passes the information through the threads , next column ) , finally the total time . it is highlighted the optimal choice of number of threads ( 16 thread ) ; it is evident as that number will change with the change of the discretization step @xmath289 . therefore it is useful to remark that an additional work will be necessary to tune the number of threads accordingly to the peculiarities of the problem ; otherwise the risk is to is to loose completely the gain obtained through parallel computing and to get worse performances even compared with the classical howard s algorithm . as in the rest of the paper all the codes are developed in mathworks matlaband performed on a processor 2,8 ghz intel core i7 ; in the tests the parallelization is simulated . c\|2c\|5c & & + dx & t. & it . & t. ( p.p . ) & it . & t. ( it.p . ) & it . ( it.p . ) & total t. + * 0.1 & 0.05 & 11 & 0.009 & 8 & 0.02 & 2 & 0.04 + * 0.05 & 2.41 & 21 & 0.05 & 13 & 0.03 & 2 & 0.14 + * 0.025 & 73.3 & 40 & 2.5 & 22 & 0.15 & 3 & 7.83 + * 0.0125 & @xmath290e5 & - & 76 & 40 & 1.293 & 5 & 383.3 + * * * * [ [ d - problem-1 ] ] 2d problem + + + + + + + + + + the next test is in a space of higher dimension . let us consider the approximation of the scaled distance function from the boundary of the square @xmath291 , solution of the eikonal equation @xmath292 where @xmath293 is the usual unit ball . for the discretization of the problem is used a standard euler discretization . similar tests than the 1d case are performed , confirming the good features of our technique and , as already shown , the necessity of an appropriate number of threads with respect to the complexity of the resolution . -norm ( left ) and distribution of the error @xmath294 , @xmath295 threads ( right ) . , title="fig:",height=170 ] -norm ( left ) and distribution of the error @xmath294 , @xmath295 threads ( right ) . , title="fig:",height=170 ] in table [ tt:3 ] performances of the classic howard s algorithm are compared with our technique . in this case the number of threads are fixed to 4 ; the parallel technique is evaluated in terms of : maximum time elapsed in one thread and max number of iterations necessary ( first and second columns ) , time and number of iterations of the iterative part ( third and fourth columns ) and total time . in both the cases the control set @xmath296 is substituted by a @xmath297points discrete version . it is evident , in the comparison , an improvement of the speed of the algorithm even larger than the simpler 1d case . this justifies , more than the 1d case , our proposal . c\|2c\|4c dx=0.025 & & + threads & t. & it . & t. ( par . p. ) & it . ( par . ) & t. ( it . p. ) & total t. + * 4 & & & 2.5 & 22 & 0.15 & 7.83 + * 9 & & & 0.9 & 18 & 0.5 & 5.08 + * 16 & 73.3 & 40 & 0.05 & 13 & 1.6 & + * 25 & & & 0.03 & 12 & 2.4 & 2.52 + * 36 & & & 0.016 & 11 & 6.04 & 6.11 + * * * * * in the table [ tt:4 ] are compared the performances for various choices of the number of threads , for a fixed @xmath298 . as in the 1d case is possible to see how an optimal choice of the number of threads can drastically strike down the time of convergence . in figure [ f : in ] is possible to see the distribution of the error . as is predictable , the highest concentration will correspond to the non - smooth points of the solution . it is possible to notice also how our technique apparently does not introduce any additional error in correspondence of the interfaces connecting the sub - domains . this is reasonable , although not evident theoretically . in fact , it is possible to prove the convergence of the scheme to the solution of using classical techniques @xcite but the rate of convergence could be different in the various subproblems , because of the ( possibly different ) local features of the problem . as shown in the tests , an important point of weakness of our technique is represented by the iterative part , which can be smaller and therefore easier than the ones solved in the parallel part , but it is highly influential in terms of general performances of the algorithm . in particular the number of the iterations of the coupling iterative - parallel part is sensible to a good initialization of the `` internal boundary '' points . as is shown in figure [ f : in ] a right initialization , even obtained on a very coarse grid , affects consistently the overall performances . in this section , all the tests are made with a initialization of the solution on a @xmath299-points grid , with @xmath97 dimension of the domain space . the time necessary to compute the initial solution is always negligeable with respect to the global procedure . ( left ) @xmath300 ( right ) ) of the approximated solution obtained with a @xmath301 and an @xmath302threads pha . , title="fig:",height=170 ] ( left ) @xmath300 ( right ) ) of the approximated solution obtained with a @xmath301 and an @xmath302threads pha . , title="fig:",height=170 ] c\|2c\|5c & & + dx & time & it . & t. ( p. p. ) & it . & t. ( it . p. ) & it . p. ) & total t. + * 0.4 & 0.004 & 4 & 0.003 & 4 & 0.002 & 1 & 0.05 + * 0.2 & 0.22 & 6 & 0.026 & 6 & 0.016 & 2 & 0.052 + * 0.1 & 164.2 & 11 & 1.102 & 8 & 2.1 & 4 & 6.78 + * 0.05 & @xmath290e5 & - & 164 & 10 & 4.98 & 3 & 494 + * * * * [ [ d - problem-2 ] ] 3d problem + + + + + + + + + + analogue results are obtained also in the approximation of a 3d problem . of course the effects of the increasing number of control points produces a greater complexity and will limit , for a same number of processors available , the possibility of a fine discretization of the domain . + let us consider the domain @xmath303 ^ 3 $ ] and the equation , where @xmath304 , unitary ball in @xmath305 . in figure [ fig3d ] there are shown two level sets of the solution obtained . a comparison with the performances of the classic howard s algorithm are shown in table [ tt:4 ] . [ multiloop ] with the growth of the dimensionality of the problem a special care should be dedicated to the resolution of the iterative step . suppose to simplify the procedure considering a square domain ( in dimension @xmath306 an interval , a square , a cube .. ) and a successive splitting in equal regular subdomains . calling @xmath20 the number of total variables and @xmath307 the number of the splitting ( which generates a division in @xmath308 subdomains ) the number of the elements in every thread of the parallel part is @xmath309 , and the number of the variables in the iterative part @xmath310{n}}(n_s-1)d$ ] . clearly the optimal choice of the number of threads is such that the elements of the iterative part are balanced with the nodes in each subdomain , so it is straight forward to find the following optimal relation between number of splitting and total elements @xmath311 it is evident that for a very high number of elements , ( figure [ fig4 ] ) , it is useless to use a great and non optimal number of threads . this contradiction comes from the bottleneck effect of the resolution on the interfaces of communication between the subdomains , indeed the complexity of such subproblem will grow with the number of threads instead to decrease , reducing our possibilities of resolution . the problem can be overcome with an additional parallel decomposition of the iterative pass , permitting us to decompose each subproblem to a complexity acceptable . imagine to be able to solve ( for computational reasons , memory storage , etc . ) only problem of dimension `` white square '' ( we refer to figure [ fig4 ] , right ) and to want to solve a bigger problem ( `` square 1 '' ) with an arbitrary number of processors available . through our technique we will decompose the problem in a finite number of subproblems `` white square '' and a ( possibly bigger than the others ) problem `` square 2 '' . we will replicate our parallel procedure for the `` square 2 '' obtaining a collection of manageable problems and a `` square 3 '' . through a reiteration of this idea we arrive to a decomposition in subproblems of dimension desired . in this section there are shown some non trivial extensions to more general situations of the method . we will discuss , in particular , how to adapt the parallelization procedure to the case of a target problem , an obstacle problem and max - min problems , where the special structure of the hamiltonian requires some cautions and remarks . an important class of problems where is useful to extend the techniques discussed is the target problems where a trajectory is driven to arrive in a _ target set _ @xmath312 optimizing a cost functional . a easy way to modify our algorithm to this case is to change the construction procedure for @xmath49 and @xmath37 : @xmath313_i:=\left\{\begin{array}{ll } \left[b(\alpha)\right]_i , & \hbox { if } x_i\notin { { \mathcal t } } , \\ \left[\mathbb{i}\right]_i , & \hbox { otherwise;}\end{array } \right . \ ; c'(\alpha)_i:=\left\{\begin{array}{ll } c(\alpha)_i , & \hbox { if } x_i\notin { { \mathcal t } } , \\ 0 , & \hbox { otherwise;}\end{array } \right.\ ] ] this , with the classical further construction of _ ghost nodes _ outside the domain @xmath0 to avoid the exit of the trajectories from @xmath0 , will solve this case . a question arises naturally in this modification : are the convergence results still valid ? the answer is not completely trivial because , for example , a monotone matrix modified as above is not automatically monotone ( the easiest counterexample is the identical matrix flipped vertically : it is monotone because invertible and equal to its inverse , but changing any row as in we get a non invertible matrix ) . to prove the convergence it is sufficient to start from the numerical scheme associated to such modified algorithm . it is quite direct to show verified the hypotheses ( h1-h4 ) getting as consequence the described properties of the algorithm . a well known benchmark in the field is the so - called zermelo s navigation problem , the main feature , in this case , is that the dynamic is driven by a force of comparable power with respect to our control . the target to reach will be a ball of radius equal to @xmath314 centred in the origin , the control is in @xmath315 . the other data are : @xmath316 ^ 2,\quad \lambda = 1 , \quad l(x , y , a)=1.\ ] ] .,title="fig:",height=170 ] .,title="fig:",height=170 ] in table [ tt:6 ] a comparison with the number of threads chosen is made . now we are in presence of characteristics not aligned with the grid , but the performances of the method are poorly effected . convergence is archived with performances comparable to the already described case of the eikonal equation . c\|2c\|4c dx=0.025 & & + threads & t. & it . & t. ( par . p. ) & it . ( par . ) & t. ( it . p. ) & total t. + * 4 & & & 1.31 & 11 & 0.13 & 5.4 + * 9 & & & 0.7 & 9 & 0.7 & 4.2 + * 16 & 37.9 & 20 & 0.031 & 7 & 1.38 & + * 25 & & & 0.02 & 7 & 2.7 & 3.9 + * 36 & & & 0.01 & 8 & 5.19 & 5.28 + * * * * * dealing with an optimal problem with constraints using the bellman s approach , various techniques have been proposed . in this section we will consider an implicit representation of the constraints through a level - set function . let us to consider the general single obstacle problem @xmath317 where the hamiltonian @xmath47 is of the form discussed in section [ s:1 ] and the standard hypothesis about regularity of the terms involved are verified . the distinctive trait of this formulation is about the term @xmath318 , assumed regular , typically stated as the opposite of the signed distance from the boudary of a subset @xmath319 . the solution of this problem is coincident , where defined , with the solution of the same problem in the space @xmath320 , explaining the name of `` obstacle problem '' ( cf . @xcite ) . through an approximation of the problem in a finite dimensional one , in a similar way as already explained , is found the following variation of the howard s problem @xmath321 where the term @xmath57 is a sampling of the function @xmath322 on the knot of the discretization grid . it is direct to show that changing the definition of the matrix @xmath49 and @xmath323 , is possible to come back to the problem . adding an auxiliary control to the set @xmath324 and re - defying the matrices @xmath49 and @xmath323 as @xmath325_i:=\left\{\begin{array}{ll } \left[b(\alpha)\right]_i , & \hbox { if } b(\alpha)v - c_g(\alpha)\geq v - w \\ \left[\mathbb{i}\right]_i , & \hbox { otherwise;}\end{array } \right . \\ c_g'(\alpha)_i:=\left\{\begin{array}{ll } c_g(\alpha)_i , & \hbox { if } b(\alpha)v - c_g(\alpha)\geq v - w \\ w_i , & \hbox { otherwise;}\end{array } \right . \end{split}\ ] ] @xmath326 ( where the @xmath327 is the @xmath328row if @xmath329 is a matrix , and the @xmath328 element if @xmath329 is a vector , and @xmath330 is the identity matrix ) , the problem becomes @xmath331 which is in the form . even in this case the verification of hypotheses ( h1-h4 ) by the numerical scheme associated to the transformation is sufficiently easy . it is in some cases also possible the direct verification of conditions of convergence in the obstacle problem deriving them from the free of constraints case . for example if we have that the matrix @xmath52 is strictly dominant ( i.e. @xmath332 for every @xmath333 , and there exists a @xmath103 such that for every @xmath251 , @xmath334 ) , then the properties of the terms are automatically verified , ( i.e. since all @xmath225 are strictly dominant and thus monotone ) . a classical problem of interest is the optimization of trajectories modelled by @xmath335 which produces a collection of curves in the plane @xmath336 with a constraint in the curvature of the path . typically this is a simplified model of a car of constant velocity @xmath323 with a control in the steering wheel . + the value function of the exit problem from the domain @xmath337 , @xmath338 $ ] discretized uniformly in 8 points is presented in figure [ f : dub ] . it is natural to imagine the same problem with the presence of constraints . such problem can be handled with the technique described above producing the results shown in the same figure [ f : dub ] , where there are presented some optimal trajectories ( in the space @xmath336 ) for the exit from @xmath337 in presence of some constraints . from the picture it is possible to notice also the constraint about the minimal radius of curvature contained in the dynamics . the last , more complicated extension of the howard s problem is about max - min problems of the form ( 1,0)380 + pha ( maxmin case ) ( 1,0)380 initialize @xmath69 @xmath181 for all @xmath198 . + k:=1 ; 1 . iterate _ ( parallel step ) _ for every @xmath182 do : + @xmath339 * find @xmath340 solution of @xmath341 . + if @xmath342 and @xmath343 , then @xmath344 , and exit ( from inner loop ) . + otherwise go to ( 1.ii ) . * @xmath345 . + set @xmath346 and go to ( 1.i ) 2 . ( sequential step ) _ for @xmath347 * find @xmath348 solution of @xmath349 . + if @xmath350 and @xmath351 , then @xmath352 , and go to ( 3 ) . + otherwise go to ( 2ii ) . * @xmath353 . + set @xmath354 and go to ( 2i ) 3 . compose the solution @xmath355 + k:=k+1 ; + if @xmath356 then _ exit _ , otherwise go to ( 1 ) . ( 1,0)380 @xmath357 such a non linear equations arises in various contexts , for example in differential games and in robust control . the convergence of a parallel algorithm for the resolution of such problem is also discussed in @xcite . also in this case , a modified version of the policy iteration algorithm can be shown to be convergent ( cf . our aim in this subsection is to give some hints to build a parallel version of such procedure . let us introduce the function @xmath358 , for @xmath359 and @xmath360 defined by @xmath361 the problem , in analogy with the previous case , is equivalent to solve the following system of nonlinear equations @xmath362 the parallel version of the howard algorithm in the case of a maxmin problem is summarized in table [ mm ] . it is worth to notice that at every call of the function @xmath363 is necessary to solve a minimization problem over the set @xmath53 , this can be performed in an approximated way , using , for instance , the classical howard s algorithm . this gives to the dimension of this set a big relevance on the performances of our technique . for this reason , if the cardinality of @xmath53 ( in the case of finite sets ) is bigger than @xmath364 , it is worth to pass to the alternative problem @xmath365 ( here there are used the isaacs conditions ) before the resolution , inverting in this way , the role of @xmath53 and @xmath364 in the resolution . one of the most known example of max - min problem is the pursuit evasion game ; where two agents have the opposite goal to reduce / postpone the time of capture . the simplest situation is related to a dynamic @xmath366 where controls are taken in the unit ball @xmath367 and capture happens when the trajectory is driven to touch the small ball @xmath368 , ( @xmath369 , in this case ) . the passage to a target problem is managed as described previously . , title="fig:",height=170 ] , title="fig:",height=170 ] in figure [ f : pe ] the approximated value function of that problem is shown . the main difficulty in the use of the howard s algorithm , i.e. the resolution of big linear systems can be overcome using parallel computing . this is important despite the fact that we must accept an important drawback : the double loop procedure ( or multi - loop procedure as sketched in remark [ multiloop ] ) does not permit to archive a superlinear convergence , as in the classical case ; we suspect ( as in figure [ f : in ] ) that such rate is preserved looking to the ( external ) iterative step , where we have to consider , anyway , that in every step of the algorithm a resolution of a reduced problem is needed . another point influential in the technique is the manner chosen to solve every linear problem which appears in the algorithm . in this paper , being not in our intentions to show a comparison with other competitor methods rather studying the properties of the algorithm in relation of the classical case , we preferred the simplicity , using a routine based on the exact inversion of the matrix . using of an iterative solver , with the due caution about the error introduced , better performances are expected ( cf . @xcite ) . through the paper we showed as some basic properties of the schemes used to discretized the problem bear to sufficient conditions for the convergence of the algorithm proposed , this choice was made to try to keep our analysis as general as possible . a special treatment about the possibility of a domain decomposition in presence of non monotone schemes is possible , although not investigated here . this work was supported by the european union under the 7th framework programme fp7-people-2010-itn sadco , sensitivity analysis for deterministic controller design . + the author thanks hasnaa zidani of the uma laboratory of ensta for the discussions and the support in developing the subject . 00 , _ an efficient policy iteration algorithm for dynamic programming equations _ , pamm 13 n.1 ( 2013 ) 467468 . , optimal control and viscosity solution of hamilton - jacobi - bellman equations . birkhauser , boston heidelberg , 1997 . , _ a bellman approach for two - domains optimal control problems in @xmath164 _ , esaim contr va . , 19 n. 3 ( 2013 ) 710739 . , _ a bellman approach for regional optimal control problems in @xmath370 _ , siam j. cont . , 52 no . 3 ( 2014 ) 17121744 . , stochastic and differential games : theory and numerical methods , birkhuser , boston , 1999 . , _ flow invariance on stratified domains _ , set - valued var . , 21 ( 2013 ) 377403 . , dynamic programming , princeton university press , princeton , nj , 1957 . , _ some convergence results for howard s algorithm _ , siam j. numer . , 47 n. 4 ( 2009 ) 30013026 . , _ a domain decomposition method for bellman equations _ , cont . , 180 ( 1994 ) 477483 . , _ systems of convex hamilton - jacobi equations with implicit obstacles and the obstacle problem _ , comm . , 8 ( 2009 ) 12911302 . , _ a discontinuous galerkin finite element method for directly solving the hamilton - jacobi equations _ , j. comput . phys . , 223 n. 1 ( 2007 ) 398415 . , two approximations of solutions of hamilton - jacobi equations , math . comp . , 43 n. 167 ( 1984 ) 119 . , partial differential equations : graduate studies in mathematics . american mathematical society 2 , 1998 . , _ semi - lagrangian approximation schemes for linear and hamilton - jacobi equations _ , applied mathematics series , siam , 2013 . , _ advances in parallel algorithms for the isaacs equation _ , in advances in dynamic games . birkhuser boston , 2005 . 515 - 544 . , dynamic programming and markov processes , the mit press , cambridge , ma , 1960 . , _ on the convergence of policy iteration in stationary dynamic programming _ , math . res . , 4 no.1 ( 1979 ) 60 - 69 . , _ convergence analysis of some algorithms for solving nonsmooth equations _ , res . , 18 ( 1993 ) 227244 . , _ a nonsmooth version of newton s method _ , math . , 58 ( 1993 ) 353367 . , _ hamilton - jacobi - bellman equations on multi - domains _ , in : _ control and optimization with pde constraints _ , birkhauser basel , 164 ( 2013 ) 93116 . , _ convergence properties of policy iteration _ , siam j. contr . opt . , 42 n. 6 ( 2004 ) 2094 - 2115 . , _ optimal control problems with state - space constraints _ , siam j. contr . opt . , 24 ( 1986 ) 552562 . , _ approximation schemes for viscosity solutions of hamilton - jacobi equations _ , j. differ . equations 59 n. 1 ( 1985 ) 143 . , _ domain decomposition algorithms for solving hamilton jacobi - bellman equations _ , num . analysis opt . , 14 ( 1993 ) 145166 . , _ on the convergence of policy iteration in stationary dynamic programming _ , math . res . , 4 n. 1 ( 1979 ) 6069 . , _ a new domain decomposition method for an hjb equation _ , j. comput . appl . math . , 159 n. 1 ( 2003 ) 195204 .	the classic howard s algorithm , a technique of resolution for discrete hamilton - jacobi equations , is of large use in applications for its high efficiency and good performances . a special beneficial characteristic of the method is the superlinear convergence which , in presence of a finite number of controls , is reached in finite time . performances of the method can be significantly improved by using parallel computing ; how to build a parallel version of method is not a trivial point , the difficulties come from the strict relation between various values of the solution , even related to distant points of the domain . in this contribution we propose a parallel version of the howard s algorithm driven by an idea of domain decomposition . this permits to derive some important properties and to prove the convergence under quite standard assumptions . the good features of the algorithm will be shown through some tests and examples . * keywords : * howard s algorithm ( policy iterations ) , parallel computing , domain decomposition + * 2000 msc : * 49m15 , 65y05 , 65n55
since dorothy denning s seminal 1987 paper on intrusion detection @xcite , ml and data mining(dm ) have steadily gained attention in security applications . darpa s 1998 network intrusion detection evaluation @xcite , and kdd(conference on knowledge discovery and data mining ) cup s 1999 challenge @xcite have raised profile of ml in security contexts . yet , constrained by hardware and system resources@xcite , large - scale ml applications did not receive much attention for many years . in 2008 , acm conference on computer and communications security(ccs ) hosted the 1st artificial intelligence in security(aisec ) workshop , which has since been a dedicated venue at a top - level security conference for the intersection of ml and security . from 2008 , the pace of research and publicity of ml in security started to accelerate in academic communities ( section 2.3 ) , and industry venues ( e.g. black hat , rsa ) also shifted interests . for instance , ml in security was still a topic of minority interest at black hat usa 2014 in august @xcite , but at rsa 2016 in february , the majority of vendors claimed to deploy ml in their products @xcite . a part of this shift may be motivated by the sudden increase in blackswan events like the discovery of crime , beast and heartbleed vulnerabilities . the discovery of these vulnerabilities suggest that organizations may be attacked via previously unknown classes of attacks . to defend against these types of attacks requires monitoring not just for known vectors attacks , but also for behavior suggestive of a compromised machine . the latter requires the gathering and analysis of much larger sets of data . advances in hardware and data processing capacities enabled large - scale systems . with increasing amount of data from growing numbers of information channels and devices , the analytic tools and intelligent behaviors provided by ml becomes increasingly important in security . with darpa s cyber grand challenge final contest looming @xcite , research interest in ml and security is becoming even more conspicuous . now is the crucial time to examine research works done in ml applications and security . to do so , we studied the state - of - art of ml research in security between 2008 and early 2016 , and systematize this research area in 3 ways : 1 . we survey cutting - edge research on applied ml in security , and provide a high - level overview taxonomy of ml paradigms and security domains . 2 . we point to research challenges that will improve , enhance , and expand our understanding , designs , and efficacy of applying ml in security . we emphasize a position which treats security as a game theory problem . while we realize there are different ways to classify existing security problems based on purpose , mechanism , targeted assets , and point of flow of the attack , our sok s section structure is based on the `` security and privacy '' category of 2012 acm computing classification system@xcite , which is a combination of specific use cases(e.g . malware , phishing ) , technique ( e.g. information flow ) , and targeted assets(e.g . web application , proxies ) . we present the state - of - art ml applications in security as the following : section 3 and table 2 & 3 discusses network security , section 4 and table 4 surveys security services , section 5 and table 5 specifies advances in software & applications security , section 6 and table 6 & 7 lays out taxonomy for system security , and section 7 and table 8 , 9 & 10 summarizes progress since 2008 in malware detection , ids , and social engineering . throughout the survey , we share our frameworks for ml system designs , assumptions , and algorithm deployments in security . we focus our survey on security _ applications _ and security - related ml and ai problems on the _ defense _ side , hence our scope excludes theories related to security such as differential privacy and privacy - preservation in ml algorithms@xcite , and excludes ml applications in side channel attacks such as @xcite . partly because there is already a 2013 sok on evolution of sybil defense@xcite in online social networks(osn ) , and partly because we would like to leave it as a small exercise to our readers , we excluded sybil defense schemes in osn as well@xcite . still with a broad base , we propose an alternative position to frame security issues , and we also recommend a taxonomy for ml applications in security use cases . yet , we do not conclude with a terminal list of `` right '' or `` correct '' approaches or methods . we believe that the range of the applications is too wide to fit into one singular use case or analysis framework . instead , we intend this paper as a systematic design and method overview of thinking about researching and developing ml algorithms and applications , that will guide researchers in their problem domains on an individual basis . we target our work to security researchers and practitioners , so we assume that our readers have general knowledge for key security domains and awareness of common ml algorithms , and we also define terms when needed . theory + + & 3.1/4/5/6.1/6.2/7.1/7.2/7.3&5/6.1/7.2&3.1/3.2/5/6.1/6.2/7.1/7.2&5/6.1 & + & 3.1/3.2/4/5/6.1/6.2/7.1/7.2/7.3&7.1/7.3&3.1/5&&6.2 + & & & 7.1&&6.2/7.2 + + & 6.1/7.2&7.2&6.2 & & + & 3.1/3.2/5/7.2/7.3&5/7.3&3.1/3.2/6.1/7.1&&6.2 + & 5&&5&5/6.1 & + & 3.1/4/7.2/7.3&7.2&4/7.2&&7.2 + & 6.1/6.2/7.1&6.1/7.1&6.2/7.1 & & + + & 4/5/6.1/7.2&&4/6.2/7.2 & & + & 3.1/6.2/7.3&7.2&3.1&&6.2/7.2 + & 3.2/5/6.1/6.2/7.1/7.2/7.3&5/6.1/7.1/7.2/7.3&3.2/5/6.1/7.1/7.2&5/6.1 & + we agree with assessment of top conferences in @xcite . we systematically went through all proceedings between 2008 and early 2016 of the top 6 network- and computer - security conferences to collect relevant papers . because of kdd s early and consistent publication record on ml applications in security , and its status as a top - level venue for ml and dm applications , we also include kdd s 2008 - 2015 proceedings . to demonstrate the wide - ranging research attention drawn to ml applications in security , we also added chosen selections from the workshop aisec , international conference on machine learning(icml ) , neural information processing systems(nips ) , and internet measurement conference(imc ) papers between 2008 - 2015 , mostly in the `` future development '' section . figure 1 shows the generalization of ml system designs when applied in security , that emerged from our survey of the papers(the legend is on the figure s bottom left ) . in different use cases , the system components may embody different names , but their functionalities and positions are captured in the figure . for example : 1 . * _ knowledge base _ * is baseline of known normality and/or abnormality , depending on use cases , they include but are not limited to blacklist(bl ) , whitelist(wl ) , watchlist ; known malware signatures , system traces , and their families ; initial set of malicious web pages ; existing security policies or rules , etc .. 2 . * _ data sources _ * are where relevant data is collected . they can be either off - line or live online data feed , e.g. malware traces collected after execution(off - line ) , url stream(online ) . * _ training data _ * are labeled data which are fed to classifiers in training . they can be standard research datasets , new data(mostly from industry ) labeled by human , synthetic datasets , or a mix . _ pre - processor and feature extractor _ * construct features from data sources , for example : url aggregators , graph representations , smtp header extractions , n - gram model builders . dynamic analyzer and static analyzer are used most often in malware - related ml tasks , and human feedback loop is added when the system s design intends to be semi - supervised or human - in - the - loop(hitl ) . theory + + & 58(49%)&7(5.9%)&24(20%)&2(1.7%)&0(0% ) + & 18(15%)&4(3.4%)&3(2.5%)&0(0%)&1(0.85% ) + & 0(0%)&0(0%)&0(0%)&0(0%)&2(1.7% ) + + & 4(3.4%)&1(0.85%)&1(0.85%)&0(0%)&0(0% ) + & 17(14.4%)&4(3.4%)&11(9.3%)&0(0%)&1(0.85% ) + & 4(3.4%)&0(0%)&1(0.85%)&2(1.7%)&0(0% ) + & 31(26%)&2(1.7%)&9(7.6%)&0(0%)&2(1.7% ) + & 20(17%)&4(3.4%)&5(4.2%)&0(0%)&0(0% ) + + & 16(13.6%)&0(0%)&7(6%)&0(0%)&0(0% ) + & 9(7.6%)&1(0.85%)&5(4.2%)&0(0%)&3(2.5% ) + & 51(43.2%)&10(8.5%)&15(12.7%)&2(1.7%)&0(0% ) + table 1 shows a matrix with rows indicating different ways of classifying the security problems , and the columns showing well - understood ml paradigms . based on the threat models and modeling purposes presented in the papers , we qualitatively group the attacker into three groups . if there are multiple attacker types in one section , the section s numbering appears multiple times accordingly . 1 . * _ passive _ * attackers make no attempt to evade detections ; their behaviors fit into descriptions of the threat models . _ semi - aggressive _ * attackers have knowledge of the detectors , and only attempt to evade detections . * _ active _ * attackers do not only have knowledge of the detectors and attempt to evade detections , but also actively try to poison , mislead , or thwart detection . _ knowledge _ * of attackers , is the information in at least one of the five aspects : the learning algorithms themselves , the algorithms feature spaces , the algorithm s parameters , training and evaluation data - regardless of being labeled or not - used by the algorithms , and decision feedback given by the algorithms @xcite . influenced by @xcite , we extend their definitions , and qualitatively categorize attackers primary purpose as to compromise _ confidentiality , availability _ or _ integrity _ of legitimate systems , services , and users . 1 . attacks on * _ confidentiality _ * compromise the confidential or secret information of systems , services , or users ( e.g. password crackers ) . 2 . attacks on * _ availability _ * make systems and services unusable with unwanted information , requests , or many errors in defense schemes ( e.g. ddos , spam ) . 3 . attacks on * _ integrity _ * masquerade maliciously intentions as benign intentions in systems , services , and users ( e.g. malware ) . we also define ml paradigms shown in the matrix : 1 . * _ supervised _ * learning uses labeled data for training . _ semi - supervised _ * learning uses both labeled and unlabeled data for training . _ unsupervised _ * learning has no labeled data available for training . _ human - in - the - loop(hitl ) _ * learning incorporates active human feedback to algorithm s decisions into the knowledge base and/or algorithms . _ game theory(gt)_-based learning considers learning as a series of strategic interactions between the model learner and actors with conflicting goals . the actors can be data generators , feature generators , chaotic human actors , or a combination@xcite . for `` means of attacks '' in table 1 , server , network , and user are straightforward and intuitive , so here we only describe `` client app '' and `` client machine '' . _ client app _ * is any browser - based means of attack on any client device , and * _ client machine _ * is any non - browser - based means of attack on any client device . as shown in table 1 , the majority of surveyed papers in different security domains use supervised learning to deal with passive or semi - aggressive attackers . however , the core requirement of supervised learning - labeled data - is not always viable or easy to obtain , and authors have repeatedly written about the difficulty of obtaining labeled data for training . based on this observation , we conclude that _ exploring semi - supervised and unsupervised learning approaches would expand the research foundation of ml applications in security domains , because semi - supervised and unsupervised learning can utilize unlabeled datasets which had not been used by supervised learning approaches before . _ moreover , during our survey , we realized that many ml applications in security assume that training and testing data come from the same distribution ( in statistical terms , this is the assumption of stationarity ) . however , in the real world , it is highly unlikely that data are stationary , let alone that the data could very well be generated by an adversarial data generator producing training and/or testing data sets , as the case in @xcite , or simply be generated responding to specific models as in @xcite . our observation from the comprehensive survey confirmed @xcite s statement , and we propose that _ gt - based learning approaches and hitl learning system designs should be explored more , in order to design more efficient security defense mechanisms to deal with active and unpredictable adversaries . at the same time , human knowledge and judgment in htil should go beyond feature engineering , to providing feedback to decisions made by ml models_. some theory - leaning papers have modeled spam filtering as bayesian games or stackelberg games@xcite . use cases in data sampling , model training with poisoned or low - confidence data have also been briefly explored in literature@xcite . based on seminal works and establishments in notable venues , the gradually increasing levels of interest in ml research applied to security is fairly visible . here we gathered some milestone events : 1 . 1987 : denning published `` an intrusion detection system '' @xcite , first framing security as a learning problem 2 . 1998 : darpa ids design challenge@xcite 3 . 1999 : kdd cup ids design challenge@xcite 4 . 2008 : ccs hosted the 1st aisec workshop . continues to operate each year@xcite 5 . 2007 , 2008 : twice , kdd hosted the international workshop on privacy , security , and trust(pinkdd)@xcite 6 . 2010 , 2012 : twice , kdd hosted intelligence and security informatics workshop(isi)@xcite 7 . 2011 : `` adversarial machine learning '' published in 4th aisec@xcite 8 . 2012 : `` privacy and cybersecurity : the next 100 years '' by landwehr et al published@xcite 9 . 2013 : manifesto from dagstuhl perspectives workshop published as `` machine learning methods for computer security '' by joseph et al @xcite . 10 . 2014 : kdd hosted its 1st `` security & privacy '' session in the main conference program@xcite 11 . 2014 : icml hosted its 1st , and so far the only workshop on learning , security , and privacy(lsp)@xcite 12 . 2016 : aaai hosted its 1st artificial intelligence for cyber security workshop(aisc)@xcite despite the surge of research interests and industry applications in the intersection of ml and security , few surveys or overviews were published after 2008 , the watershed year of increasing interest in this particular domain . in 2013 @xcitesurveyed server - side web application security , @xcite surveyed data mining applied to security in the cloud focusing on intrusion detection , @xcite discussed an ml perspective in network anomaly detection . while they are helpful and informative , the former two are limited by their scope and perspective , and the latter serves as a textbook , hence absent the quintessential of survey - mapping the progresses and charting the state - of - art . a collection of papers in 2002 and 2012 @xcite discussed applications of dm in computer security , but lacks a systematic survey on ml applications in resolving security issues . @xcite briefly compared two network anomaly detection techniques , but limited in scope . @xcite of 2009 conducted a comprehensive survey in anomaly detection techniques , some involving discussions of security domains . the dagstuhl manifesto in 2013 @xcite articulated the status quo and looked to the future of ml in security , but the majority of the literature listed were published before 2008 . @xcite of 2010 highlighted use cases and challenges for ml in network intrusion detection , but did not incorporate a high - level review of ml in security in recent years . research works on botnets among our surveyed literature focuses mainly on designing systems to detect command - and - control(c&c ) botnets , where many bot - infected machines are controlled and coordinated by few entities to carry out malicious activities@xcite . those systems need to learn decision boundaries between human and bot activities , therefore ml - based classifiers are at the core of those systems , and are often trained by labeled data in supervised learning environments . the most popular classifier is support vector machines(svms ) with different kernels , while spatial - temporal time series analysis and probabilistic inferences are also notable techniques employed in ml - based classifiers . topic clustering , mostly seen in natural language processing(nlp ) , is used to build a large - scale system to identify bot queries @xcite . in botnet detection literature , 3 core assumptions are widely shared : 1 . botnet protocols are mostly c&c @xcite 2 . individual bots within same botnets behave similarly and can be correlated to each other @xcite 3 . botnet behaviors are different and distinguishable from legitimate human user , e.g. human behaviors are more complex@xcite other stronger assumptions include that bots and humans interact with different server groups @xcite , and content features from messages generated by bots and human are independent @xcite . while classification techniques differ , wls , bls , hypothesis testing , and a classifier @xcite are usual system components . attempts have been made to abstract state machine models of network to simulate real - world network traffic and create honeypots @xcite . ground truths are often heuristic @xcite , labeled by human experts , or a combination - even at large scale , human labeled ground truths are used , for example in @xcite , game masters visual inspections serve as ground truth to detect bots in online games . in retrospect , the evolution of botnet detection is clear : from earlier and more straightforward uses of classification techniques such as clustering and nb , the research focus has expanded from the last step of classification , to the important preceding step of constructing suitable metrics , that measures and distinguishes bot - based and human - based activities@xcite . classifying dns domains that distribute or host malware , scams , and malicious content has drawn research interest especially in passive dns analysis . there are two main approaches : reputation system@xcite and classifier@xcite . reputation system scores benign and malicious domains and dns hosts , and a ml - based classifier learns boundaries between the two . nonetheless , both reputation system and classifier use various decision trees , random forest(rf ) , nave bayes(nb ) , svm , and clustering techniques for mostly supervised learning - based scoring and classification . many features used are from protocols and network infrastructures , e.g. border gateway protocol(bgp ) and updates , automated systems(as ) , registration , zone , hosts , and public bls . similar to botnet detectors , variations of bl , wl , and honeypots@xcite are used in similar functions as knowledge bases , while ground truths are often taken from public bls , limited wls , and anti - virus(av ) vendors such as mcafee and norton @xcite . but before any ml attempts take place , most studies would assume the following : 1 . malicious uses of dns are distinct and distinguishable from legitimate dns services . 2 . the data collection process - regardless of different names such as data flow , traffic recorder , or packet assembler - follows a centralized model . in other words , all the traffic / data / packets flow through certain central node or nodes to be collected . stronger assumptions include that as hijackers can not manipulate as path before it reaches them@xcite , and maliciousness will trigger an accurate ip address classifier to fail@xcite . besides analyzing the status quo , @xcite showed efforts to preemptively protect network measurement integrity and predict potentially malicious activities from web domains and ip address spaces . both offense and defense for access control , authentication , and authorization reside within the domain of security services . defeating audio and visual captchas(completely automated public turing test to tell computers and humans apart)@xcite , cracking passwords@xcite , measuring password strengths@xcite , and uncovering anonymity@xcite are 4 major use cases . on the offense , specialized ml domains such as computer vision , signal processing , and nlp automate attacks on user authentication services i.e. textual or visual passwords and captchas , and uncover hidden identities and services . on the defense side , entropy - based and ml - based systems calculate password strengths . other than traditional user authentication schemes , behavioral metrics of users are also introduced . following the generalized ml pipeline shown in figure 1 , the `` classifier '' is replaced by `` recognition engine '' in the password cracking process , and `` user differentiation engine '' in authentic metric engineering @xcite . hence the process becomes : `` data source @xmath0 pre - process & feature extraction @xmath0 recognition or user differentiation engine @xmath0 decision '' for ml - based security services . a noteworthy trend to observe , is that attacks on captchas are getting more generalized - from utilizing svm in 2008 to attack a specific type of text captcha@xcite , in 2015 a generic attach approach to attack text - based captcha @xcite was proposed . ml - based attacks on textual and visual captcha typically follow the 4-step process : 1 . _ segmentation _ : e.g. signal to noise ratio(snr ) for audio ; hue , color , value(hsv ) for visual @xcite 2 . _ signal or image representation _ : e.g. discrete fourier transformation(audio)@xcite , letter binarization(visual ) @xcite 3 . _ feature extraction _ : e.g. spectro - temporal features , character strokes @xcite 4 . _ recognition _ : k - nearest neighbor(knn ) , svm(rbf kernel ) , convolutional neural networks(cnn ) @xcite on the side of password - related topics in security services , there are 2 password models : whole - string markov models , and template - based models @xcite . concepts in statistical language modeling , such as natural language encoder and n - grams associated with markov models(presented as directed graphs with nodes labeled by n - grams ) , and context - free grammars are common probabilistic foundations to build password strength meters and password crackers @xcite . ml research in software and applications security mostly concentrate on web application security in our survey , and have used supervised learning to train popular classifiers such as nb and svm to detect web - based malware and javascript(js ) code@xcite , filter unwanted resources and requests such as malicious advertisements@xcite , predict unwanted resources and requests(e.g . future blacklisted websites)@xcite , and quantify web application vulnerabilities@xcite . while @xcite explored building web application anomaly detector with scarce training data , most use cases follow the supervised paradigm assuming plentiful labeled data : data source(web applications , static / dynamic analyzers ) @xmath0 feature extraction(often with specific pre - filter , metrics , and de - obfuscator if needed ) @xmath0 classifiers trained with labeled data . apart from this supervised setting , if a human expert s feedback is added after classifiers decisions@xcite , it forms a semi - supervised system . regardless of system designs , the usual assumption holds : _ malicious activities or actors are different from normal and benign ones likely do not change much_. the knowledge bases of normality and abnormality can vary , from historical regular expression lists@xcite to other publicly available detectors@xcite . graph - based algorithms@xcite and image recognition@xcite are both used in resource filtering , but in detecting js malware and evasions and quantifying leaks , having suitable measurements of similarities is a significant focal point . indeed , from @xcite , ml - based classifiers do well in finding similarities between mutated malicious code snippets , while the same code pieces could evade static or dynamic analyzer detections . as landwehr noticed@xcite , ml can be applied in spm . however , in automatic fingerprinting of operating systems(os ) , c4.5 decision tree , svm , rf , knn - some most commonly used ml - based classifiers in security - failed to distinguish remote machine instances with coarse- and fine - grained differences , as the algorithms can not exploit semantic knowledge of protocols or send multi - packet probes @xcite . yet by taking advantage of semantic and syntactic features , plus semi - supervised system design , @xcite showed that svm(optimized by sequential minimal optimization[smo ] algorithm ) , knn , and nlp techniques do well in android spm . on the other hand , in vulnerability management , @xcite , clustering techniques have done well in predicting future incidents and infer vulnerability patterns in code , as well as nb , svm , and rf in ranking risks and identifying proper permission levels . both vulnerability management and spm also focus on devising proper metrics for ml applications : from heuristics based on training set @xcite , jaro distance @xcite , to outside reputation system oracles @xcite , metrics are needed to compare dependency graphs , string similarities , and inferred vulnerability patterns . in most use cases , because of the need for labeled data to train supervised learning systems , many systems follow the generalized training process in figure 1 : `` knowledge base @xmath0 offline trainer @xmath0 online or offline classifier '' . when policy management decisions need feedback , a hitl design is in place where end human users feedback is directed to knowledge base . one distinguishing tradition in ml applications research in this domain , is a strong emphasis on measurement - selecting or engineering proper similarity or scoring metrics are often important points of discussion in research literature . from earlier uses of heuristics in clustering algorithms , to more recent semantic connectivity measurement applied in semi - supervised systems , both the metrics and the system designs for vulnerability and security policy management have evolved to not only identify , but also to infer and predict future vulnerable instances . compared to other security domains , ml research in information flow and ddos focus more on evasion tactics and limits of ml systems in adversarial environments . hence we grouped together the two sub - domains , and marked studies in table 7 with `` ( if ) '' and `` ( ddos ) '' accordingly . for ddos@xcite , the usual assumption is that _ patterns of attack and abuse traffic are different from normal traffic _ @xcite , but @xcite challenged it by proposing an adversary who can generate attributes that look as plausible as actual attributes in benign patterns , and caused failure in ml - based automated signature generation to distinguish benign and malicious byte sequences . then , @xcite introduced gt to evaluate ddos attack and defense in real - world . for information flow@xcite , assumptions can take various forms . in pdf classifiers based on document structural features , it is _ malicious pdf has different document structures than good pdfs _ @xcite ; in android privacy leak detector , it is _ _ the majority of an android application s semantically similar peers has similar privacy disclosure scenarios__@xcite . but @xcite poses semi - aggressive and active attackers with some information about the data , feature sets , and/or algorithms , and then attackers successfully evade ml - based pdf classifiers . another example is , pdf malware could be classified @xcite , and then a generic and automated evasion technique based on genetic programming is successfully experimented@xcite . overall , while using svm , rf , and decision trees trained with labeled data to detect and predict ddos and malicious information and data flows , ml applications in information flow and ddos challenge the usual assumption of stationary adversary behaviors . from collecting local information only , to proposing a general game theory - based framework to evaluate ddos attacks and defense , and from using static method to detect malicious pdf file to generic automated evasion , the scope of ml applications in both ddos and if have expanded and generalized over the years . program - centric or system - centric , there are 3 areas that draw most ml application research attention in malware : malware detection@xcite , classifying unknown malware into families@xcite , and auto - extract program or protocol specifications @xcite . realizing the signature and heuristic - based malware detectors can be evaded by obfuscation and polymorphism @xcite , more behavior - based matching and clustering systems and algorithms have been researched . figure 1 already shows a generalized ml system design for malware detection and classification , and a more detailed description is below : 1 . collect malware artifacts and samples , analyze them , execute them in a controlled virtual environment to collect traces , system calls , api calls , etc . @xcite . or , directly use information from already completed static and/or dynamic analyses . 2 . decide or devise similarity measurements between generalized binaries , system call graphs(scg ) , function call graphs(fcg ) , etc . , then extract features @xcite 3 . classify malware artifacts into families in - sample , or cluster them with known malware families . the classifiers and clustering engines are usually trained with labeled data@xcite . popular ones are svm and rf for classification , and hidden markov model(hmm ) and knn alongside different clustering techniques . even in the use case of auto - extract specifications , supervised learning with labeled data is needed when behavior profiles , state machine inferences , fuzzing , and message clustering are present . evasion techniques of detectors and poisoning of ml algorithms are also discussed , and typical evasion techniques include obfuscation , polymorphism , mimicry , and reflecting set generation@xcite . malware detection and matching based on structural information and behavior profiles@xcite show a tendency to use graph - based clustering and detection algorithms , and similarity measurement used in these algorithms have ranged from jaccard distance to new graph - based matching metrics . while clustering techniques have been mostly used in malware detection , a nearest neighbor technique is explored to evade malware detection . spams , malicious webpages and urls that redirect or mislead un - suspecting users to malware , scams , or adult content @xcite is perhaps as old as civilian use of the internet . research literature mostly focus on 3 major areas : detecting phishing malicious urls@xcite , filtering spam or fraudulent content @xcite , and detecting malicious user account behaviors@xcite . moreover , because phishing is a classic social engineering tactic , it is often the gateway of many studies to detect malicious urls , spam , and fraudulent content . to identify malicious urls , ml - based classifiers draw features from webpage content(lexical , visual , etc . ) , url lexical features , redirect paths , host - based features , or some combinations of them . such classifiers usually act in conjunction with knowledge bases which are usually in - browser url bls or from web service providers . if the classifier is fed with url - based features , it is common to set an url aggregator as a pre - processor before extracting features . mostly using supervised learning paradigm , nb , svm with different kernels , and lr are popular ml classifiers for filtering spam and phishing . meanwhile , gt - based learning to deal with active attackers is also evaluated in spam filtering . @xcite evaluates a bayesian game model where the defense is not fully informed of the attacker s objectives and the active adversary can exercise control over data generation , @xcite proposes a stackelberg game where spammer reacts to the learner s moves . stronger assumptions also exist : for example , @xcite assumes spammers phone blocks follow a beta distribution as conjugate prior for bernoulli and binomial distribution . another social engineering tactic is spoofing identities with fake or compromised user accounts , and detection of such malicious behaviors utilize features from user profiles , spatial- , temporal- , and spatial - temporal patterns , and user profiles are used in particular to construct normality . graph representation and trust propagation models are also deployed to distinguish genuine and malicious accounts with different behavior and representations@xcite . tracing the chronology of applying ml to defend against social engineering , one trend is clear : while content- , lexical- , and syntactic - based features are still being widely used , constructing graph representations and exploring temporal patterns of redirect paths , events , accounts , and behaviors have been on the rise as feature spaces for ml applications in defend against social engineering efforts . accordingly , the ml techniques have also changed from different classification schemes to graphic models . it is also noteworthy that in @xcite , addressing adversarial environments challenges to ml systems is elaborated as primary research areas , instead of a short discussion point . from feature sets to algorithms and systems , ids has been extensively studied . however , as @xcite cautioned , ml can be easily conflated with anomaly detection . while both are applied to build ids , important difference is that ml aims to generalize expert - defined distinctions , but anomaly detection focuses on finding unusual patterns , while attacks are not necessarily anomalous . for example , @xcite distinguished n - gram model s different use cases : anomaly detection uses it to construct normality(hence more appropriate when no attack is available for learning ) , and ml classifiers learn to discriminate between benign and malicious n - grams(hence more appropriate when more labeled data is present ) . since 2008 , works at top venues have added to the rigor for ml applications in ids . for example , a common assumption of ids is : _ anomalous or malicious behaviors or traffic flows are fundamentally different from normal ones _ , but @xcite challenges the assumption by studying low - cardinality intrusions where attackers do nt send a large number of probes . to address adversarial learning environment and minimal labels in training data , semi - supervised paradigms , especially active learning , are also used@xcite . heterogeneous designs of ids in different use cases give rise to many ad - hoc evaluations in research works , and a reproducibility and comparison framework was proposed to address the issue@xcite . meanwhile , techniques such as graph - based community detection@xcite , time series - based methods@xcite , and generalized support vector data description in cyber - physical system and adversarial environment for auto - feature selection@xcite , have also emerged . although they carry different assumptions of normality and feature representations , the supervised ml system design remains largely the same . besides the fact the more techniques and use cases have been proposed , the focus of research in ids had evolved from discovering new techniques and use cases , to rigorously evaluating fundamental assumptions and workflows of ids . for example , while feature selection has stayed as a major component , there are re - examination of assumptions and measurements on what constitutes normality and abnormality@xcite , alternative to more easily acquire data and use low - confidence data for ml systems@xcite , and proposal on validating reproducibility of results from different settings@xcite . one key goal of our sok survey is to help researchers look into the future . ml applications in security domains are attracting academic research attention as well as industrial interest , and this presents a valuable opportunity for researchers to navigate the landscapes between ml theories and security applications . there are also opportunities to explore if there are some types of ml paradigms that are especially well suited to particular security problems . apart from highlighting that 1 ) semi - supervised and unsupervised ml paradigms are more effective in utilizing unlabeled data , hence ease the difficulty of obtaining labeled data , and 2 ) gt - based ml paradigms and hitl ml system designs will become more influential in dealing with semi - aggressive and aggressive attackers , we also share the following seven speculations of future trends , based on our current sok . 1 . * metric learning * : measurement has become more and more conspicuous for ml research in security , mostly in similarity measurement for clustering algorithms@xcite . proper measurements and metrics are also used to construct ground truths to evaluate ml - based classifiers , and also have important roles in feature engineering@xcite . given the ubiquitous presence of metrics and the complex nature of constructing them , ml applications in security will benefit much from metric learning . * nlp * : malicious content , spam , and malware analysis and detections have used tools from statistical language modeling(e.g . n - gram - based representation for strings in code and http request)@xcite , as textual information explodes , nlp will become more widely used beyond source filtering and clustering e.g. @xcite use n - gram models to infer state machines of protocols . * upstream movement of ml in security defense designs. in malware detection and classifications , behavior- and signature - based malware classifiers have used inputs from static and dynamic binary analysis as features@xcite , and @xcite already shows rnn can be applied to automatically recognize functions in binary analysis . we also see ml algorithms applied in vulnerability , device , and security policy management , ddos mitigation , information flow quantifications , and network infrastructure@xcite . hence , it is reasonable to expect that more ml systems and algorithms will move upstream in more security domains . scalability * : with increasing amount of data from growing numbers of information channels and devices , scale of ml - based security defenses will become a more important aspect in researching ml applications in security @xcite . as a result , large - scale systems will enable * distributed graph algorithms * in malware analysis , as path hijacker tracing , cyber - physical system fault correlation , etc .. @xcite 5 . * specialized probabilistic models * will be applied beyond the context of classifiers , e.g. access control@xcite . high fp rates have always been a concern for system architects and algorithm researchers @xcite . * reducing fp rates * will grow from an ad - hoc component in various system designs , to independent formal frameworks , algorithms , and system designs . * privacy enforcement * was framed as a learning problem recently in @xcite , in the light of many publications on privacy - preservation in ml algorithms , and privacy enhancement by probabilistic models@xcite . this new trend will become more prominent . in this paper , we analyzed ml applications in security domains by surveying literature from top venues of our field between 2008 and early 2016 . we attempted to bring clarity to a complex field with intersecting expertises by identifying common use cases , generalized system designs , common assumptions , metrics or features , and ml algorithms applied in different security domains . we constructed a matrix showing the intersections of ml paradigms and three different taxonomy structures to classify security domains , and show that while much research has been done , explorations in gt - based ml paradigms and hitl ml system designs are still much desired ( and under - utilized ) in the context of active attackers . we point out 7 promising areas of research based on our observations , and argue that while ml applications can be powerful in security domains , it is critical to match the ml system designs with the underlying constraints of the security applications appropriately . we would like to thank megan yahya , krishnaprasad vikram , and scott algatt for their time and valuable feedback . c. landwehr , d. boneh , j. c. mitchell , s. m. bellovin , s. landau , and m. e. lesk , `` privacy and cybersecurity : the next 100 years , '' _ proceedings of the ieee _ special centennial issue , 2012 . t. ahmed , b. oreshkin , and m. coates , `` machine learning approaches to network anomaly detection , '' in _ proceedings of the 2nd usenix workshop on tackling computer systems problems with machine learning techniques _ , 2007 . p. g. kelley , s. komanduri , m. l. mazurek , r. shay , t. vidas , l. bauer , n. christin , l. f. cranor , and j. lopez , `` guess again ( and again and again ) : measuring password strength by simulating password - cracking algorithms , '' in _ sp 2012_. r. wang , w. enck , d. reeves , x. zhang , p. ning , d. xu , w. zhou , and a. m. azab , `` easeandroid : automatic policy analysis and refinement for security enhanced android via large - scale semi - supervised learning , '' in _ usenix security 2015_. k. lu , z. li , v. p. kemerlis , z. wu , l. lu , c. zheng , z. qian , w. lee , and g. jiang , `` checking more and alerting less : detecting privacy leakages via enhanced data - flow analysis and peer voting . '' in _ ndss 2015_.	the idea of applying machine learning(ml ) to solve problems in security domains is almost 3 decades old . as information and communications grow more ubiquitous and more data become available , many security risks arise as well as appetite to manage and mitigate such risks . consequently , research on applying and designing ml algorithms and systems for security has grown fast , ranging from intrusion detection systems(ids ) and malware classification to security policy management(spm ) and information leak checking . in this paper , we systematically study the methods , algorithms , and system designs in academic publications from 2008 - 2015 that applied ml in security domains . 98% of the surveyed papers appeared in the 6 highest - ranked academic security conferences and 1 conference known for pioneering ml applications in security . we examine the generalized system designs , underlying assumptions , measurements , and use cases in active research . our examinations lead to 1 ) a taxonomy on ml paradigms and security domains for future exploration and exploitation , and 2 ) an agenda detailing open and upcoming challenges . based on our survey , we also suggest a point of view that treats security as a game theory problem instead of a batch - trained ml problem . * keywords * : security , machine learning , large - scale applications , game theory , security policy management
the noise excess free ( phase - insensitive ) quantum record / upload of single - mode light into the atomic memory can be defined by the transformation @xmath1 of the light quadratures @xmath2 and @xmath3 to the atomic quadratures @xmath4 and @xmath5 in the heisenberg picture . here @xmath6 stands for the transmission coefficient of the record / upload . the quadrature noisy operators @xmath7 and @xmath8 describe the added noises in the state transfer . the variance of vacuum noise is considered to be unity . from the heisenberg uncertainty principle , the minimal variance of both @xmath7 and @xmath8 is @xmath9 ( @xmath10 ) . in this case , there is no noise excess in the record / upload process . it corresponds to transmission exhibiting only a pure loss , which can be modeled by a virtual beam splitter between light and atoms with vacuum in the free port . such the noise excess free record / upload is advantageous because it can preserve some important non - classical properties of quantum states . it will never completely break entanglement of gaussian state , vanish sub - poisson behavior of single - photon state @xcite or quantum superposition of coherent states @xcite . also only loss in the record / upload will not break security of the continuous - variable key distribution protocol with coherent states if the reverse reconciliation is used @xcite . basically , information encoded into the quadratures and gaussian entanglement or security of key distribution are not changed if the recorded / upoload state is only unitarily transformed by a known gaussian operation like the phase shift , displacement and squeezing . in this case , the record / upload is noise excess free up to that unitary operation . particularly , we focus on the noise excess free record / upload up to the squeezing operation : @xmath11 , @xmath12 . for an application , either the squeezing could be actively post - corrected in the memory , as will be proposed below , or if it is not necessary , it can be finally simply corrected on the measured data . to increase a quality of the upload , also a pre - squeezing operation @xmath13 , @xmath14 on the light mode before record / upload can be considered . it transforms the input state to its squeezed version before the record / upload into the memory . it will be shown , both the pre - squeezing of input state and the post - squeezing correction can remarkably help to reach the noise excess free record or even the lossless upload of quantum state into the memory . consider now the quantum memory experiment in ref . @xcite , see also ref . @xcite for more details . in that experimental setup , there are two simultaneously available quantum non - demolition ( qnd ) interactions between the light and atoms inside the cells . basically , they couple together the mode of light described by two complementary quadratures @xmath2 and @xmath3 and the effective collective atomic mode having complementary quadratures @xmath4 and @xmath5 . both the qnd transformations can be simply described in heisenberg picture : @xmath15 and @xmath16 where @xmath17 is an effective coupling constant @xcite . either coupling ( [ coupl1 ] ) or ( [ coupl2 ] ) can be separately activated in the same set - up @xcite . both are the particular qnd interactions , but the following analysis is generally valid for any kind of the qnd transformation between the quadratures of light and atomic memory . if the coupling ( [ coupl1 ] ) is considered , the quadrature @xmath3 can be directly written into the memory by the light - atom interaction , up to the added noise from the atomic quadrature @xmath4 . to write the complementary quadrature @xmath2 , the light pulse passing through the memory is measured by homodyne detection and the photocurrent controls the magnetic field applied with an adjustable gain to the atomic cells . by this feed - forward technique , the atomic quadrature @xmath5 can be displaced whereas the quadrature @xmath4 is not disturbed . this is standard record mechanism used in the ref . ( @xcite ) . now we assume the record of unknown quantum state up to the squeezing of the recorded state . after such the procedure , the transformation of the atomic quadratures takes the following form @xmath18 where @xmath19 is an overall gain of the feed - forward correction and @xmath20 is a scaling factor representing the squeezing of recorded state . the effective atomic mode is considered initially in vacuum state with the unit variance . to obtain the noise excess free record up to the squeezing , the parameters have to satisfy the following equations @xmath21 the solution of ( [ eq1 ] ) : @xmath22 and @xmath23 , gives a possibility to achieve the noise excess free record described by eqs . ( [ noiseless ] ) up to the squeezing . the record transmission coefficient @xmath24 of the transfer from the light mode to atomic mode shows that an unknown state written inside the memory is always squeezed by the factor @xmath23 . but even for very small @xmath25 , any state can be ( up to the squeezing ) written into the memory with no noise excess . for a feasible gain around @xmath26 , the upload can be up to the noise excess free attenuation @xmath27 up to the squeezing . such the noise excess free record even without the squeezing correction can be useful , for example , for the manipulations with gaussian entanglement . using computable measure of entanglement for gaussian state @xcite , it is straightforward to prove the following comparison . let us compare the cases after the record , without squeezing correction and with perfect squeezing correction . consider single mode from two - mode gaussian entangled state which is recorded into the memory . for any single - mode noisy gaussian operation after the noise excess free record ( up to the squeezing ) , even with an arbitrary small @xmath25 , no matter that the uncorrected squeezing reduces entanglement it will never help to subsequent gaussian channel to completely break entanglement . it means that for both the cases , the thresholds to pass entanglement through the record procedure plus consecutive gaussian channel are identical . it illustrates a practical advantage of the noise excess free record up to the squeezing . such the result is generally impossible to obtain if there is a noise excess in the record . to actively compensate the squeezing of the recorded state inside the memory , the squeezing post - correction is necessary . the squeezing post - correction should make transformation @xmath28 and @xmath29 to reduce effect of the squeezing in the atomic memory . since the total record process could exhibit pure loss , the squeezing post - correction can be implemented at a cost of an additional loss in the record . to build post - correction operation , the second kind of the coupling ( [ coupl2 ] ) with the coupling constant @xmath30 is now assumed in the same experimental set - up . now , the light mode described by @xmath31 and @xmath32 is considered in vacuum state . thus only bright local oscillator is injected into the cells . the homodyne measurement of @xmath33 quadrature of the pulse outgoing from the cells is followed by the feed - forward displacement of the atomic quadrature @xmath4 by the magnetic field . to achieve a desirable squeezing post - correction @xmath28 and @xmath29 ( @xmath34 is considered to compensate the squeezing in eqs . ( [ record ] ) ) up to a pure loss ( but without any noise excess ) , the transformations @xmath35 have to satisfy the following conditions @xmath36 it can be achieved if @xmath37 and then @xmath38 is given for any @xmath39 . the additional losses introduced by the post - correction are characterized by the transmission @xmath40 . since @xmath34 , the squeezing post - correction can be performed only in the @xmath5 quadrature . to precisely compensate the squeezing introduced in the record , it is necessary to adjust @xmath41 . from this follows that @xmath42 and total transmission coefficient of the record plus squeezing correction is given by @xmath43 the maximal transmission @xmath44 is achieved for @xmath26 ( @xmath45 ) which is feasible in the experiment in ref . @xcite . as the result , any quantum state can be in principle deterministically recorded without noise excess as is described by eq . ( [ noiseless ] ) up to the transmission @xmath46 . another open possibility how to compensate the squeezing in the recorded state is a pre - squeezing @xmath47 and @xmath13 of the state which should be recorded . from a simple calculation follows that this configuration does not lead in principle to the noise excess free record . but if both the pre - squeezing and squeezing post - correction are considered together , the local pre - squeezing helps to increase the coupling constant @xmath17 . taking both the pre - squeezing and squeezing post - correction into the consideration , the record mechanism is described by the transformations : @xmath48 where @xmath49 stands for the scaling factor corresponding to the pre - squeezing of the light mode . to approach the noise excess free record , it is necessary to adjust electronic gain @xmath50 and then the squeezing correction after the record is determined by @xmath51 . thus for a given @xmath17 , the squeezing post - correction has to be larger ( @xmath20 should be smaller ) as @xmath49 is bigger . interestingly , the transmission coefficient @xmath52 of the record up to the squeezing is effectively increased as the squeezing factor @xmath49 is larger . remarkably , by only local pre - squeezing of the state , which should be recorded into memory , the coupling constant @xmath17 is increased . since @xmath53 approaches unity for sufficiently large @xmath49 , a lossless record up to the squeezing can be approached . it is even possible for an arbitrary small @xmath25 if the sufficient pre - squeezing is available . on the other hand , if the uploaded state inside the memory should be not squeezed , the squeezing correction has to be applied . then , to reach the limit @xmath44 , @xmath54 has to be satisfied and it can be approached for any @xmath17 as the pre - squeezing is increasing . this deterministic record can be used to upload not only coherent states and gaussian entangled states , but also highly nonclassical single - photon state as an useful quantum resource . the single - photon state can be prepared by the single - photon subtraction from the squeezed light @xcite . actually , the state prepared by such the procedure approaches the squeezed single - photon state @xcite . this squeezing would be advantageous for the upload procedure as has been discussed . but , to have exactly single - photon state inside memory , it is necessary to actively apply the squeezing post - correction . this operation can be naturally included into the uploading scheme proposed above . in a result , to get single - photon state inside the memory , the single - photon subtraction can be simply applied on the squeezed state followed by the uploading procedure proposed above . but still the upload has some uncertainty , we do not know if the single photon state is actually inside the memory . therefore , the uploaded state is a mixture @xmath55 where @xmath56 which optimally approaches @xmath57 as @xmath54 . the wigner function of that state is never negative since @xmath58 @xcite , but still any mixture ( [ mix ] ) with @xmath59 is a non - classical state @xcite . for example , the statistics of ( [ mix ] ) is always sub - poissonian for arbitrary @xmath59 . in experiments with the single photons , this method can be used only if the post - selection on the fixed number of photons is finally applied to have complete control how many photons participated in the experiment . note , there is a similar requirement for the linear optical experiments with the single photons . they work correctly only if all the photons are finally detected . fortunately , there is a better way how to upload the single - photon state . instead of the deterministic electro - optical feed - forward correction , a proper post - selection of measured data from the homodyne detector is applied . it makes the upload only probabilistic , but since the uploaded state is resource , it is acceptable if the rate is not too small . for the upload of the gaussian states , there is no gain from this probabilistic method over the previous deterministic feed - forward correction . but for the non - classical non - gaussian states , for example , for the squeezed single - photon state produced by the single - photon subtraction @xcite the post - selection of the measured quadrature @xmath33 in a tiny interval @xmath60 around value @xmath61 will remarkably help . to simply understand why it will help , the unitary coupling ( [ coupl2 ] ) between light and atoms can expand for very small @xmath62 , where @xmath63 is an interaction time , as @xmath64 , where @xmath65 and @xmath66 are annihilation operators of the light and atomic modes . in this convention , the vacuum noise has variance @xmath0 . the ideal projection on the eigenstate @xmath67 of @xmath33 corresponding to eigenvalue @xmath61 can be rewritten in a form of @xmath68 . initially , light is in the fock state @xmath69 and atomic mode is in the vacuum state @xmath70 . straightforwardly , we can prove that @xmath71 and from it follows that in the limit of weak coupling and very tight post - selection @xmath72 , it is possible to upload single - photon state to the memory without any loss and noise . to calculate the upload for a finite @xmath73 , we use the formalism of wigner functions . the squeezed single - photon state is represented by the wigner function @xmath74 where @xmath20 is squeezing parameter . the atomic memory is considered initially in vacuum state described by the wigner function @xmath75 . the application of the coupling ( [ coupl2 ] ) on the product @xmath76 followed by the homodyne detection and above mentioned post - selection will generate the wigner function @xmath77+\mbox{erf}\left[\frac{\sqrt{2}}{a } \left(b+\kappa x\right)\right ] \right)-\nonumber\\ \frac{4\sqrt{2}}{\sqrt{\pi^3 } s d } \exp\left(-2\frac{b^2}{a^2}-2x^2\frac{d^2}{a^2}-2p^2\frac{a^2}{d^2 } \right)\times\nonumber\\ \left(b\mbox{cosh}\left[\frac{4b\kappa x}{a^2}\right]+\kappa x \mbox{sinh}\left[\frac{4b\kappa x}{a^2}\right]\right)\nonumber\\\end{aligned}\ ] ] of the state uploaded into the memory , where @xmath78 and @xmath79 $ ] is error function . the success rate of the upload is @xmath80-\sqrt{\frac{2}{\pi } } \exp\left(-\frac{2b^2}{d^2}\right)\frac{2a^2b}{d^3}.\ ] ] if the squeezing post - correction @xmath81 and @xmath82 is applied , then as @xmath73 goes to zero , the wigner function of atomic state approaches @xmath83 where @xmath84 is exactly the wigner function of the single photon . the squeezing post - correction does not depend on the post - selection threshold @xmath73 . to approach controllable upload , the squeezing post - correction proposed in the previous section can not be used , since it can work only up to additional losses which destroys the desired control of upload . but , fortunately , if both @xmath85 are small then the squeezing post - correction can be approximately omitted and the uploaded state is @xmath86 as the result , the single - photon state is probabilistically uploaded into the memory with full control and remarkably , is not attenuated comparing to the previous deterministic case . thus the post - selection compensates the loss which occurs in the deterministic upload . since the probabilistic upload is obtained for a small @xmath17 , the post - selection also compensates strength of the coupling between light and atoms . but as @xmath17 decreases , the post - selection threshold @xmath73 has to decrease and it reduces the success rate of the upload . to simply measure quality of the single - photon state uploaded into the memory versus the success rate of the upload , the fidelity and negativity of wigner function can be used . the fidelity between the uploaded state described by the wigner function @xmath87 and target single - photon state with the corresponding wigner function @xmath84 is defined as @xmath88 whereas the negativity in the origin of the phase space is given simply by @xmath89 . the expression for the fidelity is rather complex but for both the @xmath17 and @xmath73 small , the expansion @xmath90 shows how the result of a sufficiently tight post - selection practically does not depend on a small @xmath17 . without the squeezing post - correction , the maximal fidelity is exactly @xmath91 in the limit @xmath92 . for a small @xmath73 , the negativity of wigner function is @xmath93 the maximum of negativity @xmath94 can be always approached if @xmath73 is sufficiently small . the @xmath95 is independent on @xmath20 and @xmath17 because the negativity of wigner function is not changed by the pre - squeezing . both the fidelity and negativity can be achieved with the success rate approaching @xmath96 which grows as pre - squeezing parameter @xmath20 is smaller . here the pre - squeezing plays also positive role , it significantly enhances the probability of success . the quantitative results of the single photon upload are depicted in fig.1 . the pre - squeezing @xmath97 can evidently help to increase the success rate for both the fidelity as well as for the negativity , three curves are depicted for @xmath98 . thus squeezing of the single - photon state from the squeezed light by the single photon subtraction is actually an advantage for the upload into the memory . contrary to the deterministic case , the post - selection can effectively increase the coupling between light and atoms . to approach such the lossless upload of single photon state , the in - coupling and out - coupling losses have to be minimized . the in - coupling losses will mix the single - photon state with vacuum state and in a result , a less sub - poissonian state is actually uploaded . the efficiency of homodyne detector can be can be very close to unity and also electronic noise of homodyne detection can be negligible , if the power of local oscillator is high . the out - coupling losses can be joined with the efficiency of the homodyne detection to an effective total out - coupling efficiency @xmath99 . the fidelity ( [ fi ] ) is a decreasing function of @xmath99 , as can be seen from fig . 2 . to approach the losses upload , the coupling constant @xmath17 has to be reasonably high and then , the pre - squeezing in the opposite direction ( @xmath100 ) can help to reach unit fidelity of the single - photon state at a cost of the success rate . a superposition of the coherent states is considered as an another resource for the highly non - linear quantum operations and quantum computing @xcite . it is rather a superposition of many fock states , than just single fock state , therefore it can be interesting to investigate whether the upload can transfer such the state of light into the memory . let us consider the following superposition @xmath101 of coherent states which is pre - squeezed before the upload , where @xmath102 . then the state can be described by the wigner function @xmath103,\end{aligned}\ ] ] where @xmath104 . at the beginning , the vacuum state is inside the atomic memory . after the coupling ( [ coupl2 ] ) and successful post - selection of the measured @xmath33 quadrature in the interval @xmath105 around @xmath61 , the atomic mode is projected to the state with the following wigner function @xmath106\mbox{erf}(p)+\mbox{cerf}(p)\right),\nonumber\end{aligned}\ ] ] where @xmath107 + \mbox{erf}\left[\frac{\sqrt{2}}{a}\left(b+a\kappa p\right)\right],\nonumber\\\end{aligned}\ ] ] and @xmath108\right]+\nonumber\\ & & \mbox{re}\left[\mbox{erf}\left[\frac{\sqrt{2}}{a}\left(ix_0a+b+\kappa p\right)\right]\right].\end{aligned}\ ] ] the success rate of such the uploading process is @xmath109+\right.\nonumber\\ \left.\mbox{erf}[\frac{\sqrt{2}(ix_0a+b)}{d}]-\mbox{erf } [ \frac{\sqrt{2}(ix_0a - b)}{d}]\right).\nonumber\\\end{aligned}\ ] ] if the squeezing post - correction @xmath110 and @xmath111 is applied , the wigner function ( [ wiga ] ) approaches the eq . ( [ cat ] ) as @xmath73 goes close to zero , but with a reduced amplitude @xmath112 . if @xmath73 is enough small and @xmath17 is sufficiently smaller than @xmath20 ( @xmath113 ) then the squeezing post - correction can be approximately omitted . but then also amplitude @xmath114 of the states in the uploaded superposition decreases . fortunately , this amplitude can be successfully compensated by a sufficient pre - squeezing ( @xmath20 is small ) . thus sufficiently large superposition of coherent states can be also uploaded into the quantum memory without any loss . in a contrast to a simple phase - insensitive character of the single photon state , the superposition of coherent states exhibits more complex phase - space behavior . in one quadrature there two symmetrically located gaussian peaks like for the classical mixture of the coherent states . on the other hand , the complementary quadrature shows multiple interference fringes covered by a gaussian envelope . because the fidelity averages both the interference as well as distinguishable peaks , it is not good measure of a quality of the uploaded superposition . from this reason , the marginal probability distribution @xmath115 exhibiting the interference is rather evaluated separately , to directly observe the interference in the superposition . the explicit form of marginal distribution @xmath116 can be approximated by @xmath117\ ] ] for a small post - selection threshold @xmath73 . assuming @xmath118 , the final distribution is @xmath119.\ ] ] for @xmath120 ( @xmath121 ) , @xmath122 approaches the exact marginal distribution of the superposition of the coherent state only the amplitude @xmath123 is reduced . as a result , the attenuation in the deterministic scheme , which for the coherent - state superposition introduces an excess noise in the uploaded state , is substituted here by only a reduction of @xmath124 without any additional noise . the uploaded state closely approaches the ideal pure state superposition with the reduced amplitude . the pre - squeezing plays here also a remarkable role . in fig . 2 ( upper picture ) , there is a marginal distribution of the uploaded state without pre - squeezing . evidently , the interference is vanishing . but if the sufficient pre - squeezing ( -6db ) is applied ( lower picture ) then the fringes can be observed and also the success rate is higher . in conclusion , the deterministic noise excess free record of unknown state and the probabilistic lossless upload of the resource state ( single - photon state , superposition of coherent states ) into quantum memory are proposed , both based on the original quantum memory experiment in ref . first , the deterministic scheme is able to record unknown quantum state only with a reduced transmission ( maximally @xmath0 ) . the pre - squeezing of the state of light going to the memory can compensate small coupling strength between light and atomic memory . but this record is without any trigger controlling the uploading process what is crucial , for example , for the upload of single photons . therefore , the probabilistic scheme which is able to upload single - photon state is proposed . it approaches unit fidelity of the upload at the cost of success rate , even for a weak coupling between the light and atoms . in this case , the pre - squeezing enhances the success rate of the upload . the same method can be used to upload the pure coherent - state superposition , at the cost of the success rate and reduced amplitude of the superposition , but with preservation of the purity of the superposition . in this case , the pre - squeezing not only enhances the success rate , but also effectively enlarges the reduced amplitude of the coherent states in the superposition and thus increases non - classicality of the uploaded state . since the considered scheme has been used in the real experiment set - up , both the methods can be directly used to upload the non - classical resources into quantum memory . the proposed schemes are direct steps towards to storage and operations on highly non - classical quantum states in the continuous - variable quantum memory . * acknowledgments * this research has been supported by projects msm 6198959213 and lc06007 of the czech ministry of education and grant 202/08/0224 of ga cr . i also acknowledges a support by the alexander von humboldt foundation and eu grant fp7 212008 - compas .	in recent continuous - variable quantum memory experiment , one quadrature of light pulse is directly uploaded by the light - atom coupling whereas complementary quadrature is obtained by homodyne measurement of the out - coupled light . subsequently , information from homodyne measurement is written into the memory by feed - back electro - optical control of the atomic state . using the same experimental set - up , a deterministic noise excess free record of unknown quantum states to continuous - variable quantum atomic memory is proposed . further , the memory experiment is extended by the pre - squeezing of the recorded state of light and squeezing post - correction of the recorded state . to upload a resource to the memory , as single - photon state or superposition of coherent states , the post - selection of the measurement results from homodyne detection is suggested . then such the probabilistic upload approaches even a lossless transfer of these highly nonclassical states into the quantum memory . a non - classicality of quantum states is extremely fragile to disturbing effects of a noise coupled to the system . this is a real problem for the transmission and operations with the non - classical states . recently , the same problem arises also in a construction of quantum memory . the quantum memories have been recently demonstrated using electromagnetically - induced transparency @xcite and off - resonant faraday effect @xcite . it is useful to distinguish three basic steps : initial record ( or upload ) of the quantum state , its storage for a longer time inside the memory and final read - out from the memory . in addition , quantum state written to the memory can either carry encoded information or it can be a resource for other quantum operations . to distinguish these two cases , the record of quantum states is used for the states carrying information and the upload of quantum state for the resources . in the former case , encoded information makes the state to be at least partially unknown , whereas in the latter case , the uploaded state is known completely and the upload can be optimized with a respect to that specific state . in this case , a relatively fast probabilistic upload is sufficient to transfer the resource state into the memory . the quantum memory should definitely preserve the non - classical properties of quantum states . it is a necessary for any application of quantum memory , for example , for the quantum repeaters for the long - distance quantum key distribution . focusing on the record / upload process , the noise during this process has to be investigated . there is always vacuum noise which accompanies any loss in the record / upload , but even more destroying is an impact of a noise excess above this vacuum noise penalty . therefore , an noise excess free record / upload to quantum memory should be find . in the other words , the noise excess free means that the record / upload is only pure lossy process . such the record / upload will be not entanglement breaking channel @xcite , will not completely vanish sub - poissonian statistics of single photon state @xcite and also will not smear out interference effects in the superposition of coherent states @xcite . also security of the continuous - variable quantum key distribution is preserved for any loss in the transfer of quantum state if there is no noise excess @xcite . second , more complicated open problem is how to construct the lossless upload of the nonclassical state . such the upload allows to prepare highly non - classical inside the memory . the quantum memory for coherent states of light has been recently demonstrated using a pair of glass cells filled by cs atoms ( at room temperature ) @xcite . in this experiment , the coherent state is generated by the amplitude and phase modulations of the strong continuous - wave ( cw ) millisecond laser pulses . the off - resonant interaction of cw light pulses with the collective spin atoms inside the cell ( faraday effect ) can be describe as quantum non - demolition coupling between effective quadratures of the light and atomic modes . the coupling automatically records a single quadrature of light to the effective quadrature of collective spin of the atoms . the complementary quadrature of light pulse passing through the memory is then measured by homodyne detection and feed - forward by the electro - optical control into the atomic memory . in the experiment @xcite , an average fidelity of recorded coherent states has been shown to be higher than a threshold corresponding to measure - prepare strategy . to approach unit fidelity in the same set - up , a difficult high squeezing of the collective spin state of the atoms before the uploading procedure is required . since the average fidelity has been considered as a figure of merit , the demonstrated record of the coherent state is not the excess noise free and the recorded state has been far a way from a pure state . there were few attempts to achieve noise excess free memory , involving several coherences of each atom in the cell @xcite or multi - pass version of the deterministic memory @xcite . further , the scheme with two passes has been theoretically examined with an additional magnetic field to achieve the deterministic and noise excess free memory with a loss exponentially decreasing as the coupling constant is growing @xcite . these experimentally challenging extensions are mainly orientated to approach the deterministic record / read - out of unknown quantum state carrying information into / from memory . on the other hand , to upload the known highly non - classical resource state , such as the single photon state or the superposition of coherent states into the memory , the probabilistic approach could be more simple and effective . in this paper , the noise excess free version of deterministic record of unknown quantum state into the quantum memory is described and the probabilistic upload of highly non - classical states ( single - photon state , superposition of coherent states ) approaching lossless light - atomic transfer is proposed . in the section ii . , is shown that the original set - up from ref . @xcite can be directly used to obtain noise excess free deterministic record of any quantum state of light for an arbitrary weak non - demolition coupling and without any noise pre - squeezing of the atomic memory . the recorded state is noise excess free but only up to a squeezing of the collective atomic variables . therefore , a squeezing post - correction is suggested to approach phase - insensitive record symmetrical in both the quadratures . further , it is demonstrated that a pre - squeezing of the state of light before entering into the memory remarkably enhances the coupling strength and subsequently , it reduces the attenuation of the uploaded state inside the memory . unfortunately , the application of the squeezing post - correction actively is always at a cost of additional loss and for a higher squeezing of the recorded state in the memory , the additional attenuation is larger . in the result , the purely lossy phase - insensitive record with maximum @xmath0 of transmission can be achieved without any noise excess . it can be restrictive for the upload of the non - classical resources , such as the single - photon state or the superposition of the coherent states . if the upload is lossy , such the single photon is uncontrollably transferred to the memory , without any trigger saying whether the photon is really uploaded . therefore , in the section iii . , a probabilistic upload is proposed based on a post - selection of a result from the homodyne detector in the same memory set - up . the single photon can be probabilistically uploaded into the memory without loss and with an arbitrary precision , although the coupling between the light and atoms in the memory can be weak . the superposition of coherent states can be probabilistically uploaded with a reduced amplitude , but with the purity approaching unity . remarkably , for the probabilistic upload the squeezing post - correction can be omitted if the light - atom coupling is sufficiently weak . to increase the success rate of the single - photon upload , the pre - squeezing of the state of light before the upload can be used . the superposition of coherent states can be uploaded in the same way , but the pre - squeezing plays here more important role , it effectively enhances the amplitude of the coherent states in the uploaded superposition , resulting in more distinguishable interference of the coherent states . because all these proposals are within the existing experiment set - up , they can be directly implemented if the suitable non - classical sources are available .
kondo effect in electronic transport through quantum dots ( qds ) strongly coupled to external leads is a many body phenomenon which occurs at temperatures @xmath1 lower than the so - called kondo temperature @xmath2 , @xmath3 . the kondo temperature is characteristic of a particular system and depends on the corresponding parameters ( energy of the dot level , coupling strength to external leads , coulomb parameter ) . spin fluctuations in the dot , generated by coupling of the dot to external leads , give rise to a narrow peak in the dot s density of states ( dos ) at the fermi level . this kondo peak enhances transmission through the dot , and leads to the unitary limit of the linear conductance at zero temperature , where the conductance reaches @xmath4 . the kondo peak in dos also leads to the so - called zero - bias ( kondo ) anomaly in the differential conductance . the kondo effect in electronic transport through single quantum dots was extensively studied in the last two decades , both theoretically @xcite and experimentally @xcite . recently , the kondo - assisted transport through double - dot systems in the t - shaped @xcite , parallel @xcite , and series @xcite geometries also has been addressed . there are , however , only a few papers on the kondo phenomenon in electronic transport through triple - dot structures @xcite . such complex dot systems are of current interest from both fundamental and application points of view . especially the interference effects in electronic transport attract much attention , as the multi - dot systems offer a unique possibility to study fundamental phenomena which earlier were observed in solid state physics and/or quantum optics . by specific design and fabrication of qd systems , one can investigate for instance the aharonov - bohm oscillations @xcite , fano resonance @xcite , and others . very recently , the dicke effect , which is well known in atomic optics @xcite , has been predicted also for electronic transport through quantum systems @xcite . the key feature of the dicke resonance in optics is the presence of a strong and very narrow spontaneous emission line ( in addition to much broader lines ) of a collection of atoms which are separated by a distance smaller than the wavelength of the emitted light @xcite . generally , the narrow ( broad ) line is associated with a state which is weakly ( strongly ) coupled to the electromagnetic field , and such long - lived ( short - lived ) state is called _ subradiant _ ( _ superradiant _ ) mode . in this paper we consider another realization of the dicke - like effect , i.e. the kondo - dicke resonances in electronic transport through a system of three coupled quantum dots , where two side - coupled dots are noninteracting ( vanishing coulomb parameter , @xmath5 ) , while the central dot is coupled to the leads and is in the deep kondo regime . in addition , we also observe behavior of the kondo peak in dos , which resembles the transition from _ subradiant _ to _ superradiant _ mode ( or _ vice versa _ ) in the usual dicke effect . the theoretical approach we use in this paper is based on the slave - boson technique in the mean - field approximation . the paper is organized as follows . in section 2 we describe the model of a three - dot system , and present the corresponding hamiltonian in the slave - boson formulation . the mean - field approach to the problem is described in section 3 . numerical results on the kondo problem are shown and discussed in section 4 for symmetric and asymmetric systems . final conclusions are given in section 5 . in this paper we consider a system of three single - level quantum dots . the whole system is coupled to external leads as shown schematically in fig.1 . the two side dots , labelled as qd1 and qd3 , are coupled to the central dot qd2 _ via _ a direct hopping term . the dot qd2 , in turn , is additionally attached to external electron reservoirs ( see fig.1 ) . moreover , the central dot is assumed to be in the kondo regime , while the side dots are beyond the kondo regime ( for simplicity they are assumed to be noninteracting , @xmath5 ) . the system under consideration can be modelled by the anderson impurity - like hamiltonian of the following form : @xmath6 the first term describes the left ( @xmath7 ) and right ( @xmath8 ) electrodes in the non - interacting quasi - particle approximation , with @xmath9 ( @xmath10 ) being the creation ( annihilation ) operator of an electron with the wave vector @xmath11 and spin @xmath12 in the lead @xmath13 , and with @xmath14 denoting the corresponding single - particle energy . the next three terms of the hamiltonian describe the system of three coupled quantum dots . here , @xmath15 is the energy of the discrete level in the @xmath16-th dot ( @xmath17 ) , while @xmath18 denotes the hopping parameter between the @xmath19-th dot ( @xmath20 ) and the dot qd2 . both @xmath15 and @xmath18 can be spin dependent in a general case . furthermore , the term with @xmath0 describes the intra - dot coulomb interactions in the dot qd2 , with @xmath0 denoting the corresponding coulomb integral . the inter - dot coulomb repulsion , similarly as the intra - dot coulomb interaction for the dots qd1 and qd3 , is assumed to be small and therefore neglected . the last term of the system s hamiltonian describes electron tunnelling from the leads to the dot qd2 ( or _ vice versa _ ) , with @xmath21 denoting the relevant matrix elements ( @xmath22 ) . coupling of the dot qd2 to external leads can be parameterized in terms of @xmath23 . we assume that @xmath24 is constant within the electron band of the leads , @xmath25 for @xmath26 , and @xmath27 otherwise . here , @xmath28 denotes the electron band width . @xmath29 describe coupling of the dot qd2 to the external leads , whereas @xmath18 is the hopping parameter between the dot qd2 and the @xmath19-th dot , @xmath20,scaledwidth=40.0% ] the following considerations are restricted to the limit of infinite intra - dot coulomb parameter for the dot qd2 , @xmath30 , so the double occupancy on the central dot is forbidden ( only one electron can reside in the central dot ) . this assumption allows us to employ the infinite-@xmath0 slave - boson mean - field ( sbmf ) approach @xcite to analyze transport properties . however , one should bear in mind that the sbmf theory is valid only for low bias at zero temperature . in the slave - boson approach , one introduces a set of auxiliary operators for the central dot . the dot s creation ( annihilation ) operator @xmath31 ( @xmath32 is then replaced with @xmath33 ( @xmath34 ) . here , @xmath35 ( @xmath36 ) is the slave - boson ( pseudofermion ) annihilation operator for an empty ( singly occupied with a spin-@xmath12 electron ) state of the dot qd2 . to eliminate non - physical states , the following constraint has to be imposed on the new quasi - particles , @xmath37 equation ( 2 ) is a completeness relation of the hilbert space for the central dot . accordingly , the hamiltonian ( 1 ) can be replaced by an effective hamiltonian , expressed in terms of the auxiliary boson @xmath35 and pseudo - fermion @xmath36 operators as @xmath38 the last term with the lagrange multiplier @xmath39 has been introduced to incorporate the constraint for qd2 , given by eq.(2 ) , which prevents double occupancy of the central dot . apart from this , @xmath40 in eq.(3 ) stands for the spin - degeneracy of the qd2 . one can check that the number operator @xmath41 ( see eq.(2 ) ) commutes with this new hamiltonian , so the total particle number of @xmath42-electrons and slave bosons is conserved . the mean field approximation ( mfa ) adopted to the slave - boson method relies on replacing the boson field @xmath35 by its expectation value . thus , in the lowest order ( in the @xmath43 expansion ) , the slave - boson operator is substituted by a real and independent of time @xmath44 number , @xmath45 . this approximation neglects fluctuations around the average value @xmath46 of the slave boson operator , but takes into account spin fluctuations ( kondo regime ) exactly at @xmath47 and in the limit of @xmath48 . on the other hand , one should bear in mind that mfa can not be applied in the mixed - valence regime , where strong charge fluctuations take place . that is why the mfa approach restricts our considerations to the low bias regime ( @xmath49 ) . introducing the following renormalized parameters @xmath50 , @xmath51 and @xmath52 , the effective mfa hamiltonian can be rewritten as @xmath53 one needs now to find the unknown parameters @xmath54 and @xmath39 . these can be determined from the constraint imposed on the slave bosons ( eq.(2 ) ) , and from the equation of motion for the slave boson operators . as a result , one finds two equations , @xmath55 @xmath56 there is also another way to obtain the above equations . the free parameters @xmath54 and @xmath39 can be determined by minimizing the ground state energy of the effective mfa hamiltonian with respect to these parameters . application of the hellmann - feynman theorem to the hamiltonian ( 4 ) , together with the conditions for minimal energy @xmath57 , gives a set of self - consistent equations . these equations in the fourier space read @xmath58 @xmath59 where @xmath60 and @xmath61 are the fourier transforms of the lesser green functions ; the former being defined as @xmath62 and a similar definition also holds for @xmath63 . to rewrite eq.(8 ) in a different form , we write first the equation of motion for @xmath64 . then , we multiply this equation by @xmath65 and take the thermal average . upon expressing the obtained equation in terms of the lesser green functions , and taking its hermitian conjugation , one finds @xmath66 when substituting eq.(9 ) into eq.(8 ) , one finally arrives at the equation @xmath67 formulas ( 7 ) and ( 10 ) are our final equations for the unknown parameters . to solve them one still needs to determine the lesser green function @xmath68 . this can be derived from the corresponding equation of motion . similarly , the equation of motion can also be used to derive the retarded green function ( which is also required in our calculations ) . as a result one finds @xmath69 @xmath70 } { \left\|\varepsilon-\tilde{\epsilon}_{2\sigma}+i\tilde{\gamma}_{\sigma}- \frac{\tilde{t}_{32\sigma}^2}{\varepsilon-\epsilon_{3\sigma}}- \frac{\tilde{t}_{12\sigma}^2}{\varepsilon-\epsilon_{1\sigma}}\right\|^2},\ ] ] where @xmath71 is the fermi - dirac distribution function in lead @xmath13 , @xmath72 and @xmath73 are the renormalized parameters describing coupling of the dot qd2 to external leads . electric current @xmath74 flowing through the system can be determined from the formula @xmath75t_{\sigma}(\varepsilon),\ ] ] where @xmath76 is the transmission probability . it is worth noting that the above current formula is similar to that used for noninteracting qd systems . in our case , however , the transmission probability depends on the renormalized parameters @xmath77 , @xmath78 and @xmath79 , which depend on the gate and transport voltages . at @xmath80 , the current and linear conductance are given by the formula @xmath81 and @xmath82 in numerical calculations we assume that the dot levels and the hoping parameters are independent of electron spin , @xmath83 ( for @xmath84 ) , @xmath85 for @xmath20 , and @xmath86 ( for @xmath22 ) . the energy levels @xmath87 are measured from the fermi level of the leads in equilibrium ( @xmath88 ) . in the following we set the bare level of the dot qd2 at @xmath89 , and the bandwidth is assumed to be @xmath90 . in this paper all the energy quantities will be expressed in the units of @xmath91 ( @xmath92 ) . the energy levels @xmath93 and @xmath94 can be tuned by applying gate voltages to qd1 and qd3 . taking into account the above parameters , the kondo temperature @xmath2 of the central dot for @xmath95 can be estimated to be @xmath96 ( @xmath97 ) . ) of the interdot coupling strength @xmath98 , and for @xmath99.,scaledwidth=46.0% ] ( @xmath100 ) , and for the interdot coupling @xmath101.,scaledwidth=46.0% ] we start from the local density of states ( ldos ) at equilibrium , @xmath102 , for the central dot qd2 , which can be calculated from the formula @xmath103,\ ] ] where @xmath104 $ ] denotes the imaginary part of @xmath105 . to understand basic features of the ldos @xmath102 in the kondo regime and the influence of side - coupled dots , we consider first the situation when the dot levels are fixed while the coupling between the dots can be changed . the corresponding ldos is shown in fig.2 for several values of @xmath106 . assume first that the central dot is detached from the two dots qd1 and qd3 , @xmath107 ( solid line in fig.2 ) . the peak at the fermi level in @xmath102 reveals then the usual kondo phenomenon in a single dot , which leads to the kondo anomaly in transport through the dot ( to be discussed later ) . if now one additional and noninteracting dot is attached to the kondo dot _ via _ a hopping term , the shape of the kondo peak is changed due to the interference effects ( not shown in fig.2 ) . as it has been shown by wu _ et al _ @xcite , the dos can be then decomposed into a breit - wigner and a fano line shape . assume now that another non - interacting dot has been connected to the kondo dot , as in the case shown in fig.1 . let the bare levels of the side - coupled dots are fixed at @xmath2 and @xmath108 , respectively . when the interdot coupling strength @xmath109 is nonzero , two additional asymmetric satellite peaks emerge in ldos @xmath102 . as shown in fig.2 , these peaks move away from the central peak as @xmath109 increases . at the same time , the widths of the satellite peaks increase , while the central peak becomes narrower . in fig.3 the ldos for the dot qd2 is shown for different energy levels of the side - coupled dots , while coupling of these dots to the central dot is fixed , @xmath101 . it has been assumed there that the energy levels of the dots qd1 and qd3 are located symmetrically with respect to the fermi level at equilibrium , @xmath110 and @xmath111 . for a nonzero @xmath112 , the ldos reveals three well defined peaks . the central peak ( the main kondo peak ) is located at the fermi level of the leads , while the other two satellite peaks are roughly at @xmath113 . when @xmath114 , position of these satellite peaks is only weakly dependent on @xmath112 . between the maxima , the ldos goes to zero at @xmath115 . the width of central peak becomes narrower and narrower with decreasing @xmath112 , and the peak has almost @xmath116-like shape for very small values of @xmath112 , @xmath117 . this means that the kondo state associated with this maximum becomes then more and more localized at the fermi level . in analogy to the dicke effect , which is well known in optics , the narrow central peak in ldos may be considered as a long - lived ( _ subradiant _ ) state , whereas the other two peaks as corresponding to short - lived ( _ superradiant _ ) states . it is interesting to note that one can observe transition from _ subradiant _ to _ superradiant _ mode ( and _ vice versa _ ) by tuning separation of the side - coupled dots levels from the fermi level , e.g. by gate voltages . thus , a particular state , for example the one at the fermi level , is _ subradiant_-like for small values of @xmath112 ( @xmath117 ) , and _ superradiant_-like for large values of @xmath112 ( @xmath118 ) . when the energy levels of the side - coupled dots coincide with the fermi level , @xmath119 , the kondo peak ( central peak ) in ldos is suppressed and disappears . ) .,scaledwidth=46.0% ] for large values of the side dots level separation ( @xmath118 ) , the satellite peaks are narrow , while the central peak is then relatively broad . this situation is qualitatively similar to that observed in ref . however , the width of the central peak is now a complex non - monotonic function of the level separation @xmath112 ( due to self - consistent parameters @xmath54 , @xmath39 ) . the width of the peak increases for small values of @xmath112 achieving a maximum and then decreases very slowly ( even for large values of @xmath112 ) . moreover , in the nonlinear response regime , when a nonzero bias voltage is applied to the system , the width of the central peak is further suppressed , especially for large @xmath112 . position of the maximum moves to smaller values of @xmath112 as bias voltage increases . another new feature is a broad range of very small ldos between the central maximum and the satellite peaks for larger values of @xmath112 . to summarize this section we conclude that the width of the kondo peak in ldos can be changed by tuning both the level position of side dots and the inter dot coupling strength . such systems give us an opportunity to study cross - over from the _ subradiant _ to _ superradiant _ mode , and _ vice versa_. ( measured in the units of @xmath91).,scaledwidth=46.0% ] ) , and indicated energy levels of the side - coupled dots , @xmath120.,scaledwidth=46.0% ] now we consider nonlinear differential conductance @xmath121 . in fig.4 we show differential conductance as a function of the applied voltage for the case when the dot qd1 is coupled to the kondo dot with a constant strength , @xmath122 , while the coupling strength of the dot qd3 is gradually switched on . for comparison we have also shown there the curve for the kondo dot totally decoupled from both side dots , @xmath95 . owing to the kondo resonance , the unitary limit of the conductance is then reached for zero bias limit . this is well know behavior of transport through a single kondo dot . let us now turn on the coupling of the kondo dot to one of the side dots , say to the dot qd1 . the zero bias anomaly becomes then suppressed as a consequence of the interplay between the fano interference and the kondo effect @xcite . for a sufficiently large @xmath123 , the zero bias anomaly may even disappear . however , the situation becomes more complex when the second side dot ( qd3 ) is attached to the kondo dot and starts to play a role in transport . with increasing coupling strength @xmath124 of the dot qd3 to the dot qd2 , one observes revival of the zero bias anomaly peak , which for @xmath125 reaches unitary limit for zero bias . interplay of the dicke and kondo resonances leads , however , to a faster decrease of the conductance at small voltages , and slower at higher voltages . as a consequence , a tail in the conductance survives at higher voltages , where the conductance of the kondo dot detached from the side dots already disappears . this behavior is a consequence of the narrowing of the central kondo peak in ldos and the occurrence of two satellite peaks ( see discussion in the previous subsection ) . this behavior is clearly seen in fig . 5 , where the bias dependence of the differential conductance is shown for several values of the inter - dot coupling parameters , @xmath126 . the zero bias conductance in such a symmetrical case does not depend on the value of @xmath109 , which reflects the ldos behavior ( see fig.2 ) . similar behavior also occurs when the coupling strength @xmath106 are fixed while the level separation of the side - coupled dots is changed ( see fig.6 ) . for large values of @xmath112 the conductance decreases with increasing bias monotonically . however , for small values of @xmath112 the differential conductance is a nonmonotonic function of the applied bias voltage . it falls down rapidly in the region of small bias voltages achieving a minimum value at some voltage , and then rises again . these features are quite reasonable because the satellite peaks move away from the central peak as @xmath112 increases . thus , for large values of @xmath112 the satellite peaks do not contribute to electronic transport . for a small @xmath112 , however , the satellite peaks enter the transport window and contribute to current , leading to an increase in the conductance above a certain bias voltage . it is worth to note , that the zero bias anomaly for @xmath119 is totally suppressed and the differential conductance slowly rises as the bias voltage is applied . this suppression is a consequence of the suppression of the kondo peak in dos due to the interplay of the dicke and kondo resonances . ) .,scaledwidth=48.0% ] the only nonzero contribution to the noise in electronic transport at zero temperature , @xmath47 , is the shot noise , while the thermal noise is then totally suppressed . the quadratic form of the mfa hamiltonian allows us to express the zero frequency shot noise in the form , @xmath127.\ ] ] deviation of the shot noise @xmath128 from the poisson value , @xmath129 , is usually characterized by the fano factor @xmath130 , which for the poissonian noise is exactly equal to 1 . in fig . 7 , the bias dependence of the fano factor is displayed for different interdot coupling parameters @xmath124 , whereas the parameter @xmath123 is fixed at @xmath131 . for the t - shape geometry ( one side dot is decoupled from the central dot qd2 , @xmath132 ) , the fano interference plays a significant role in transport processes , and as discussed above leads to suppression of the zero bias anomaly below the unitary limit . this is also manifested in the shot noise characteristics . the zero bias limit of the fano factor is larger than zero , and this behavior is a result of the weakening of the kondo peak at the fermi level due to the fano interference . with increasing @xmath123 , but still having @xmath132 , the fano interference outweighs the kondo effect and the fano factor @xmath133 tends to unitary limit ( shot noise tends to the poissonian value ) . when coupling of the dot qd3 to the dot qd2 is turned on , the zero bias fano factor drops and finally for a symmetric system , @xmath134 , reaches zero as for the single quantum dot . as a result , the system is ideally transparent for electrons at the fermi energy @xmath135 ( at the equilibrium ) . ( measured in the units of @xmath91).,scaledwidth=48.0% ] figure 8 shows the fano factor as a function of the applied voltage for various inter - dot couplings , @xmath126 . applied bias voltage partially suppresses the kondo effect , which results in the growth of the fano factor . this growth is faster for smaller values of @xmath109 . , @xmath111 , and for @xmath136.,scaledwidth=48.0% ] for different values of @xmath112 ; @xmath110 , @xmath111 . the bias is fixed at @xmath137.,scaledwidth=48.0% ] when the levels of the side coupled dots are located at the fermi level ( @xmath119 ) , the zero bias fano factor achieves the poissonian value and then slowly falls down with increasing bias voltage ( see fig.9 ) . in contrary , the zero bias limit of the fano factor for a nonzero level separation of the side - coupled dots is zero and tends to the poissonian limit with increasing bias voltage . one can then note a sudden increase of the fano factor for a small value of the level separation , @xmath138 , while for @xmath139 the fano factor remains small over a wider range of the bias voltage before it grows to the poisson limit . in fig.10 the fano factor @xmath133 is plotted as a function of @xmath109 for a fixed bias voltage @xmath137 , and for different positions of the side - coupled dots levels ( @xmath110 , @xmath111 ) . when the bare levels of side - coupled dots are in the transport window , the fano factor is then significantly affected by the interdot coupling . for a very small value of @xmath112 , the fano factor grows rapidly from a small value at zero interdot coupling to its maximum values , and then falls down to half of the poissonian value at sufficiently strong couplings . it is worth to note that the fano factor does not reach the poissonian limit for any @xmath112 , except the case of @xmath119 . in turn , for @xmath140 the fano factor @xmath133 is less sensitive to the interdot coupling strength . the central peak in the ldos is then very broad and gives a large contribution to the differential conductance . on can also note that for a large value of @xmath112 , @xmath133 remains small in a wider range of the interdot coupling before it rises to half of the poisson value . contrary to the case of small @xmath112 , one can now see a minimum in the the fano factor ( for a finite bias ) . thus , the kondo - assisted transport is optimized for @xmath141 . in this paper we have considered the effects due to interplay of the kondo and dicke resonances in electronic transport through a three - dot system . one of the dot ( that attached to the leads ) was in the kondo regime , while the two side dots were out of the kondo regime . using the method based on the slave boson mean field approximation we have calculated local density of states , differential conductance , and shot noise . the results clearly show that the side dots strongly modify the transport characteristics . the kondo peak in the local density of states of the central dot becomes narrower due to the interplay of the kondo and dicke effects . the kondo peak is even totally suppressed when the bare energy levels of the side dots are equal and located at the fermi level . this leads to suppression of the zero bias kondo peak in the differential conductance . the interplay of the kondo an dicke resonances has also a significant impact on the shot noise characteristics . this work , as part of the european science foundation eurocores programme spintra , was supported by funds from the ministry of science and higher education as a research project in years 2006 - 2009 and the ec sixth framework programme , under contract n. eras - ct-2003 - 980409 .	electronic transport through a triple quantum dot system , with only a single dot coupled directly to external leads , is considered theoretically . the model includes coulomb correlations in the central dot , while such correlations in the two side - coupled dots are omitted . the infinite-@xmath0 mean - field slave - boson approach is used to obtain basic transport characteristics in the kondo regime . when tuning position of the side - coupled dots levels , transition from _ subradiant _ to _ superradiant _ like mode ( and _ vice versa _ ) has been found in the spectral function , in analogy to the dicke effect in atomic physics . bias dependence of the differential conductance and zero frequency shot noise is also analysed .
a series of revolutionary technological advances , optical transmission systems have enabled the growth of internet traffic for decades @xcite . most of the huge bandwidth of fiber systems is in use @xcite and the capacity of the optical core network can not keep up with the traffic growth @xcite . the usable bandwidth of an optical communication system with legacy standard single - mode fiber ( smf ) is effectively limited by the loss profile of the fiber and the erbium - doped fiber amplifiers ( edfas ) placed between every span . it is thus of high practical importance to increase the spectral efficiency ( se ) in optical fiber systems . even with new fibers , the transceiver will eventually become a limiting factor in the pursuit of higher se because the practically achievable signal - to - noise ratio ( snr ) can be limited by transceiver electronics @xcite . digital signal processing ( dsp ) techniques that are robust against fiber nonlinearities and also offer sensitivity and se improvements in the linear transmission regime are thus of great interest . a technique that fulfills these requirements and that has been very popular in recent years is _ signal shaping_. there are two types of shaping : geometric and probabilistic . in geometric shaping , a nonuniformly spaced constellation with equiprobable symbols is used , whereas in probabilistic shaping , the constellation is on a uniform grid with differing probabilities per constellation point . both techniques offer an snr gain up to the ultimate shaping gain of 1.53 db for the additive white gaussian noise ( awgn ) channel ( * ? ? ? iv - b ) , ( * ? ? ? viii - a ) . geometric shaping has been used in fiber optics to demonstrate increased se @xcite . probabilistic shaping has attracted considerable attention in fiber optics @xcite . in particular , @xcite use the probabilistic amplitude - shaping scheme of @xcite that allows forward - error correction ( fec ) to be separated almost entirely from shaping by concatenating a distribution matcher @xcite and an off - the - shelf systematic fec encoder . probabilistic shaping offers several advantages over geometric shaping . using the scheme in @xcite , the labeling of the quadrature amplitude modulation ( qam ) symbols can remain an off - the - shelf binary reflected gray code , which gives large achievable information rates ( airs ) for bit - wise decoders and makes exhaustive numerical searching for an optimal labeling obsolete . a further feature of probabilistic shaping that , for fiber - optics , has only been considered in @xcite is that it can yield rate adaptivity , i.e. , the overall coding overhead can be changed without modifying the actual fec . probabilistic shaping also gives larger shaping gains than purely geometric shaping ( * ? ? ? 4.8 ( bottom ) ) for a constellation with a fixed number of points . given these advantages , we restrict our analysis in this work to probabilistic shaping on a symbol - by - symbol basis . shaping over several time slots has been studied theoretically @xcite and is beyond the scope of the present study . in this paper , we extend our previous work on probabilistic shaping for optical back - to - back systems @xcite and investigate the impact of shaping for qam formats on the nonlinear interference ( nli ) of an optical fiber channel with wavelength division multiplexing ( wdm ) . for the analysis , we use a recently developed modulation - dependent gaussian noise ( gn ) model @xcite in addition to full - field split - step fourier method ( ssfm ) simulations . this gn model includes the impact of the channel input on the nli by taking into account higher - order standardized moments of the modulation , which allows us to study the impact of probabilistic shaping on the nli from a theoretical point of view . the contributions of this paper are twofold . firstly , we show that one shaped qam input , optimized for the awgn channel , gives large shaping gains also for a multi - span fiber system . this allows potentially for a simplified implementation of probabilistic shaping because just one input pmf can be used for different fiber parameters . secondly , no significant additional shaping gain is obtained for such a multi - span system with 64qam when the pmf is optimized to the optical fiber channel using a gn model . the relevance of this result is that numerical optimizations of the channel input pmf are shown to be obsolete for many practical long - haul fiber systems . in the following , we review the basic principles of probabilistic shaping . the focus is on airs rather than bit - error ratios after fec . both symbol - wise airs and airs for bit - wise decoding are discussed . for a more detailed comparison , we refer the reader to ( * ? ? ? iii ) , ( * ? ? ? 4),@xcite,@xcite . consider an independent and identically distributed ( iid ) discrete channel input @xmath0 and the corresponding continuous outputs @xmath1 . the channel is described by the channel transition probability density @xmath2 , as shown in the center of fig . [ fig : model ] . the symbol - wise inputs @xmath3 are complex qam symbols that take on values in @xmath4 according to the probability mass function ( pmf ) @xmath5 on @xmath6 . without loss of generality , the channel input is normalized to unit energy , i.e. , @xmath7=1 $ ] . the constellation size @xmath8 is the modulation order and denoted by @xmath9 . unless otherwise stated , we consider qam input that can be decomposed into its constituent one - dimensional ( 1d ) pulse amplitude modulation ( pam ) constellation without loss of information . this means that every qam symbol can be considered as two consecutive pam symbols that represent the real and imaginary parts of the qam symbol . the probability of each two - dimensional ( 2d ) qam constellation is the product of the respective 1d pam probabilities , denoted by @xmath10 . the analysis in this work is conducted with 2d qam symbols ; it is explicitly stated when a 1d input is considered , which is done mainly for the ease of notation and graphical representation . the mutual information ( mi ) between the channel input and output sequences , normalized by the sequence length , is defined as @xmath11,\ ] ] where @xmath12 $ ] denotes expectation and @xmath13 is the marginal distribution of @xmath14 . the mi in is an air for a decoder that uses soft metrics based on @xmath2 . since the optical channel is not known in closed form , we can not directly evaluate . a technique called mismatched decoding @xcite is used in this paper , which gives an air for a decoder that operates with the auxiliary channel @xmath15 instead of the true @xmath2 . in this paper we consider memoryless auxiliary channels of the form @xmath16 which means that , in the context of fiber - optics , all correlations over polarization and time are neglected at the decoder . we assume a fixed auxiliary channel , i.e. , @xmath17 , and restrict the analysis in this paper to 2d circularly symmetric gaussian distributions @xmath18 where @xmath19 is the noise variance of the auxiliary channel , @xmath20 , and @xmath21is complex . for details on the impact of higher - dimensional gaussian auxiliary channels , see @xcite . irrespective of the particular choice of the auxiliary channel , we get a lower bound to @xmath22 by using @xmath23 instead of @xmath2 ( * ? ? ? vi ) , @xmath24 \triangleq { \ensuremath{\text{r}_{\text{sym}}}\xspace } , \end{aligned}\ ] ] where the expectation is taken with respect to @xmath25 , and @xmath26 . the value of @xmath27can be estimated from monte carlo simulations of @xmath28 input - output pairs @xmath29 of the channel as @xmath30 the symbol - wise air @xmath31 is achievable for a decoder that assumes @xmath23 . for the practical bit - interleaved coded modulation schemes @xcite that are also used in fiber - optics @xcite , a bit - wise demapper is followed by a binary decoder , as shown in fig . [ fig : model ] . in this setup , the symbol - wise input @xmath32is considered to consist of @xmath33 bit levels @xmath34 that can be stochastically dependent , and the decoder operates on bit - wise metrics . an air for this bit - metric decoding ( bmd ) scheme is the bmd rate @xcite @xmath35^+ \\ & = \bigl [ { \ensuremath{\mathbb{h}}}({\ensuremath{\boldsymbol{b}}\xspace } ) - \sum_{i=1}^m { \ensuremath{\mathbb{h}}}(b_i\|{\ensuremath{y}\xspace } ) \bigr]^+ , \end{aligned}\ ] ] which is the air considered for the simulations in this work . in , the index @xmath36 indicates the bit level , @xmath37 $ ] denotes entropy and @xmath38^+$ ] is @xmath39 . note that @xmath40is bounded above by the symbol - wise mi , @xmath41 @xcite . the first term of is the sum of the mis of @xmath33 parallel bit - wise channels . the term @xmath42 in corrects for a rate overestimate due to dependent bit levels . for independent bit levels , i.e. , @xmath43 , the term @xmath42 is zero and @xmath40becomes the well - known generalized mutual information calculated with soft metrics that are matched to the channel . we calculate @xmath40 , which is an instance of , in monte carlo simulations of @xmath28 samples as @xmath44\nonumber \\ & - \frac{1}{n } \sum_{k=1}^{n}\sum_{i=1}^m \left[\log_2\left ( 1+e^{(-1)^{b_{k , i } } \lambda_{k , i } } \right)\right ] , \end{aligned}\ ] ] where @xmath45 are the sent bits . the air @xmath40is a function of the soft bit - wise demapper output @xmath46 . these log - likelihood ratios ( llrs ) are computed with the auxiliary channel as @xmath47 where @xmath48 and @xmath49 denote the set of constellation points whose @xmath36^th^ bit is 1 and 0 , respectively . the first term of is the llr from the channel and the second term is the a - priori information . for uniformly distributed input , the a - priori information is 0 . using the 2d gaussian auxiliary channel of , we have @xmath50 these llrs can be computed equivalently in 1d if a symmetric auxiliary channel is chosen , a product labeling is used ( * ? ? ? 2.5.2 ) and @xmath6 is generated from the product of 1d constellations . we search for the input distribution @xmath51 that maximizes @xmath40of , @xmath52\leq 1}{\max}~ { \ensuremath{\text{r}_{\text{bmd}}}\xspace},\ ] ] where the underlying channel is awgn . probabilistic shaping for the nonlinear fiber channel is discussed in sec . [ sec : mismatched_shaping_fiber ] . as the awgn channel is symmetric , the 1d pmfs are also symmetric around the origin , i.e. , @xmath53 which in 2d corresponds to a fixed probability per qam ring . a common optimized input for is to use shaped input distributions from the family of maxwell - boltzmann ( mb ) distributions ( * ? ? ? iv ) , ( * ? ? ? viii - a ) . the method to find an optimized input for a particular snr is discussed in detail in ( * ? ? ? iii - c ) and briefly reviewed in the following paragraph . in bits/4d - sym for uniform qam input ( dashed lines ) and qam with the shaped , snr - dependent mb pmf of ( solid lines ) . the awgn capacity ( dotted line ) is shown as a reference . ] let the positive scalar @xmath54 denote a constellation scaling of @xmath3 with a fixed constellation @xmath6 . furthermore , let the pmf of the input be @xmath55 where @xmath56 is another scaling factor . for each choice of @xmath56 , there exists a scaling @xmath54 that fulfills the average - power constraint @xmath57=1 $ ] . we optimize the scalings @xmath54 and @xmath56 such that @xmath40is maximized while using a distribution from and operating at the channel snr that is defined as @xmath58}{\sigma^2}=\frac{{e_\text{s}}}{n_0}=\frac{1}{n_0 } , \end{aligned}\ ] ] where the 1d signal power @xmath59 is normalized to 1 due to the average - power constraint and @xmath60 is the noise variance per dimension . this optimization can be carried out with efficient algorithms , see ( * ? ? ? iii - c ) . in fig . [ fig : bmd_awgn_uniformshaped ] , @xmath40 in bits per four - dimensional symbol ( bit/4d - sym ) is shown versus the snr of an awgn channel . we choose to plot @xmath40per 4d symbol to have values that are consistent with the dual - polarization airs of sec [ sec : mismatched_shaping_fiber ] . further , only @xmath40is shown as it virtually achieves @xmath27 ( * ? ? ? * table 3 ) , ( * ? ? ? * fig . 1 ) , and @xmath40is the more practical air compared to @xmath27 . the dotted curves represent uniformly distributed input and the solid lines show the @xmath40for qam with optimized mb input . the awgn capacity @xmath61 is given for reference . significant gains are found from probabilistic shaping over uniform input , with sensitivity improvements of up to 0.43 db for 16qam , 0.8 db for 64qam and more than 1 db for 256qam . in order to find the optimized mb input , the snr of the channel over which we transmit , denoted _ channel _ snr , must be known or estimated _ a priori _ at the transmitter . this transmitter - side estimate of the snr is referred to as _ shaping _ snr . in a realistic communication system , it can be difficult to know the channel snr at the transmitter because of varying channel conditions such as the number and properties of co - propagating signals , dsp convergence behavior , and aging of components . hence , shaping without knowledge of the channel snr could simplify the implementation of probabilistic shaping . we will see later that an offset from the shaping snr to the channel snr has a minor effect on @xmath40for the awgn channel if a suitable combination of qam format and shaping snr is used in the proper snr regime . c\|\|c\|c\|\|c\|c\|\|c\|c & * a ) * & * b ) * & * c ) * & * d ) * & * e ) * & * f ) * + @xmath9-qam & 16 & 16 & 64 & 64 & 256 & 256 + .fixed distributions of @xmath9-qam ( with @xmath62 ) that lead to at most 0.1 snr loss compared to the full shaping gain [ cols="^ " , ] = @xmath638 dbm ( solid , crosses ) and @xmath64 = 3 dbm ( dashed , circles ) . although these two power levels have approximately the same @xmath65 , the pmf for 3 dbm is not shaped as much in order to avoid increased nli . ] in fig . [ fig : egn_model_optim_2000 km ] , @xmath40is shown versus @xmath64per channel for different input distributions of 64qam . all results are obtained from the spm - xpm model . the dotted curves show @xmath40for a 1d pmf ( red ) and a 2d pmf ( gray ) , both optimized with the spm - xpm model , and the airs for uniform , mb - shaped input , pmf * d ) * are also included . the two optimized pmfs give identical values of @xmath40 , and their shapes are very similar , as the insets in fig . [ fig : egn_model_optim_2000 km ] show . we conclude that , for the considered system , there is virtually no benefit in using the optimized 2d input . additionally , the mb shaped input gives identical gains to the 1d - optimized input at low power and around the optimal one . it is only in the high - power regime that slightly increased airs are obtained with the optimized input . this indicates that , also for a multi - span fiber channel , the shaping gain is very insensitive to variations in the input distribution , and an optimized input gives shaping gains that are no larger than those with an mb pmf . in fact , it is sufficient for the considered system simply to use the fixed input distribution * d ) * from table [ table : input_dists_perfectshaping ] to effectively obtain the maximum shaping gain . in bit/4d - sym for 64qam versus @xmath64per channel in dbm for the spm - xpm model . the airs for the 1d - optimized input ( red dotted ) , 2d - optimized pmf ( gray dotted ) and for all other shaped inputs lie on top of each other over a wide range of launch powers . inset : the optimized 1d pmf in its 2d representation and the 2d pmf , each for @xmath64of @xmath631.5 dbm . ] in this work , we have studied probabilistic shaping for long - haul optical fiber systems via both numerical simulations and a gn model . we based our analysis on awgn results that show that just two input pmfs from the family of maxwell - boltzmann distributions are sufficient per qam format to realize large shaping gains over a wide range of snrs . we have found that these fixed shaped distributions also represent an excellent choice for applying shaping to a multi - span fiber system . using just one input distribution for 64qam , large shaping gains are reported from transmission distances between 1,400 km to 3,000 km . for a fixed distance of 2,000 km , we have studied the impact of probabilistic shaping with maxwell - boltzmann distributions and other pmfs . the adverse effects of shaping in the presence of modulation - dependent nonlinear effects of a wdm system have been shown to be present . an nli penalty from shaping is found to be relatively minor around the optimal launch power in a multi - span system . this means that , for the considered system , just one input pmf for 64qam effectively gives the maximum shaping gain and an optimization for the fiber channel is not necessary . this could greatly simplify the implementation and design of probabilistic shaping in practical optical fiber systems . we expect similar results for other qam formats such as 16qam or 256qam when they are used in fiber systems that are comparable to the ones in this work . we have also found that the gn model is in excellent agreement with the ssfm results , confirming its accuracy for shaped qam input . for nonlinear fiber links in which the contribution of @xmath66is significant , e.g. , those with in - line dispersion management or single - span links with high power , further optimizations of the shaping scheme can be both beneficial for a large shaping gain and incur low nli . additionally , instead of shaping on a per - symbol basis , constellation shaping over several time slots to exploit the temporal correlations by xpm is an interesting future step to increase se . also , optimizing distributions in four dimensions could be beneficial for highly nonlinear polarization - multiplexed fiber links . the authors would like to thank prof . frank kschischang ( university of toronto ) for encouraging us to use the spm - xpm model to study probabilistic shaping for the nonlinear fiber channel . the authors would also like to thank the anonymous reviewers for their valuable comments that helped to improve the paper . p. bayvel , r. maher , t. xu , g. liga , n. a. shevchenko , d. lavery , a. alvarado , and r. i. killey , `` maximizing the optical network capacity , '' _ philosophical transactions of the royal society of london a _ , vol . 374 , no . 2062 , jan . r. maher , a. alvarado , d. lavery , and p. bayvel , `` modulation order and code rate optimisation for digital coherent transceivers using generalised mutual information , '' in _ proc . european conference and exhibition on optical communication ( ecoc ) _ , valencia , spain , paper mo.3.3.4 , sep . 2015 . u. wachsmann , r. f. h. fisher , and j. b. huber , `` multilevel codes : theoretical concepts and practical design rules , '' _ ieee transactions on information theory _ , vol . 45 , no . 5 , pp . 13611391 , jul . 1999 . i. b. djordjevic , h. g. batshon , l. xu , and t. wang , `` coded polarization - multiplexed iterative polar modulation ( pm - ipm ) for beyond 400 gb / s serial optical transmission , '' in _ proc . optical fiber communication conference ( ofc ) _ , san diego , ca , usa , paper omk2 , mar . h. g. batshon , i. b. djordjevic , l. xu , and t. wang , `` iterative polar quantization based modulation to achieve channel capacity in ultra - high - speed optical communication systems , '' _ ieee photonics journal _ , vol . 2 , no . 4 , pp . 593599 , aug . t. h. lotz , x. liu , s. chandrasekhar , p. j. winzer , h. haunstein , s. randel , s. corteselli , b. zhu , and d. w. peckham , `` coded pdm - ofdm transmission with shaped 256-iterative - polar - modulation achieving 11.15-b / s / hz intrachannel spectral efficiency and 800-km reach , '' _ journal of lightwave technology _ , vol . 31 , no . 4 , pp . 538545 , feb . j. estaran , d. zibar , a. caballero , c. peucheret , and i. t. monroy , `` experimental demonstration of capacity - achieving phase - shifted superposition modulation , '' in _ proc . european conference on optical communications ( ecoc ) _ , london , uk , paper we.4.d.5 , sep . t. liu and i. b. djordjevic , `` multidimensional optimal signal constellation sets and symbol mappings for block - interleaved coded - modulation enabling ultrahigh - speed optical transport , '' _ ieee photonics journal _ , vol . 6 , no . 4 , pp . 114 , aug . 2014 . a. shiner , m. reimer , a. borowiec , s. o. gharan , j. gaudette , p. mehta , d. charlton , k. roberts , and m. osullivan , `` demonstration of an 8-dimensional modulation format with reduced inter - channel nonlinearities in a polarization multiplexed coherent system , '' _ optics express _ , vol . 22 , no . 17 , pp . 2036620374 , aug . 2014 . o. geller , r. dar , m. feder , and m. shtaif , `` a shaping algorithm for mitigating inter - channel nonlinear phase - noise in nonlinear fiber systems , '' _ journal of lightwave technology _ , vol . pp , no . 99 , jun . 2016 . b. p. smith and f. r. kschischang , `` a pragmatic coded modulation scheme for high - spectral - efficiency fiber - optic communications , '' _ journal of lightwave technology _ , vol . 30 , no . 13 , pp . 20472053 , jul . 2012 . m. p. yankov , d. zibar , k. j. larsen , l. p. christensen , and s. forchhammer , `` constellation shaping for fiber - optic channels with qam and high spectral efficiency , '' _ ieee photonics technology letters _ 26 , no . 23 , pp . 24072410 , dec . t. fehenberger , g. bcherer , a. alvarado , and n. hanik , `` ldpc coded modulation with probabilistic shaping for optical fiber systems , '' in _ proc . optical fiber communication conference ( ofc ) _ , los angeles , ca , usa , paper th.2.a.23 , mar . 2015 . f. buchali , g. bcherer , w. idler , l. schmalen , p. schulte , and f. steiner , `` experimental demonstration of capacity increase and rate - adaptation by probabilistically shaped 64-qam , '' in _ proc . european conference and exhibition on optical communication ( ecoc ) _ , valencia , spain , paper pdp.3.4 , sep . 2015 . c. diniz , j. h. junior , a. souza , t. lima , r. lopes , s. rossi , m. garrich , j. d. reis , d. arantes , j. oliveira , and d. a. mello , `` network cost savings enabled by probabilistic shaping in dp-16qam 200-gb / s systems , '' in _ proc . optical fiber communication conference ( ofc ) _ , anaheim , ca , usa , paper tu3f.7 , mar . t. fehenberger , d. lavery , r. maher , a. alvarado , p. bayvel , and n. hanik , `` sensitivity gains by mismatched probabilistic shaping for optical communication systems , '' _ ieee photonics technology letters _ 28 , no . 7 , pp . 786789 , apr . 2016 . f. buchali , f. steiner , g. bcherer , l. schmalen , p. schulte , and w. idler , `` rate adaptation and reach increase by probabilistically shaped 64-qam : an experimental demonstration , '' _ journal of lightwave technology _ , vol . 34 , no . 7 , pp . 15991609 , apr . m. p. yankov , f. da ros , e. p. da silva , s. forchhammer , k. j. larsen , l. k. oxenlwe , m. galili , and d. zibar , `` constellation shaping for wdm systems using 256qam/1024qam with probabilistic optimization , '' mar . 2016 . [ online ] . available : http://arxiv.org/abs/1603.07327 g. bcherer , p. schulte , and f. steiner , `` bandwidth efficient and rate - matched low - density parity - check coded modulation , '' _ ieee transactions on communications _ , vol . 63 , no . 12 , pp . 46514665 , dec . 2015 . r. dar , m. feder , a. mecozzi , and m. shtaif , `` on shaping gain in the nonlinear fiber - optic channel , '' in _ proc . ieee international symposium on information theory ( isit ) _ , honolulu , hi , usa , jun . 2014 . a. ganti , a. lapidoth , and i. e. telatar , `` mismatched decoding revisited : general alphabets , channels with memory , and the wide - band limit , '' _ ieee transactions on information theory _ 46 , no . 7 , pp . 23152328 , nov . 2000 . m. secondini , e. forestieri , and g. prati , `` achievable information rate in nonlinear wdm fiber - optic systems with arbitrary modulation formats and dispersion maps , '' _ journal of lightwave technology _ , vol . 31 , no . 38393852 , dec . 2013 . t. fehenberger , t. a. eriksson , a. alvarado , m. karlsson , e. agrell , and n. hanik , `` improved achievable information rates by optimized four - dimensional demappers in optical transmission experiments , '' in _ proc . optical fiber communication conference ( ofc ) _ , anaheim , ca , usa , paper w1i.4 , mar . t. a. eriksson , t. fehenberger , p. andrekson , m. karlsson , n. hanik , and e. agrell , `` impact of 4d channel distribution on the achievable rates in coherent optical communication experiments , '' _ journal of lightwave technology _ , vol . 34 , no . 9 , pp . 22562266 , may 2016 . d. arnold , h .- a . loeliger , p. vontobel , a. kavcic , and w. zeng , `` simulation - based computation of information rates for channels with memory , '' _ ieee transactions on information theory _ , vol . 52 , no . 8 , pp . 34983508 , aug . 2006 . p. poggiolini , g. bosco , a. carena , v. curri , y. jiang , and f. forghieri , `` the gn - model of fiber non - linear propagation and its applications , '' _ journal of lightwave technology _ , vol . 32 , no . 4 , 694721 , feb . 2014 . r. dar , m. feder , a. mecozzi , and m. shtaif , `` inter - channel nonlinear interference noise in wdm systems : modeling and mitigation , '' _ journal of lightwave technology _ , vol . 33 , no . 5 , pp . 10441053 , mar . 2015 . t. koike - akino , k. kojima , d. s. millar , k. parsons , t. yoshida , and t. sugihara , `` pareto - efficient set of modulation and coding based on rgmi in nonlinear fiber transmissions , '' in _ proc . optical fiber communication conference ( ofc ) _ , anaheim , ca , usa , paper th1d.4 , mar .	different aspects of probabilistic shaping for a multi - span optical communication system are studied . first , a numerical analysis of the additive white gaussian noise ( awgn ) channel investigates the effect of using a small number of input probability mass functions ( pmfs ) for a range of signal - to - noise ratios ( snrs ) , instead of optimizing the constellation shaping for each snr . it is shown that if a small penalty of at most 0.1 db snr to the full shaping gain is acceptable , just two shaped pmfs are required per quadrature amplitude modulation ( qam ) over a large snr range . for a multi - span wavelength division multiplexing ( wdm ) optical fiber system with 64qam input , it is shown that just one pmf is required to achieve large gains over uniform input for distances from 1,400 km to 3,000 km . using recently developed theoretical models that extend the gaussian noise ( gn ) model and full - field split - step simulations , we illustrate the ramifications of probabilistic shaping on the effective snr after fiber propagation . our results show that , for a fixed average optical launch power , a shaping gain is obtained for the noise contributions from fiber amplifiers and modulation - independent nonlinear interference ( nli ) , whereas shaping simultaneously causes a penalty as it leads to an increased nli . however , this nonlinear shaping loss is found to have a relatively minor impact , and optimizing the shaped pmf with a modulation - dependent gn model confirms that the pmf found for awgn is also a good choice for a multi - span fiber system . achievable information rates , bit - wise decoders , gaussian noise models , mutual information , nonlinear fiber channel , probabilistic shaping , wavelength division multiplexing .
dark energy ( de ) seems to play an important role of an agent that drives the present acceleration of the universe with the help of large negative pressure . an effective viscous pressure can also play its role to develop the dynamical history of an expanding universe @xcite-@xcite . it is found @xcite that viscosity effects are viable at low redshifts , which observe negative pressure for the cosmic expansion with suitable viscosity coefficients . in general , the universe inherits dissipative processes @xcite , but perfect fluid is an ideal fluid with zero viscosity . although , perfect fluid is mostly used to model the idealized distribution of matter in the universe . this fluid in equilibrium generates no entropy and no frictional type heat because its dynamics is reversible and without dissipation . the dissipative processes mostly include bulk and shear viscosities . the bulk viscosity is related with an isotropic universe whereas the shear viscosity works with anisotropy of the universe . the cmbr observations indicate an isotropic universe , leading to bulk viscosity where the shear viscosity is neglected @xcite . long before the direct observational evidence through the sn ia data , the indication of a viscosity dominated late epoch of accelerating expansion of the universe was already mentioned @xcite . the origin of the bulk viscosity in a physical system is due to its deviations from the local thermodynamic equilibrium . thus the existence of bulk viscosity may arise the concept of accelerating expansion of the universe due to the collection of those states which are not in thermal equilibrium for a small fraction of time @xcite . these states are the consequence of fluid expansion ( or contraction ) . the system does not have enough time to restore its equilibrium position , hence an effective pressure takes part in restoring the system to its thermal equilibrium . the measurement of this effective pressure is the bulk viscosity which vanishes when it restores its equilibrium @xcite-@xcite . so , it is natural to assume the existence of a bulk viscous coefficient in a more realistic description of the accelerated universe today . physically , the bulk viscosity is considered as an internal friction due to different cooling rates in an expanding gas . its dissipation reduces the effective pressure in an expanding fluid by converting kinetic energy of the particles into heat . thus , it is natural to think of the bulk viscous pressure as one of the possible mechanism that can accelerate the universe today . however , this idea needs a viable mechanism for the origin of the bulk viscosity , although there are many proposed best fit models . many models have been suggested to discuss the vague nature of de . during the last decade , the holographic dark energy ( hde ) , new agegraphic dark energy ( nade ) , their entropy corrected versions and correspondence with other de models have received a lot of attention . the hde model is based on the holographic principle which states that _ the number of degrees of freedom in a bounded system should be finite and has a relationship with the area of its boundary _ @xcite . moreover , in order to reconcile the validity of an effective local quantum field , cohen et al . @xcite provided a relationship between the ultraviolet ( uv ) and the infrared ( ir ) cutoffs on the basis of limit set by the formation of a black hole . this is given by @xcite @xmath1 where constant @xmath2 is used for convenience , @xmath3 is the reduced planck mass and @xmath4 is the ir cutoff . this model has been tested by using different ways of astronomical observations @xcite-@xcite . also , it has been discussed widely in various frameworks such as in the general relativity , modified theories of gravity and extra dimensional theories @xcite-@xcite . the nade model was developed in view of the heisenberg uncertainty principle with general relativity . this model exhibits that de originates from the spacetime and matter field fluctuations in the universe . in this model , the length measure is taken as the conformal time instead of age of the universe and its energy density is @xmath5 where @xmath6 is the conformal time . the causality problem occurs in the usual hde model , while it is avoided here . many people have explored the viability of this model through different observations @xcite-@xcite . another proposal to discuss the accelerating universe is the modified gravity theories @xcite . the @xmath0 gravity is the generalization of teleparallel gravity by replacing the torsion scalar @xmath7 with differentiable function @xmath0 , given by @xmath8 where @xmath9 is the coupling constant and @xmath10 . this leads to second order field equations formed by using weitzenb@xmath11ck connection which has no curvature but only torsion . the equation of state ( eos ) parameter , @xmath12 , is used to explore the cosmic expansion . bengochea and ferraro @xcite tested power - law @xmath0 model for accelerated expansion of the universe . they performed observational viability tests and concluded that this model exhibits radiation , matter and de dominated phases . incorporating exponential model along with power - law model , linder @xcite investigated the expansion of the universe in this theory . he observed that power - law model depends upon its parameter while exponential model acts like cosmological model at high redshift . bamba et al . @xcite discussed the eos parameter for exponential , logarithmic as well as combination of these @xmath0 models and they concluded that the crossing of phantom divide line is observed in combined model only . karami and abdolmaleki @xcite constructed this parameter for hde , nade and their entropy corrected models in the framework of @xmath0 gravity . they found that the universe lies in phantom or quintessence phase for the first two models whereas phantom crossing is achieved in entropy corrected models . sharif and rani @xcite described the graphical representation of k - essence in this modified gravity with the help of eos parameter . some other authors @xcite explored the expansion of the universe with different techniques in @xmath0 gravity . also , the effects of viscous fluid in modified gravity theories @xcite-@xcite are analyzed to display accelerating expansion . in this paper , we construct the viscous eos parameter for different viable de models in the framework of @xmath0 gravity with pressureless matter . for this purpose , we consider a time dependent viscous model with its constant viscous reduction to explore the de era in general fluid . the graphical behavior indicates the acceleration of the universe for suitable viscous coefficients . the scheme of paper is as follows : section * 2 * provides basic formalism and discussion about the field equations of @xmath0 gravity . in section * 3 * , the viscous eos parameter is constructed for different de models . also , we discuss the graphical behavior of this parameter for these models . the last section summarizes the results . the @xmath0 theory of gravity ( as the generalization of the teleparallel gravity ) is uniquely determined by the tetrad field @xmath13 @xcite . it is an orthonormal set of four - vector fields defined on lorentzian manifold . the metric and tetrad fields can be related as @xmath14 where @xmath15 is the minkowski metric for the tangent space . here we use greek alphabets @xmath16 to denote spacetime components while the latin alphabets @xmath17 are used to describe components of tangent space . the non - trivial tetrad field @xmath18 , yielding non - zero torsion , can be written as @xmath19 satisfying the following properties @xmath20 the variation of eq.([1 * ] ) with respect to the tetrad field leads to the following field equations @xcite @xmath21f_{t } + s_{i}~^{\mu\nu}\partial_{\mu}(t ) f_{tt}+\frac{1}{4}h^{\nu}_{i}f=\frac{1}{2}\kappa^{2}h^{\rho}_{i}t^{\nu}_{\rho},\ ] ] where @xmath22 . the torsion scalar is defined as @xmath23 where @xmath24 and torsion tensor @xmath25 are given as follows @xmath26 which are antisymmetric . the energy - momentum tensor for perfect fluid is @xmath27 where @xmath28 is the four - velocity in comoving coordinates , @xmath29 and @xmath30 denote the total energy density and pressure of fluid inside the universe . the flat homogenous and isotropic frw universe is described by @xmath31 where @xmath32 is the scale factor such that @xmath33 in the form of redshift @xmath34 . the corresponding tetrad components are @xcite-@xcite @xmath35 which obviously satisfies eq.([4 ] ) . using eqs.([6 ] ) and ( [ 10 ] ) , the torsion scalar turns out in the form of hubble parameter @xmath36 as @xmath37 . the corresponding modified friedmann equations become @xmath38 for the realistic model , we take viscosity term which introduces the effective pressure in the energy - momentum tensor @xcite , i.e. , @xmath39 defined by @xmath40 here @xmath41 is the time dependent bulk viscosity function . to avoid the violation of the second law of thermodynamics , @xmath42 . the field equations ( [ 13 ] ) and ( [ 14 ] ) may be rewritten as @xmath43 we assume here the pressureless ( dust ) matter , i.e. , @xmath44 and the expressions for torsion contributions @xmath45 and effective pressure become @xmath46 it is noted that if we insert @xmath47 in eq.([17 ] ) with non - viscous case , we arrive at the usual friedmann equations in general relativity . the corresponding viscous eos parameter becomes @xmath48 the phantom and quintessence regions are mostly described with the help of constant eos parameter such as , @xmath49 , which corresponds to the quintessence era whereas phantom era is referred to @xmath50 and the phantom divide line is given by @xmath51 . if we consider a torsion dominated universe , then eq.([13 ] ) reduces to @xmath52 inserting the above value in the energy conservation equation for torsion , it follows that @xmath53 @xmath54 the eos parameter @xmath55 describes a vacuum , phantom dominated or quintessence dominated universe for @xmath56 or @xmath57 respectively for viscous case . for the non - viscous case @xmath58 , these conditions reduce to @xmath59 or @xmath60 . viscous models have interesting insights about the evolution of the expanding universe . here we consider a simple time dependent bulk viscous model as follows @xcite @xmath61 where @xmath62 and @xmath63 are positive coefficients . the cosmological evolution can be explored for different values of these coefficients @xcite-@xcite . this bulk viscosity model is motivated due to the terms involved , i.e. , viscosity is related to the velocity and acceleration which give the phenomenon of scalar expansion in fluid dynamics . the viscous model having constant @xmath64 and velocity term @xmath65 are discussed in @xcite , thus a linear combination of these two with acceleration term @xmath66 may give more physical results . in general , the existence of viscosity coefficients in a fluid is due to the thermodynamic irreversibility of the motion . if the deviation from reversibility is small , the momentum transfer between various parts of the fluid can be taken to be linearly dependent on the velocity derivatives . this case corresponds to the constant viscous model . when viscosity is proportional to hubble parameter the momentum transfer involves second order quantities in the deviation from reversibility leading to more physical results . the proper choices of their coefficients may lead to the crossing of phantom divide line . to determine the evolution of effective eos parameter incorporating @xmath0 and viscous models , we assume the hubble parameter in the form @xcite @xmath67 here @xmath68 and @xmath69 are positive constants , the constant @xmath70 is either positive or negative and @xmath71 is guaranteed for the accelerated expansion of the universe due to the violation of strong energy condition @xmath72 . for @xmath73 , it leads to the scale factor @xmath74 which ends up the universe with future finite time big rip singularity . using eq.([a ] ) with @xmath73 , the torsion scalar becomes @xmath75 with @xmath76 . also , taking the value of @xmath36 , the energy conservation equation , @xmath77 for dust matter yields the solution @xmath78 where @xmath79 is an arbitrary constant . in the following , we discuss three de models by taking into account of the viscosity . first , we consider the following de model @xcite @xmath80 where @xmath81 and @xmath82 are constants . for @xmath83 , this model leads to the teleparallel gravity . it is interesting to note that the model ( [ 29 ] ) is the result of correspondence between energy densities of @xmath0 and hde model . in flat frw universe , the ir cutoff @xmath4 in eq.([1 ] ) becomes the future event horizon @xmath84 resulting the he d energy density . using eq.([a ] ) in the correspondence , @xmath81 takes the form @xmath85 and @xmath82 is an integration constant . replacing @xmath0 and viscous models in eq.([21 ] ) , the viscous eos parameter takes the form @xmath86}{(\kappa^2 \rho_{m0}(1+z)^{3+\frac{2}{h}}+3\alpha h^2)(1+z)^{\frac{1}{h}}}\\\label{31}&-&\frac{2\alpha h}{\kappa^2 \rho_{m0}(1+z)^{3+\frac{2}{h}}+3\alpha h^2}-1.\end{aligned}\ ] ] the graphical behavior of time dependent viscous eos parameter with respect to redshift is shown in * figure 1. we draw this parameter by taking arbitrary values of the coefficients @xmath87 of viscous model , where @xmath81 depends upon the constant @xmath88 which is 0.818 for flat model @xcite . also , we fix the redshift range from 0 to 5 to discuss the behavior of the universe at low redshifts . the left graph in figure 1 * shows the evolution of the universe initially from matter dominated era for higher values of @xmath34 and then converges to quintessence era at @xmath89 for @xmath90 and @xmath91 . the phantom divide line is being crossed by the @xmath92 as @xmath34 approaches to zero . by decreasing @xmath93 and @xmath63 from @xmath94 , the universe remains in phantom dominated era ( shown in the right graph ) . for the constant viscous case , we take @xmath95 in eq.([24 ] ) , thus the constant viscous eos parameter becomes @xmath96 * figure 2 * represents the same behavior as indicated by time dependent viscous eos parameter . however , the phantom crossing for the constant viscous coefficient occurs at @xmath97 , it shows phantom behavior for @xmath98 ( right graph ) . assuming the exponential @xmath0 model @xcite @xmath99 where @xmath100 is an arbitrary constant . inserting @xmath0 and viscous models in eq.([21 ] ) , the viscous eos parameter takes the form @xmath101 \left[2(1+z)^3+\frac{36h^4}{(1+z)^{\frac{4}{h}}}\exp(-\frac{6h^2b}{(1+z)^{\frac{2}{h}}})\right]^{-1}.\end{aligned}\ ] ] figure * 3 * represents the graphical behavior of time dependent viscous @xmath102 versus @xmath34 . in the left graph , the plot shows the evolution of the universe from matter to de phase for higher values of redshift , approximately for @xmath103 . at @xmath104 for particular values @xmath90 and @xmath105 , the eos parameter indicates the quintessence era and approaches to @xmath106 as @xmath107 . as we decrease the values of @xmath108 , the @xmath92 represents the phantom era of the universe . now for constant viscous eos parameter , we take @xmath109 in eq.([d ] ) yields @xmath110 \left[2(1+z)^3\right.\\\label{e}&+&\left.\frac{36h^4}{(1+z)^{\frac{4}{h } } } \exp(-\frac{6h^2b}{(1+z)^{\frac{2}{h}}})\right]^{-1}.\end{aligned}\ ] ] its plot versus @xmath34 is in figure * 4 * , showing same behavior as that of time dependent case . approximately , the universe meets the quintessence era at @xmath111 and converges towards @xmath112 as @xmath34 approaches to zero ( in left graph ) . in right graph , the evolution of eos parameter represents the phantom era of the universe for @xmath113 by decreasing the value of @xmath64 , i.e. , @xmath114 . finally , we take the model @xmath115 which includes linear and nonlinear terms of torsion scalar and @xmath116 are constants . similar to the first model ( [ 29 ] ) , this model comes through the correspondence of nade model with @xmath0 gravity . the energy density of the nade model inherits the conformal time @xmath117 . incorporating the correspondence , here @xmath118 is an integration constant and @xmath70 is @xmath119 where @xmath120 for flat universe . replacing eq.([38 ] ) in ( [ 21 ] ) , the viscous eos parameter becomes @xmath121}{(2\kappa^2 \rho_{m0}(1+z)^{5+\frac{2}{h}}+\gamma ( -6h^2)^{1+h})(1+z)^{\frac{1 - 2h}{h}}}\end{aligned}\ ] ] @xmath122 the graphical behavior of time dependent viscous @xmath92 is given in * figure 5. initially , it shows the deceleration phase @xmath123 of the universe for higher values of @xmath34 . as we decrease the value of redshift up to @xmath124 , it meets the quintessence region for the particular values @xmath125 and @xmath126 , and crossing of the phantom divide line takes place for @xmath34 tends to zero . the right graph indicates that the universe remains in this era for @xmath127 . the constant viscous model for this case is @xmath128 figure 6 * shows its plot versus redshift . it provides the crossing of phantom divide line for a high value @xmath129 , whereas @xmath130 corresponds to the phantom region for decreasing @xmath34 of the accelerating expansion of the universe . viscous models have been discussed in cosmological evolution of the universe as compared to the ideal perfect fluid . the term of shear viscosity vanished when a completely isotropic unverse is assumed and only the bulk viscosity contributes for the accelerating universe to get negative pressure . in this paper , we have considered viscosity by taking dust matter in the framework of @xmath0 gravity . we have taken three different viable de models and a time dependent viscous model to construct the viscous eos parameter for these models . the graphical representation is also developed by considering arbitrary values of the coefficients in viscous model for a specific expression of hubble parameter . the results and the comparison with non - viscous case are given as follows . all the three models in viscous fluid indicates the behavior of the universe from matter dominated phase to quintessence era and then converges to phantom era of the de dominated phase for decreasing @xmath34 . it shows the phantom universe by taking the particular values of viscous coefficients . the constant viscous cases also exhibit phantom behavior . the non - viscous case @xmath131 shows a universe which always stays in phantom for @xmath132 or quintessence for @xmath133 regions @xcite . however , the third model has resulted the phantom phase of the universe for the higher values of viscous coefficients as compared to the first and second @xmath0 models . in each case , the time dependent case shows the phantom crossing by taking small values of viscous coefficients while constant viscous case needs higher values for crossing . the combination of torsion and viscosity influences the accelerating expansion of the universe in such a way that it strictly depends upon the viscous coefficients of the model . we have to fix the ranges for these coefficients in order to get our desired results . we conclude that the viscosity model leads to different behavior of the accelerating universe in de era under the effects of viscous fluid . on the other hand , viscosity may result the crossing of the phantom divide line and phantom dominated universe @xcite as shown in * figures 1 * and * 6*. in the non - viscous case @xcite , the universe remains in the phantom and quintessence eras for the relevant scale factors . beyond the ideal situation , we remark that the de era of the universe in a real fluid may be observed and hence accelerating expansion of the universe is achieved .	we study the bulk viscosity taking dust matter in the generalized teleparallel gravity . we consider different dark energy models in this scenario along with a time dependent viscous model to construct the viscous equation of state parameter for these dark energy models . we discuss the graphical representation of this parameter to investigate the viscosity effects on the accelerating expansion of the universe . it is mentioned here that the behavior of the universe depends upon the viscous coefficients showing the transition from decelerating to accelerating phase . it leads to the crossing of phantom divide line and becomes phantom dominated for specific ranges of these coefficients . * keywords : * @xmath0 gravity ; viscosity ; effective equation of state . + * pacs : * 04.50.kd ; 95.36.+x
the problem of instantaneous wavefunction collapse ( reduction ) due to measurement @xcite has been a stumbling block for many physicists since the creation of quantum mechanics . the unavoidable `` spookiness '' @xcite of the quantum collapse is related to the impossibility to find a traditional physical mechanism responsible for the collapse . mathematically , the `` spookiness '' can be expressed via violation of the bell inequalities @xcite . even though this violation @xcite is common knowledge nowadays , the mechanism and interpretation of the collapse remain debatable @xcite . a natural approach to understanding the physics of the wavefunction reduction is through analysis of the gradual evolution at a shorter time scale , i.e. , `` inside '' the collapse . a few decades ago there was an idea that such an evolution can be fully described by decoherence . however , nowadays it is becoming common knowledge that gradual collapse of individual quantum systems is governed by a continuous flow of information during the measurement , thus showing essentially the same `` spookiness '' as the textbook collapse . this understanding was significantly influenced by experiments with superconducting qubits in the last decade @xcite , which demonstrated the actual evolution `` inside '' the collapse . there are many approaches to the theoretical description of the evolution `` inside '' the collapse , i.e. , the description of partial or continuous quantum measurement . in spite of very different mathematical treatments , many of these approaches are essentially equivalent . probably the most well - known approach is based on positive operator - valued measure ( povm ) and kraus operators @xcite . let us also mention quantum trajectories @xcite , quantum filtering @xcite , monte carlo wavefunction approach @xcite , quantum state diffusion @xcite , restricted path integral @xcite , quantum bayesian formalism @xcite ( see also @xcite and chap . 2.2 of @xcite ) and many other approaches , e.g. , @xcite . among these approaches , one of the simplest formalisms is the quantum bayesian formalism , which is based only on elementary quantum mechanics and common sense . for solid - state systems , the gradual collapse due to continuous measurement was first described using the quantum bayesian formalism @xcite , and soon after that was also described by the quantum trajectory approach @xcite . from late 1990s to mid-2000s the analysis was mainly focused on the continuous measurement of a charge qubit by a quantum point contact ( qpc ) or a single - electron transistor ( set ) @xcite . the next considered system was based on a partially / continuously measured superconducting phase qubit @xcite ; the first experimental demonstration of a partial collapse @xcite and uncollapsing @xcite was realized with this system . after the development of circuit qed qubit measurement @xcite and the transmon @xcite , much attention was paid to this system since it experimentally allowed truly continuous quantum measurement of qubits @xcite . in this measurement setup the qubit state affects the frequency of a coupled resonator , which in turn is probed by an applied microwave in the homodyne way . for the circuit qed measurement of a qubit the quantum trajectory approach was developed in ref . @xcite and the quantum bayesian approach was introduced in ref . in particular , the quantum bayesian theory was used in several circuit qed experiments on quantum feedback and quantum trajectories @xcite , and several experiments used the quantum trajectory theory @xcite . while the description of the qubit evolution in the process of circuit qed measurement is generally similar to that for measurement by qpc or set , there is one considerable difference . the measurement by a qpc or set is of the broad - band type , while the circuit qed measurement is narrow - band . correspondingly , instead of one output signal @xmath0 in the qpc / set case , there are in general two output signals in the circuit qed case , since a narrow - band signal can be represented as @xmath1 , where @xmath2 is the carrier frequency . the existence of two signals ( two quadratures ) leads to the importance of the question of which amplifier is used in the process of measurement . in the case of a phase - sensitive amplifier , only one quadrature is amplified , and therefore only one signal [ say , @xmath0 ] is available . this makes the phase - sensitive case similar to the measurement by qpc or set ( however , it is still important exactly which quadrature is amplified ) . for the case of a phase - preserving amplifier , both output signals @xmath0 and @xmath4 are available ( and both are noisier than in the phase - sensitive case @xcite ) ; this makes the description of qubit evolution significantly different from that for the qpc / set case . an advantage of the quantum bayesian formalism in comparison with the quantum trajectory formalism is its simplicity , so that it does not require special theoretical training , and can be used by non - experts . this simplicity is due to a transparent physical meaning , which directly relates the quantum back - action to the information acquired during measurement . in ref . @xcite the quantum bayesian formalism for the circuit qed measurement of a qubit was developed for the so - called `` bad cavity '' limit , in which the damping ( bandwidth ) @xmath5 of the measurement resonator is much larger than the rate of qubit collapse ( quantum back - action ) due to measurement . in this limit the qubit is practically unentangled with the resonator and experiences two kinds of back - action due to measurement . the `` spooky '' back - action ( which can also be called `` quantum '' , `` informational '' , or `` non - unitary '' @xcite ) moves the qubit state along the meridians of the bloch sphere and is directly related to the continuous information on the qubit state ( @xmath6 or @xmath7 ) obtained during the measurement . this back - action does not have a physical mechanism , similarly to the einstein - podolsky - rosen - bell example @xcite . the other type , the `` phase '' backaction ( called `` realistic '' , `` classical '' or `` unitary '' back - action in @xcite ) has a physical mechanism : fluctuation of the ( ac stark - shifted ) qubit frequency due to a fluctuating number of photons in the resonator . the phase back - action moves the qubit state along the parallels of the bloch sphere . in spite of the clear physical mechanism , the phase back - action also has some `` spookiness '' ; for example , it is possible to choose the qubit movement along the parallels or meridians afterwards , by choosing the amplified microwave quadrature @xcite ( this prediction has been confirmed experimentally @xcite ) . in the present paper we extend the quantum bayesian formalism to the case when the `` bad cavity '' limit is not applicable . as we will see , in this case the evolution equations remain practically the same as in the `` bad cavity '' regime @xcite ; however , now they should be applied to the entangled qubit - resonator system . in the derivation we will assume that the qubit evolves only due to measurement ; in particular , we assume no rabi oscillations . the rabi oscillations can be added later phenomenologically ; however , such addition is not really correct if the rabi frequency is comparable to or larger than the resonator damping @xmath5 . in this respect the theory discussed here has the same limitation as the `` polaron frame approximation '' usually used in the quantum trajectory approach @xcite ( see also @xcite ) . actually , our theory is equivalent to the quantum trajectory theory with this approximation . however , the evolution equations are formally different and have a simple and intuitive physical meaning . we expect that our approach may have advantage over the quantum trajectory theory in numerical simulations , similar to the qpc / set case . ( in the latter case the reason for the numerical advantage was that the quantum trajectory equation is essentially the lowest - order approximation in the time step , while the quantum bayesian evolution is the exact solution in the absence of rabi oscillations , and this permits using larger time steps even in the presence of rabi oscillations . ) our derivation will be based on elementary quantum mechanics . we will also need some basic facts related to coherent states ; for completeness , they are discussed in appendix a. the paper is mainly addressed to non - experts in continuous quantum measurement and non - experts in quantum optics ; this is why we include brief discussion of facts well - known to experts and focus on simple logic . we hope that our derivation is accessible at the advanced - undergraduate level . while we discuss the circuit qed measurement of one qubit , it is straightforward to extend the discussion to the measurement of several qubits , including entanglement by measurement @xcite . the paper is organized in the following way . in sec . ii we discuss the system and the model . in sec . iii we review the results of ref . @xcite for the `` bad cavity '' regime of circuit qed measurement . the main section of this paper is sec . iv , in which we derive the quantum bayesian formalism for circuit qed measurement with a moderate bandwidth . we first introduce a natural idea of `` history tail '' , which consists of the microwave field emitted by the measurement resonator , and thus carries information about the resonator state at previous time moments ( sec . [ sec : history - tail ] ) . then we develop the bayesian formalism by applying a natural measurement procedure to pieces of the `` history tail '' of short duration @xmath8 ( sec . [ sec : main - idea ] ) . the textbook collapse due to this measurement leads to evolution of the entangled qubit - resonator state . we first derive the results for phase - sensitive measurement ( sec . [ sec : imperfect ] ) and then for phase - preserving measurement ( sec . [ sec : ph - pres ] ) . the obtained evolution equations for short @xmath8 are also converted into the differential form ( sec . [ sec : differential ] ) and integrated for an arbitrary long duration ( sec . [ sec : arb - duration ] ) we conclude in sec . v. appendix a reviews basic facts related to coherent states . in appendix b we derive the formulas for the phase back - action in the `` bad cavity '' regime via a simple language based on vacuum noise . we consider a superconducting qubit ( transmon ) measured in the circuit qed setup ( fig . [ fig : schematic ] ) . the idea of the measurement @xcite is based on the dispersive coupling of the qubit with a microwave resonator , whose frequency slightly changes depending on whether the qubit is in the state @xmath6 or @xmath7 ( both are the eigenstates of qubit energy ) . this frequency shift affects the phase and amplitude of a probing microwave , which is transmitted through or reflected from the resonator ( theoretically , there is no significant difference between the transmission and reflection configurations ; however , in practice it is often better to use reflection ) . the outgoing microwave is amplified , and then the ghz - range signal is downconverted by mixing it with the original microwave tone , so that the low - frequency ( @xmath9 mhz ) output of the iq mixer provides information about the qubit state . the rate of the information acquisition is limited by the output noise , which is mainly determined by the first amplifying stage ( pre - amplifier ) . in recent years nearly quantum - limited superconducting parametric amplifiers @xcite became the standard pre - amplifiers , replacing formerly used cryogenic high - electron - mobility transistors ( hemts ) , which have a much higher noise level . is transmitted through ( or reflected from ) the resonator , whose frequency slightly changes , @xmath10 , depending on the qubit state . after amplification , the microwave is sent to the iq mixer , which produces two quadrature signals : @xmath0 and @xmath4 . in the case of a phase - sensitive amplifier we define @xmath0 as the signal corresponding to the amplified quadrature , while for a phase - preserving amplifier we define @xmath0 as the quadrature carrying information about the qubit state.,width=336 ] the hamiltonian of the qubit interacting with the resonator in the dispersive approximation @xcite is h_q&r/=(_q/2 ) _ z + _ + a^a _ z , [ ham - disp]where @xmath11 is the ( effective ) qubit frequency , @xmath12 is the ( effective ) resonator frequency , @xmath13 is the dispersive coupling , @xmath14 and @xmath15 are the creation and annihilation operators for the resonator ( so that @xmath16 is the number of photons in the resonator ) , and the pauli operator @xmath17 acts on the qubit state in the energy basis @xmath7 and @xmath6 . as we see from this hamiltonian , the resonator frequency increases by @xmath18 when the qubit state changes from @xmath6 to @xmath7 ; conversely , the qubit frequency increases by @xmath18 per each additional photon in the resonator ( ac stark shift ) . the typical value of @xmath19 is crudely 1 mhz , while the qubit and resonator frequencies are typically between 4 and 9 ghz , with the detuning @xmath20 of crudely 1 ghz , where @xmath21 . the microwave drive of the resonator can be described by the standard additional hamiltonian h_d/=(t ) e^-i_dt a^+ ^(t ) e^i_dt a , [ ham - drive ] where @xmath22 is the drive frequency and @xmath23 is the properly normalized drive amplitude . [ this form is the rotating wave approximation ( rwa ) of the physical hamiltonian @xmath24(a+a^\dagger ) $ ] . ] we do not consider the case when the resonator is driven by a squeezed microwave or a squeezed vacuum . we assume that the field in the resonator decays with the rate @xmath25 ( so that the energy decays with the rate @xmath5 ) due to coupling with transmission lines and possibly due to other mechanisms of decay ( at zero temperature ) . for the ensemble - averaged evolution , the effect of damping with rate @xmath5 can be described via the standard lindblad term in the master equation ; however , we will not use it , since we are interested in evolution of an individual quantum system rather than an ensemble . note that the derivation of the dispersive hamiltonian ( [ ham - disp ] ) for a transmon is somewhat involved ( see , e.g. , ref . @xcite and appendix of ref . @xcite ) because at least 3 transmon levels should be taken into account to find the coupling @xmath13 ( 4 levels are needed for the lowest - order dependence of @xmath13 on @xmath26 ) . the small-@xmath26 value of the coupling @xmath13 can be approximated @xcite as = , where @xmath27 is the coupling in jaynes - cummings hamiltonian and @xmath28 is the transmon anharmonicity ( @xmath29 is the transition frequency between transmon levels @xmath7 and @xmath30 ) . with increasing @xmath26 the value of @xmath13 changes ( as well as @xmath12 ) , and a better description of the evolution should be based on the eigenlevels of the transmon - resonator system , rather than bare levels @xcite . in the present paper we do not take these complications into account and use the simple hamiltonian ( [ ham - disp ] ) ; however , there is a natural way to include these effects into our formalism phenomenologically . one more subtlety is that the resonator damping @xmath5 leads to the qubit energy relaxation @xcite via the purcell effect , which we do not take into account . however , in many present - day experiments this effect is suppressed by a purcell filter @xcite , so description by the simple hamiltonian ( [ ham - disp ] ) again becomes a good approximation . in this paper we will be using the rotating frame , based on the drive frequency @xmath22 for the resonator and the frequency @xmath11 for the qubit . this essentially means that instead of fast - oscillating coefficients in the lab - frame wavefunction @xmath31 ( here @xmath26 is the number of photons in the resonator ) , we implicitly operate with slower - varying coefficients @xmath32 and @xmath33 . equivalently , we can change the hamiltonian ( [ ham - disp ] ) and ( [ ham - drive ] ) to the rotating - frame hamiltonian h_rot/= ( _ r-_d ) a^a + a^a _ z + a^+ ^ a. [ ham-2]note that in appendix a we use tilde signs for the rotating - frame variables , which are omitted in the main text . our goal in this paper is to find ( in a simple way ) the evolution of the qubit - resonator state in the process of measurement . for that we assume that the qubit evolves only due to measurement , so we explicitly assume the absence of a rabi drive applied to the qubit and absence of qubit energy relaxation . since the hamiltonian ( [ ham - disp ] ) is of the quantum non - demolition ( qnd ) type @xcite , then if the initial qubit state is @xmath6 , it will remain @xmath6 during the whole measurement process . similarly , the initial qubit state @xmath7 will remain @xmath7 . in these two simple cases , evolution of the resonator state is decoupled from the qubit , but the effective resonator frequency @xmath34 depends on the qubit state ( the upper sign is for the qubit state @xmath7 ) . then the classical evolution of the resonator field @xmath35 can be described in the standard rwa way as @xcite _ = -i ( _ r -_d ) _ - _ -i , [ alpha - dot1]where the rotating frame is based on the drive frequency @xmath22 . the quantum evolution is described by exactly the same equation @xcite , with the classical field state replaced by the coherent state @xmath36 ( see appendix a ) . note that @xmath37 is the average number of photons in the resonator . besides the notation @xmath38 , we will interchangeably use a notation that explicitly shows the corresponding qubit state , _ 1 _ + , _ 0 _ - . from eq . ( [ alpha - dot1 ] ) , we see that the resonator field depends on the qubit state . in particular , the steady state is _ , [ alpha - st](it is easy to see that these complex numbers always belong to the circle in the complex plane , which is centered at @xmath39 and passes through the origin . ) the outgoing field @xmath40 in the transmission and reflection configurations ( fig . [ fig : trans - refl ] ) can be described as @xcite f_trans= , f_refl= + , [ f - out]where @xmath41 is the resonator damping due to coupling with the outgoing transmission line ( @xmath42 ) , and in this normalization @xmath43 is the average number of propagating photons per second . ( note that the phase of @xmath40 can be chosen arbitrarily ; in our choice the coefficient between @xmath40 and @xmath44 is real and positive . ) by combining eqs . ( [ alpha - st ] ) and ( [ f - out ] ) it is easy to see that in the case @xmath45 the reflection configuration operates with smaller fields for the same response ( and therefore larger phase response ) than the transmission configuration , and in this sense it is preferable from the practical point of view . however , for our purposes in this paper the two configurations are equivalent ( the well - defined difference @xmath46 can theoretically be simply subtracted ) . we will implicitly assume the transmission configuration ( without loss of generality ) , while all the results are applicable to both the transmission and reflection configurations . the outgoing microwave field @xmath40 is then amplified ( either in a phase - preserving or a phase - sensitive way ) and sent to the iq mixer ( fig . [ fig : schematic ] ) . a phase - sensitive amplifier amplifies only a certain phase ( quadrature ) @xmath47 of the microwave field and de - amplifies the @xmath48-shifted phase ( orthogonal quadrature ) . for the complex number @xmath40 this means amplification of only a certain direction on the complex plane along @xmath49 . for a faster qubit measurement , the obvious choice is to amplify the quadrature that connects the complex numbers @xmath50 and @xmath51 corresponding to qubit states @xmath6 and @xmath7 . we will consider amplification of an arbitrary quadrature , including this optimal case . the iq mixer produces two low - frequency signals , which correspond to the real and imaginary parts of an amplified @xmath40 ; however it is easy to rotate the axes of the complex plane by using the linear combinations of the two outputs . since only one quadrature is amplified by a phase - sensitive detector , there is no information in the mixer output corresponding to the orthogonal quadrature . therefore , the phase - sensitive amplifier essentially produces _ only one output signal _ after the mixer , which we will call @xmath0 . note that the amplified phase @xmath47 can in principle vary in time ; then we also vary the quadrature , corresponding to @xmath0 . a phase - preserving amplifier equally amplifies any quadrature , so both outputs of the iq mixer are important . ( note that usual non - parametric amplifiers , including hemt , are phase - preserving . ) in this case we will call @xmath0 the linear combination of the outputs corresponding to the quadrature connecting @xmath50 and @xmath51 , so that @xmath0 carries information about the measured qubit state @xmath6 or @xmath7 . the output signal for the orthogonal quadrature will be called @xmath4 ; it does not carry information about the qubit state , but will still be important for producing phase back - action . since @xmath52 and @xmath53 evolve before reaching steady values , we will correspondingly vary the quadratures corresponding to @xmath0 and @xmath4 . the main reason why phase - sensitive amplifiers are often preferred for the qubit measurement is that their quantum limitation for the output noise is twice smaller than that for phase - preserving amplifiers @xcite . the output noise of a phase - sensitive amplifier should exceed the `` half quantum '' , which exactly corresponds to the width of the ground state of the oscillator , representing the amplified field ( so that the energy is @xmath54 ) . in other words , this is the amplified vacuum noise of the coherent state of the field , and the ideal phase - sensitive amplifier does not add its own noise ( the output noise can be smaller if a squeezed state is amplified ) . the output noise power of a phase - preserving amplifier is at least two `` half quanta '' : one comes from the amplified vacuum noise , and one more is added by the amplifier @xcite . as discussed above , the dynamics of the system is very simple when the qubit is either in the state @xmath6 or @xmath7 during the whole measurement process . the goal of this paper is to describe the evolution when the initial qubit state is a superposition @xmath55 or , more generally , an arbitrary density matrix @xmath56 . in this section we review results of ref . @xcite for the `` bad cavity '' limit , which assumes @xmath57 , where @xmath58 is the qubit ensemble dephasing due to measurement , discussed below . in this case we can neglect transient evolution of the resonator state , and there is practically no entanglement between the qubit and the measurement resonator , because the two steady states ( [ alpha - st ] ) are very close to each other , @xmath59 . therefore , the evolution of the qubit state can be considered by itself . it is assumed that parameters of the measurement setup ( @xmath5 , @xmath60 , etc . ) do not change in time . we review here the `` bad cavity '' limit mainly for later comparison with the more general case @xmath61 ; the derivation in the next section does not rely on results discussed in this section . ensemble dephasing of the qubit in the `` bad cavity '' regime is @xcite = 2 i m ( _ 1,st^ * _ 0,st ) = \|_1,st-_0,st\|^2 . [ gamma - bc1]we see that the condition @xmath57 is equivalent to @xmath59 . in the case when @xmath62 , the ensemble dephasing @xcite can be expressed as [ see eq . ( [ alpha - st ] ) ] = , [ gamma - bc2]and the ac stark shift contribution @xmath63 to the effective qubit frequency @xmath64 is @xcite _ q= 2\|n . [ stark - bc]if @xmath65 is comparable to @xmath5 ( but still @xmath66 ) , then eqs . ( [ gamma - bc2 ] ) and ( [ stark - bc ] ) should be modified ( see sec . [ sec : ensemble - aver ] ) , but the bayesian formalism reviewed in this section does not change . evolution of the qubit density matrix @xmath67 due to measurement of an _ arbitrary duration _ @xmath68 can be described by simple equations @xcite @xmath69 , \,\,\,\qquad \label{ph - sens - diag - bc}\\ & & \frac{\rho_{10}(t+\tau)}{\rho_{10}(t)}= \frac{\sqrt{\rho_{11}(t+\tau)\ , \rho_{00}(t+\tau ) } } { \sqrt{\rho_{11}(t)\ , \rho_{00}(t)}}\ , \ , \nonumber \\ & & \hspace{1.7 cm } \times \ , \exp [ -i k \tilde{i}_{\rm m}(\tau)\ , \tau ] \ , e^{-\gamma \tau } \ , e^{-i\delta \omega_{\rm q}\tau } , \qquad \label{ph - sens - off - bc } \end{aligned}\ ] ] where i_m ( ) = i_m ( ) - , i_m()= _ t^t+ i(t ) dt , [ ph - sens - im - bc]so that @xmath70 is the measured output signal @xmath0 averaged over the time interval @xmath71 $ ] , while for @xmath72 we also subtract the mean value @xmath73 , with @xmath74 and @xmath75 being the average output signals , corresponding to the qubit states @xmath6 and @xmath7 . the measurement response is @xmath76 where @xmath77 is the phase difference between the amplified quadrature @xmath47 and the optimal quadrature @xmath78 , which gives the largest response @xmath79 . the variance of @xmath80 due to the amplifier noise is d= , [ d - def - bc]where @xmath81 is the single - sided spectral density of the noise [ for different definitions of the spectral density , eq . ( [ d - def - bc ] ) should be changed correspondingly ] . the phase back - action depends on the coefficient @xmath82 , which equals k= _ d= _ d. [ ph - sens - k - bc ] + the dephasing rate @xmath83 is due to non - ideality of the measurement , = - = - , where @xmath58 is the qubit ensemble dephasing [ see eqs . ( [ gamma - bc1 ] ) and ( [ gamma - bc2 ] ) ] . the quantum efficiency of the measurement process can be introduced in two different ways , = 1-==_amp_col , = ^2 _ d , where @xmath84 ( @xmath85 ) takes into account quantum efficiency @xmath86 of the phase - sensitive amplifier and efficiency @xmath87 of the microwave signal collection , while @xmath88 also includes the effect of choosing a non - optimal quadrature for amplification . here @xmath89 is the ratio of the microwave energy reaching amplifier to the total energy loss by the resonator , so that @xmath90 describes the loss in the transmission line before reaching the amplifier . the amplifier efficiency @xmath91 is the ratio between the output noise @xmath92 of an ideal quantum - limited amplification chain to the actual output noise @xmath81 . the last term in eq . ( [ ph - sens - off - bc ] ) is due to the ac stark shift @xmath63 given by eq . ( [ stark - bc ] ) ( note that in ref . @xcite the rotating frame was already accounting for this term and the equation was written for the conjugate variable @xmath93 ) . note that @xmath94 and therefore @xmath95 . equations ( [ ph - sens - diag - bc])([ph - sens - im - bc ] ) can be used to find the qubit evolution in an experiment by using experimental output signal record @xmath0 ; in numerical simulations @xmath80 can be picked randomly from the probability density distribution @xmath96 ^ 2}{2d } \bigg ] , \qquad \label{ph - sens - gauss}\end{aligned}\ ] ] where @xmath97 and @xmath98 is the standard gaussian distribution in the case when the qubit is in the state @xmath99 . for an infinitesimally small averaging time @xmath68 , this is equivalent to using i(t)= + [ _ 11 ( t)-_00(t ) ] + _ i(t ) , s__i = s_i , [ ph - sens - i(t)]where @xmath100 is the white noise with spectral density @xmath81 . the qubit evolution equations ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) have a very simple physical meaning . the evolution ( [ ph - sens - diag - bc ] ) of the diagonal matrix elements of the density matrix is the classical bayesian update for the probabilities , _ jj(t+ ) = , where @xmath101 is the likelihood , given by eq . ( [ ph - sens - gauss ] ) . note that another form of eq . ( [ ph - sens - diag - bc ] ) in terms of the non - centered measurement result @xmath70 is = . the evolution ( [ ph - sens - off - bc ] ) of the off - diagonal matrix element contains the natural term due to change of the diagonal elements ( conservation of relative purity ) , the phase back - action term , decoherence due to non - ideality , and contribution from the ac stark shift . the phase back - action has a natural mechanism : when measuring a non - optimal quadrature , @xmath102 , the output signal gives us information about the fluctuating number of photons in the resonator , and therefore the fluctuating ac stark shift . the factor @xmath82 in eq . ( [ ph - sens - off - bc ] ) is the coefficient characterising this linear relation between the ac stark shift and output signal fluctuations . the evolution equations ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) have been derived in ref . @xcite in the following way . the bayesian evolution ( [ ph - sens - diag - bc ] ) of the diagonal matrix elements was essentially postulated from the necessary correspondence between the classical and quantum evolution . this follows from common sense as much as the standard collapse postulate in quantum mechanics . for the off - diagonal elements , the logic of the derivation ( sketched below ) was essentially the same as in the first derivation @xcite for measurement by a qpc . using the general inequality @xmath103 and evolution ( [ ph - sens - diag - bc ] ) for the diagonal elements , it is easy to derive inequality for the ensemble dephasing , @xmath104 . in the quantum - limited case [ in this case @xmath105 is resolved with signal - to - noise ratio of 1 after time @xmath106 , and therefore @xmath107/(32\chi^2\bar{n})\,$ ] ] , and for @xmath108 , the lower bound of this inequality for @xmath58 coincides with the actual value ( [ gamma - bc2 ] ) . therefore , in this case the evolution of @xmath109 should be precisely the first term in eq . ( [ ph - sens - off - bc ] ) and possibly a result - independent phase ( which is naturally associated with the qubit frequency shift in the last term ) ; otherwise the ensemble dephasing would be larger than in eq . ( [ gamma - bc2 ] ) . thus , in the ideal case eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) have been derived `` logically '' , by comparing unavoidable evolution due to acquired information with the ensemble dephasing . in the non - optimal case ( @xmath102 ) , the derivation in ref . @xcite took into account the phase back - action by explicitly analyzing the information on the fluctuating photon number in the resonator provided by the measurement result @xmath110 . in this way eq ( [ ph - sens - k - bc ] ) for the correlation factor @xmath82 was obtained , leading to the term with @xmath82 in eq . ( [ ph - sens - off - bc ] ) . finally , the term @xmath111 in eq . ( [ ph - sens - off - bc ] ) was obtained by averaging over the extra noise from a non - ideal amplifier @xcite and averaging over the signal that was lost due to imperfect microwave collection . this is how the qubit evolution equations ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) have been obtained in ref . @xcite . actually , the derivation for the phase back - action coefficient @xmath82 in ref . @xcite was presented only for the case of resonant microwave frequency , @xmath112 . in appendix b we show the derivation , which is still valid in the case of a significant detuning , @xmath113 . this derivation is based on an analysis of the effect of vacuum noise entering the resonator from the transmission line . in this analysis the vacuum noise is treated essentially classically , consistent with the poisson statistics @xmath114 for the photon number . note that averaging of the evolution equations ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) over random @xmath110 with the probability distribution ( [ ph - sens - p(i ) ] ) produces ensemble - averaged equations @xmath115 in which there is no dependence on the measured phase @xmath77 ( as required by causality ) because + = = - . let us briefly discuss the role of the `` weak response '' condition @xmath116 in the formalism reviewed in this section . in the case of not too small a number of photons in the resonator , @xmath117 , this inequality follows from the `` bad cavity '' condition @xmath59 , and therefore is not needed as an additional condition . however , for @xmath118 it is possible to have @xmath59 even when @xmath119 . in this case the bayesian formalism ( [ ph - sens - diag - bc])([ph - sens - gauss ] ) is still applicable , but the ensemble dephasing @xmath58 and ac stark shift @xmath120 are not necessarily given by eqs . ( [ gamma - bc2 ] ) and ( [ stark - bc ] ) , in particular , because @xmath121 and @xmath122 may be significantly different , @xmath123 . the formulas for @xmath58 and @xmath120 in this case are given in ref . @xcite and also derived in sec . [ sec : ensemble - aver ] [ @xmath58 is given by eq . ( [ gamma - bc1 ] ) , while @xmath63 is given by eq . ( [ delta - omega - sum ] ) ] . note that the bayesian evolution equations ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) are exactly the same as for the continuous qubit measurement by qpc or set @xcite . however , the dependence ( [ ph - sens - deltai - bc ] ) and ( [ ph - sens - k - bc ] ) of the response @xmath124 and phase back - action coefficient @xmath82 on the measured quadrature @xmath77 is a specific feature of the circuit qed ( or cavity qed ) setup . as was discussed in sec . ii , in the case of a phase - preserving amplifier we choose @xmath0 to be the output signal , corresponding to the optimal quadrature @xmath78 , while the output @xmath4 corresponds to the orthogonal quadrature @xmath125 . therefore , @xmath108 for @xmath0 and @xmath126 for @xmath4 . the qubit state evolution due to a phase - preserving measurement for an _ arbitrary duration _ @xmath68 is described by equations @xcite @xmath69 , \,\,\,\qquad \label{ph - pres - diag - bc}\\ & & \frac{\rho_{10}(t+\tau)}{\rho_{10}(t)}= \frac{\sqrt{\rho_{11}(t+\tau)\ , \rho_{00}(t+\tau ) } } { \sqrt{\rho_{11}(t)\ , \rho_{00}(t)}}\ , \ , \nonumber \\ & & \hspace{1.7 cm } \times \ , \exp [ -i k \tilde{q}_{\rm m}(\tau)\ , \tau ] \ , e^{-\gamma \tau } \ , e^{-i\delta \omega_{\rm q}\tau } , \qquad \label{ph - pres - off - bc } \end{aligned}\ ] ] which have exactly the same form as eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) , except @xmath127 in eq . ( [ ph - sens - off - bc ] ) is replaced by @xmath128 in eq . ( [ ph - pres - off - bc ] ) . the measurement result @xmath127 is given by eq . ( [ ph - sens - im - bc ] ) , and similarly q_m ( ) = _ t^t+ q(t ) dt - q_0 , [ ph - pres - qm - bc]with equal average values , @xmath129 , for the two qubit states . since @xmath0 and @xmath4 are equally amplified , the variances of @xmath127 and @xmath128 due to amplifier noise are both equal to d= , s_q = s_i . the phase back - action is now caused by @xmath130 , and the coefficient @xmath82 is the same as in eq . ( [ ph - sens - k - bc ] ) for @xmath126 , k== , i = i_1-i_0 . [ ph - pres - k]the dephasing rate @xmath83 due to non - ideality is = - 2 = - , where ensemble dephasing @xmath58 is still given by eqs . ( [ gamma - bc1 ] ) and ( [ gamma - bc2 ] ) ( it can not depend on the detector because of causality ) , and the extra factor of 2 is related to equal contributions due to fluctuations of @xmath0 and @xmath4 . the quantum efficiency can again be defined in two different ways , = 1- = _ amp_col , = = _ amp_col , [ eta - phase - pres - def]where @xmath84 ( @xmath131 ) compares the measurement with the ideal phase - preserving case , while @xmath88 and @xmath132 compare the operation using only @xmath0 channel with the ideal phase - sensitive case ( @xmath133 because of twice larger noise in an ideal phase - preserving amplifier ) . equations ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) describe the qubit evolution when the signals @xmath0 and @xmath4 are obtained from an experiment , while in numerical simulations @xmath127 can be generated using eqs . ( [ ph - sens - p(i ) ] ) and ( [ ph - sens - gauss ] ) , while @xmath128 can be picked from the gaussian probability distribution p(_m ) = . for infinitesimally small @xmath68 , the signal @xmath0 can also be generated using eq . ( [ ph - sens - i(t ) ] ) , and for @xmath4 we can use q(t)=q_0 + _ q ( t ) , s__q= s_q= s_i , with equal spectral densities , @xmath134 , of uncorrelated noise in @xmath0 and @xmath4 channels . equations ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) have been derived in ref . @xcite in three different ways , leading to the same result . in the first derivation , eq . ( [ ph - pres - diag - bc ] ) has been again postulated from the necessary correspondence with classical evolution of probability , and the phase back - action coefficient @xmath82 in eq . ( [ ph - pres - k ] ) has been calculated from information on fluctuation of photon number , provided by @xmath4 . this gives the inequality @xmath135 , whose lower bound in the ideal case coincides with the actual value ( [ gamma - bc2 ] ) . thus in the ideal case eqs . ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) can be derived `` logically '' , while the non - ideal case ( @xmath136 ) can be analyzed by averaging over the extra noise of the amplifier ( @xmath137 ) and over information contained in the lost fraction of the microwave signal ( @xmath138 ) . in the second derivation @xcite , eqs . ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) have been obtained from eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) by considering a phase - preserving amplifier as a phase - sensitive amplifier with rapidly rotating amplified phase @xmath47 , so that the difference @xmath77 from the optimal phase is also changing . then averaging the evolution in eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) over the period of phase rotation , we obtain eqs . ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) . finally , the third derivation in ref . @xcite has been based on considering a phase - preserving amplifier as two phase - sensitive amplifiers , which amplify orthogonal quadratures in two microwave channels , obtained from the microwave signal by using a symmetric beam splitter . then using eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) for each channel , we again obtain eqs . ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) . note that since in the `` bad cavity '' regime the qubit is practically not entangled with the measurement resonator , it is easy to include the qubit evolution due to rabi oscillations , energy relaxation , etc . ( this extra evolution should be much slower than @xmath5 , but can be slower , comparable , or faster than @xmath58 ) . for that we need to take the derivative of the evolution equations ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) for the phase - sensitive case or eqs . ( [ ph - pres - diag - bc ] ) and ( [ ph - pres - off - bc ] ) for the phase - preserving case , and simply add the terms due to other mechanisms of evolution ( this is equivalent to interleaving the both types of evolution ) . as always @xcite , in taking the derivative it is important to specify whether the it or stratonovich definition of the derivative is used . now let us discuss the main subject of this paper : the bayesian formalism for continuous qubit measurement in the circuit qed setup ( fig . [ fig : schematic ] ) in the case when the `` bad cavity '' limit is not applicable . therefore , we now assume that the resonator damping rate @xmath5 is comparable to the speed of the qubit evolution due to measurement back - action , which can be characterized by the qubit ensemble dephasing @xmath58 . in this case there is significant entanglement between the qubit and resonator , so we should consider the evolution of the combined qubit - resonator system . also , since the typical measurement time is comparable to @xmath139 , our formalism should focus on the transient evolution . the parameters of the measurement setup ( @xmath60 , @xmath47 , @xmath5 , etc . ) are allowed to change in time ( this change should be much slower than @xmath22 for rwa to be valid , but can be comparable to or even faster than @xmath5 ) . in general , this problem is rather complicated , but we use a simplifying assumption : we assume that the qubit evolution is only due to measurement , i.e. , there are no rabi oscillations , qubit energy relaxation , etc . in practice this means that the frequency of rabi oscillations and rate of energy relaxation should be much smaller than @xmath5 ( then the extra evolution can be added phenomenologically , as discussed in the previous section ) . we will assume that the initial state of the qubit - resonator system is unentangled , and the resonator starts in a coherent state @xmath140 , \|(0)= ( c_0 \|0+c_1 \|1)\|_in , [ psi - in]where @xmath141 and , for example , @xmath140 is vacuum . generalization to a mixed initial state , ^q&r ( 0 ) = _ q , in \|_in_in\| , [ rho - in]or a slightly more general state [ see eq . ( [ rho - qr - def ] ) below ] will be straightforward . in the analysis we will use the rotating frame , corresponding to the hamiltonian ( [ ham-2 ] ) . we will first discuss a simple general point of view , which describes the evolution due to measurement , then derive equations for the ensemble - averaged evolution , and then discuss the evolution during an individual realization of the measurement process . until sec . [ sec : imperfect ] we will assume the ideal case , in particular @xmath142 see fig . [ fig : trans - refl](a ) , in which we need to assume @xmath143 . suppose the initial state of the qubit is @xmath6 . since the measurement is of the qnd type and the qubit does not evolve by itself , it will remain in the state @xmath6 , and since the resonator is initially in a coherent state , its state will remain to be an ( evolving ) coherent state ( see appendix a ) . therefore , the qubit - resonator system will evolve in the rotating frame as \|(t)=\|0 e^-i_0 ( t ) \|_0 ( t ) , where the coherent state amplitude @xmath52 and the overall phase @xmath144 evolve according to eqs . ( [ tilde - alpha - dot ] ) and ( [ phi - dot-2 ] ) with the resonator frequency @xmath145 and drive ( rotating frame ) frequency @xmath22 , @xmath146 here the drive amplitude @xmath60 can be time - dependent and the damping @xmath5 can in general be also time - dependent . note that the evolution ( [ phi0-dot ] ) of the overall phase is often not considered in textbooks , but for us it is very important . derivation of eq . ( [ phi0-dot ] ) from the schrdinger equation with the hamiltonian ( [ ham-2 ] ) is very simple . now let us consider a larger physical system , which includes the field leaking from the resonator to the transmission line ( actually , we also necessarily need to consider the incoming field from the transmission line , but we assume that it is always vacuum ) . this larger system keeps a record of the previous evolution in a form of a `` flying away tail '' ( a propagating microwave ) , which we will call a `` history tail '' ( fig . [ fig : tail ] ) . let us divide this tail into sufficiently short pieces of duration @xmath8 ( @xmath147 ) ; each of them will also be a coherent state , as follows from the property 2.6 discussed in appendix a for a beam splitter ( in our case a leaking `` mirror '' at the end of the resonator ) . the @xmath148th piece of the history tail ( counting back in time ) will be @xmath149 , which is the resonator state at time @xmath150 , passed through the beam splitter with transmission amplitude @xmath151 [ this value follows from the energy conservation , @xmath152 . therefore , the wavefunction , including the history tail is \|(t)= e^-i_0 ( t ) \|0 \|_0 ( t)_m \|_0 ( t - m t ) . [ evol-0]if @xmath5 depends on time , then the factor @xmath153 in this equation should be replaced with @xmath154 . note that the coherent states in the tail are unentangled with each other and with the resonator state ( see property 2.6 in appendix a ) . also note that the number of terms in the direct product ( [ evol-0 ] ) increases with time ; this seems unphysical , but it is only a matter of notation ; we can keep the number of terms constant by adding vacuum states of the pieces of field incoming from the transmission line . if the qubit is initially in the state @xmath7 , then it remains in @xmath7 , so that the wavefunction of the system including the history tail is given by eq . ( [ evol-0 ] ) with @xmath6 replaced with @xmath7 , @xmath50 replaced with @xmath51 , and @xmath155 replaced with @xmath156 , where @xmath157 and @xmath158 evolve according to eqs . ( [ tilde - alpha - dot ] ) and ( [ phi - dot-2 ] ) with the resonator frequency @xmath159 , @xmath160 now let us make a simple but very important logical step in the derivation . since the qubit does not evolve by itself , we can consider two evolutions at the same time : for the qubit in the state @xmath6 and in the state @xmath7 , so that the coefficients in the initial superposition ( [ psi - in ] ) do not change in time ( fig . [ fig : tail ] ) . this follows from the general linearity of quantum mechanics and somewhat resembles the `` many worlds '' interpretation . therefore , for the initial state ( [ psi - in ] ) of the qubit - resonator system we obtain the wavefunction evolution ( including the flying away history tail ) @xmath161 where @xmath162 and @xmath163 are constant in time , while @xmath164 , @xmath155 , @xmath51 , and @xmath156 evolve according to eqs . ( [ alpha0-dot ] ) , ( [ phi0-dot ] ) , ( [ alpha1-dot ] ) and ( [ phi1-dot ] ) , starting with @xmath165 and @xmath166 . the approach to qubit measurement via the wavefunction evolution in eq . ( [ evol-01 ] ) is physically transparent and quite powerful . in particular , it will easily allow us to describe evolution of the qubit - resonator system in the process of measurement by applying the textbook collapse postulate to measurement of the tail pieces ( fig . [ fig : tail ] ) . however , let us first discuss the ensemble - averaged evolution . if the result of the tail measurement is not taken into account , we need to average the quantum state over all possible measurement results , which is equivalent to tracing the state ( [ evol-01 ] ) over the tail . this leads to a density operator in the qubit - resonator hilbert space , @xmath167 , in which the parts diagonal in the qubit subspace are @xmath168 while the off - diagonal parts contain the inner product of the tails for the two different evolutions , @xmath169}\ , \|1\rangle \langle 0 \| \otimes ( t)\rangle \langle \alpha_0 ( t)\| \nonumber \\ & & \hspace{0.1 cm } \times \prod_m \langle \alpha_0 ( t - m\ , \delta t ) \sqrt{\kappa \ , \delta t } \ , \|\ , \alpha_1 ( t - m\ , \delta t ) \sqrt{\kappa \ , \delta t } \rangle , \qquad \label{rho-10-qr-1}\end{aligned}\ ] ] and similarly for @xmath170 . the inner product for each time piece @xmath8 is given by eq . ( [ inner - product ] ) in appendix a , so that we find @xmath171}\ , \|1\rangle \langle 0 \| \otimes ( t)\rangle \langle \alpha_0 ( t)\| \ , \nonumber \\ & & \hspace{1 cm } \times \exp \left ( - \int_0^t \frac{\kappa}{2}\,\|\alpha_1(t')-\alpha_0(t')\|^2 \ , dt ' \right ) \nonumber \\ & & \hspace{1 cm } \times \exp \left ( -i \int_0^t \kappa\ , { \rm im}\ , [ \alpha_1^(t ' ) \ , \alpha_0 ( t')]\ , dt ' \right ) . \qquad \label{rho-10-qr}\end{aligned}\ ] ] in this equation the second line obviously describes dephasing with the rate _ d(t ) = \|_1(t)-_0(t)\|^2 , [ gamma - d]which is directly related to distinguishability of the field emitted into the transmission line and therefore to the information that can in principle be obtained from measurement . the third line in eq.([rho-10-qr ] ) is the changing phase factor which can be ascribed to the shift of the qubit frequency in the process of measurement , _ q,1(t ) = i m [ _ 1^ ( t ) _ 0 ( t ) ] . [ delta - omega-1]however , a similar frequency shift comes from the term @xmath172}$ ] in eq . ( [ rho-10-qr ] ) ; using eqs.([phi0-dot ] ) and ( [ phi1-dot ] ) , we obtain the corresponding value _ q,2(t ) = re \ { ^(t ) [ _ 1 ( t)- _ 0 ( t ) ] } , [ delta - omega-2]so that the total frequency shift of the qubit ( which can be called the ac stark shift ) is _ q , s = _ q,1 + _ q,2 . [ delta - omega-12 ] thus , the ensemble - averaged evolution of the qubit - resonator state can be described ( neglecting the overall phase ) by the wavefunction \|(t)= ( t)+ e^-i _ 0^t _ q , s(t ) dt c_1 \|1 \|_1 ( t ) , subjected to dephasing @xmath173 between the two components . if we also want to trace the state over the resonator , then we have an additional inner product @xmath174 , which changes the dephasing rate ( [ gamma - d ] ) by ( t ) = [ delta - gamma](this change can be positive or negative ) and introduces additional contribution to the qubit frequency shift , _ q,3(t)= i m [ _ 1^(t)_0(t ) ] , [ delta - omega-3]as follows from eq . ( [ inner - product ] ) . the _ qubit - only _ density matrix elements then become @xmath175 \ , dt'\bigg ] \nonumber \\ & & \hspace{0.9 cm } \times \ , \exp \bigg [ -i\int_0^t [ \delta\omega_{\rm q , s}(t')+\delta\omega_{\rm q,3}(t ' ) ] \ , dt ' \bigg ] . \qquad \end{aligned}\ ] ] [ note that @xmath176 are numbers , while @xmath177 in eqs . ( [ rho-00-qr])([rho-10-qr ] ) are operators . ] using eqs . ( [ alpha0-dot ] ) and ( [ alpha1-dot ] ) , it is easy to show that @xmath178 , \label{gamma - sum}\\ & & \delta\omega_{\rm q , s}(t)+\delta\omega_{\rm q,3}(t)= 2\chi \ , { \rm re } [ \alpha_1^ * ( t ) \ , \alpha_0(t ) ] , \qquad \label{delta - omega - sum}\end{aligned}\ ] ] which coincide with the results of refs . @xcite for the qubit dephasing and ac stark shift . note that the results ( [ gamma - d ] ) and ( [ gamma - sum ] ) for the dephasing rate coincide in the steady state ( because then @xmath179 ) , but they are different during the transient evolution . the rate ( [ gamma - d ] ) reflects the information loss due to emitted field , while the rate ( [ gamma - sum ] ) also includes the effect from changing entanglement between the qubit and the resonator . similarly , the results ( [ delta - omega-12 ] ) and ( [ delta - omega - sum ] ) for the ac stark shift coincide in the steady state ( then @xmath180 ) , but differ during transients . equation ( [ delta - omega-12 ] ) is applicable to the entangled qubit - resonator state , while eq . ( [ delta - omega - sum ] ) assumes tracing over the resonator state . also note that all these results for the dephasing rate and ac stark shift are applicable only in the case of a non - evolving qubit ( i.e. , when the evolution is only due to measurement ) . therefore , they are applicable to the ramsey sequence ( with short pulses ) , but , strictly speaking , not applicable to rabi oscillations , spectroscopic measurement of the ac stark shift , etc . for the echo sequence our results are not applicable directly , but the exact results can still be easily obtained using the same derivation ( assuming sufficiently short pulses applied to the qubit ) . in the `` bad cavity '' limit we can neglect the transients and use the steady - state values @xmath181 and @xmath182 . if additionally @xmath116 , then eqs . ( [ gamma - sum ] ) and ( [ delta - omega - sum ] ) reduce to eqs . ( [ gamma - bc2 ] ) and ( [ stark - bc ] ) . if @xmath183 , then for the qubit dephasing and ac stark shift in sec . iii we need to use steady - state versions of eqs . ( [ gamma - sum ] ) and ( [ delta - omega - sum ] ) or , equivalently , eqs . ( [ gamma - d ] ) and ( [ delta - omega-12 ] ) . to describe an individual measurement realization with a random result and evolution depending on this result , we will measure the pieces of the `` history tail '' in eq . ( [ evol-01 ] ) see fig . [ fig : tail ] . note that each piece can be measured in a different way , so the measurement properties can be changing in time . moreover , in general the sequence of measurement of the pieces can also be arbitrary ( e.g. , in a `` delayed choice '' experiment ) . we are interested in describing homodyne measurement with a phase - sensitive or phase - preserving amplifier . let us start with describing an _ ideal _ ( with perfect quantum efficiency ) _ phase - sensitive _ homodyne measurement . we will use the following physical model to describe such a measurement : 1 . a large coherent - state field @xmath184 ( @xmath185 ) from a pump is added to the piece of the tail . the number of photons @xmath26 is measured in the resulting state . 3 . for a particular random @xmath26 obtained in this measurement , the wavefunction is collapsed in the standard textbook way . this procedure describes well the optical homodyne measurement ( note that the photon number does not actually need to be resolved with single - photon precision since the fluctuations are significant ) . it is also similar to what is done experimentally in a phase - sensitive superconducting parametric amplifier . for example , in refs . @xcite the phase - sensitive parametric amplifier works by adding a pump microwave to the microwave leaked from the resonator using a directional coupler ( a microwave analog of a beam splitter ) . then the resulting microwave is sent to a nonlinear oscillator , whose frequency depends on the oscillation amplitude . this frequency change is then sensed via the corresponding phase change at the mixer . thus we measure the power of the pump with added signal , i.e. , within the time interval @xmath8 we essentially measure the corresponding number of photons @xmath26 ( again , single - photon precision in measuring @xmath26 is not needed ) . in a more complicated case of sideband pumping ( double - pumping ) of the parametric amplifier @xcite , the added resonant pump wave is modulated in amplitude ; however , the general principle remains practically the same . the case of a parametric pumping at the doubled frequency is different , but it is still practically equivalent to the measurement described by our model . note that in order to add the field @xmath184 , we need a beam splitter ( directional coupler ) which almost fully passes the signal , so the applied pump field should be much larger than the already large field @xmath184 . also note that a small part of the signal in this case will be lost , so that perfect quantum efficiency is impossible . however , we will not consider these details , and will also not consider ways to go around these problems ( e.g. , by using a balanced homodyne detection ) . it is rather simple to analyze the measurement using our model . for describing measurement of @xmath148th piece of the tail , let us rewrite eq . ( [ evol-01 ] ) as @xmath186 where @xmath187 is the measured @xmath148th piece of the tail , while remaining terms in eq . ( [ evol-01 ] ) are denoted as the normalized wavefunctions @xmath188 . [ actually , if we measure each piece of the tail immediately as it emerges , then it is sufficient to consider @xmath189 , which contain only the qubit and resonator states and do not contain unmeasured pieces of the tail . however , with eqs . ( [ psi-1])([psi - j ] ) we can in general consider a `` delayed choice '' version of the measurement . ] after the first step in the procedure ( addition of the pump field @xmath184 ) , the tail pieces @xmath190 become @xmath191 $ ] [ see eq.([displacement - comp ] ) for displacement by operator @xmath192 , therefore , the state ( [ psi-1 ] ) becomes @xmath193 at the second step of our procedure we need to measure the number of photons @xmath26 in the pump - plus - piece - of - tail part of the state ( [ psi-2 ] ) . the probability distribution for obtaining a particular @xmath26 is @xmath194 as follows from eqs . ( [ psi-2 ] ) and ( [ prop-2 - 1 ] ) . this distribution is normalized , @xmath195 , because @xmath196 . the third step of the procedure is the orthodox collapse of the state ( [ psi-2 ] ) onto the particular ( random ) measurement result @xmath26 . this means that instead of the states @xmath197 in eq . ( [ psi-2 ] ) , we pick only the amplitude corresponding to @xmath198 , and then renormalize the wavefunction ( [ psi-2 ] ) , so that it becomes [ see eq . ( [ alpha - def-2 ] ) ] @xmath199 where the normalization @xmath200 ensures that @xmath201 . note that the the overall phase of @xmath202 is not important . as we see , the `` quantum back - action '' due to the collapse changes the amplitudes of the pre - measured state ( [ psi-1 ] ) : @xmath203 and @xmath204 . this is the main idea for the description of the evolution due to measurement in the quantum bayesian formalism . the procedure can be applied to measurement of other pieces of the `` history tail '' in the same way . let us transform eqs . ( [ p(n)-1 ] ) and ( [ tilde - c-1 ] ) into a more useful form , using the assumption of a large pump amplitude , @xmath205 and @xmath206 [ see eq . ( [ alpha - def-2 ] ) ] , @xmath207}{\sqrt{2\pi\sigma^2 } } } \nonumber \\ & & \hspace{2.6 cm } \times \exp [ i n\ , { \rm arg } ( \alpha_{\rm p}+\alpha_{j,\rm t } ) ] \ , \|n\rangle , \qquad \label{alpha - gauss}\\ & & \hspace{-0.0 cm } \bar{n}_j = \|\alpha_{\rm p}\|^2 + 2\ , { \rm re } ( \alpha_{\rm p}^\alpha_{j,\rm t})+ \|\alpha_{j,\rm t}\|^2 \nonumber \\ & & \hspace{0.5 cm } \approx \|\alpha_{\rm p}\|^2 + 2\ , { \rm re } ( \alpha_{\rm p}^\alpha_{j,\rm t } ) , \label{n - bar-01 } \\ & & \sigma=\|\alpha_{\rm p}\|\gg 1 , \end{aligned}\ ] ] where @xmath208 is the average number of photons for the state @xmath209 ( we can neglect the last term @xmath210 for @xmath208 since @xmath211 ) and @xmath212 is the standard deviation . note that we use the same @xmath213 for both states because @xmath214 . also note that for exact normalization of the gaussian state ( [ alpha - gauss ] ) at finite @xmath215 the denominator @xmath216 should be slightly changed ; however , this is not important for the derivation . the probability distribution ( [ p(n)-1 ] ) for measuring @xmath26 photons in this case becomes p(n)= + , [ p(n)-gauss]and the updated amplitudes @xmath217 and @xmath218 given by eq.([tilde - c-1 ] ) become _ j= . [ tilde - c-2]we see that if the measurement result @xmath26 is closer to @xmath121 than to @xmath122 , then the amplitude @xmath219 increases ( by absolute value ) in this update . this is the expected feature of the quantum bayesian formalism : if the measurement result is more consistent with the qubit state @xmath6 , then the amplitude of this state increases . to simplify the phase factor in eq . ( [ tilde - c-2 ] ) , let us write @xmath220 as @xmath221 , and then expand @xmath222 to the second order , so that e^i n arg ( _ p+_j , t ) = e^i n arg ( _ p ) e^in im(_j , t/_p ) [ 1-re(_j , t/_p ) ] [ phase-1](we need the second order because @xmath223 and we wish to keep the terms linear in @xmath224 ) . the @xmath225-independent phase factor @xmath226 can be ignored as an overall phase . in the remaining phase in eq.([phase-1 ] ) let us represent @xmath26 as @xmath227 with the center point \|n_c(\|n_0+\|n_1)/2 . from eq . ( [ n - bar-01 ] ) , neglecting the term @xmath228 , we find @xmath229 $ ] . if @xmath230 , then the phase in eq . ( [ tilde - c-2 ] ) ( neglecting the overall phase ) is @xmath231 $ ] , which can be written ( neglecting the terms of order @xmath232 ) as @xmath233 $ ] . moving the phase difference to @xmath234 ( i.e. , considering the phase for @xmath235 as an unimportant overall phase ) , we find that the phase evolution in eq . ( [ tilde - c-2 ] ) can be described by multiplying @xmath163 by @xmath236 . this is exactly what we would expect from the phase of the inner product @xmath237 , and it is fully consistent with the result ( [ delta - omega-1 ] ) for the ac stark shift contribution . thus , for @xmath230 the phase shift produced by the collapse is the same as the ensemble - averaged phase shift . when @xmath238 , there is an additional phase factor @xmath239 in eqs . ( [ tilde - c-2 ] ) and ( [ phase-1 ] ) [ we now use @xmath240 , neglecting a phase correction of order @xmath232 ] . moving the phase difference to @xmath234 , we find that @xmath163 should be additionally multiplied by @xmath242}$ ] . thus , the evolution due to measurement [ see eqs . ( [ psi-1 ] ) , ( [ tilde - psi ] ) , and ( [ tilde - c-2 ] ) ] can be described as @xmath243}{\rm norm } , \qquad \label{tilde - c0}\\ & & \tilde{c}_1= \frac{c_1\,\exp [ -(n-\bar{n}_1)^2/4\sigma^2]}{\rm norm } \ , e^{-i\,\delta \varphi } , \label{tilde - c1}\\ & & { \rm norm } = \sqrt{\sum\nolimits_{j=0,1 } \|c_j\|^2 \exp [ -(n-\bar{n}_j)^2/2\sigma^2 ] } , \\ & & \delta\varphi = -\frac{n-\bar{n}_{\rm c } } { \sigma}\,\sqrt{\kappa \delta t } \,\ , { \rm im}\ { [ \alpha_1(t_m)-\alpha_0(t_m)]\,e^{-i\phi_{\rm a}}\ } \nonumber \\ & & \hspace{0.8 cm } + \kappa \delta t\ , { \rm im}[\alpha_{1 } ^(t_m)\,\alpha_{0}(t_m ) ] , \,\,\,\ \label{delta - phi } \\ & & \phi_{\rm a } = \phi_{\rm p}={\rm arg } ( \alpha_{\rm p } ) , \end{aligned}\ ] ] where @xmath244 is the time moment when the measured piece of the `` history tail '' leaked from the resonator , and @xmath47 is the phase of the pump , which determines the amplified quadrature . we emphasize that the measured piece becomes unentangled with the rest of the wavefunction [ see eq . ( [ tilde - psi ] ) ] and therefore can be disregarded when the measurement of the next piece is analyzed . note that the pump phase @xmath47 affects the response [ see eq . ( [ n - bar-01 ] ) ] \|n_1-\|n_0 = 2 re \{[_1(t_m)-_0(t_m ) ] e^-i_a } and also affects the phase shift @xmath245 in eq.([delta - phi ] ) . thus , the choice of the measured quadrature affects evolution of the system ( as in the `` bad cavity '' case @xcite ) . however , it is simple to show that the state update ( [ tilde - c0])([delta - phi ] ) averaged over the measurement result ( [ p(n)-gauss ] ) does not depend on the choice of @xmath47 . the formalism developed in secs . [ sec : main - idea ] and [ sec : gaussian ] allows us to consider measurement of the `` history tail '' pieces in an arbitrary sequence and thus to describe various `` delayed choice '' experiments . however , usually this is not needed , and we can assume measurement of the pieces as soon as they leak from the resonator . in this case it is sufficient to describe the system by an entangled qubit - resonator wavefunction ( we still consider an ideal case ) \|(t)= c_0(t ) \|0 \|_0(t)+ c_1(t ) \|1 \|_1(t ) , [ psi - cont - meas]with the coefficients @xmath246 and @xmath247 evolving in time due to measurement . then the evolution equations are essentially the same as eqs . ( [ tilde - c0])([delta - phi ] ) , but now there is no delay in measurement , and we have included the phase factors @xmath248 and @xmath249 in eq . ( [ evol-01 ] ) into the coefficients @xmath162 and @xmath163 in eq . ( [ psi - cont - meas ] ) [ actually , we include the phase difference @xmath250 into @xmath163 , neglecting the overall phase @xmath248 ] . also , instead of the measured number of photons @xmath26 , let us introduce the output signal @xmath252 . similarly , @xmath253 and @xmath254 are the corresponding average values , and @xmath255 is the variance of @xmath70 . then eqs . ( [ tilde - c0])([delta - phi ] ) can be rewritten as @xmath256}{\rm norm } , \qquad \label{c0 - 1}\\ & & c_1(t+\delta t)= \frac{c_1 ( t)\,\exp [ -(i_{\rm m}-i_1)^2/4d]}{\rm norm } \ , e^{-i\,\delta \varphi } , \quad \,\,\ , \label{c1 - 1}\\ & & { \rm norm } = \sqrt{\sum\nolimits_{j=0,1 } \|c_j(t)\|^2 \exp [ -(i_{\rm m}-i_j)^2/2d ] } , \qquad \\ & & \delta\varphi = - \frac{i_{\rm m}-(i_0+i_1)/2}{\sqrt{d}}\,\sqrt{\kappa \delta t } \,\ , { \rm i m } [ ( \alpha_1-\alpha_0)\,e^{-i\phi_{\rm a } } ] \nonumber \\ & & \hspace{1.0 cm } + \ , \delta \omega_{\rm q , s } \ , \delta t , \label{delta - phi-1}\end{aligned}\ ] ] where the qubit ac stark shift @xmath257 due to leaking field is given by eqs . ( [ delta - omega-1])([delta - omega-12 ] ) . note that eqs . ( [ c0 - 1])([delta - phi-1 ] ) do not change if we multiply @xmath70 , @xmath74 , @xmath75 , and @xmath258 by an arbitrary factor . therefore , we can consider them just as experimental output signals ( in arbitrary units ) , so that i_m=_t^t+t i(t ) dt for a continuous measurement output @xmath0 , and the variance @xmath259 is related to the single - sided spectral density @xmath81 of the output signal as d= . [ d - def ] now let us introduce the angle difference @xmath77 between the amplified quadrature along @xmath184 and the `` information - carrying '' quadrature along @xmath260 , _ d = _ a - arg[_1 ( t ) -_0(t ) ] , and also introduce the maximum response @xmath261 , which _ would _ correspond to @xmath108 , so that i = i_1 -i_0 = i_max ( _ d ) . then we can write the factor @xmath262 $ ] in eq . ( [ delta - phi-1 ] ) as @xmath263 . also counting the measurement signal from the central point @xmath73 , i_m = i_m - , we can rewrite evolution equations ( [ c0 - 1])([delta - phi-1 ] ) as @xmath264 where @xmath265 ensures @xmath266 . note that for short @xmath8 the variance @xmath259 of the noisy signal @xmath110 is much larger than @xmath267 , and then @xmath268 , so that the change of @xmath162 and @xmath163 is small . however , from the structure of eqs . ( [ c0 - 2])([delta - phi-2 ] ) it is easy to see that they _ remain valid for an arbitrary long @xmath8 _ if @xmath79 , @xmath77 , @xmath269 and noise @xmath81 do not change with time . therefore , the practical upper limit for the time step @xmath8 ( e.g. , in numerical simulations ) is determined by transients , which change the resonator states @xmath270 and @xmath157 , and by possible changes of the amplified quadrature phase @xmath47 . so far we considered an ideal phase - sensitive measurement , so that the evolution description using a wavefunction was sufficient . to describe a measurement with imperfect efficiency , we need to use the language of density matrices ( we still assume that the qubit evolves only due to measurement ) . then instead of eq . ( [ psi - cont - meas ] ) , the evolution of the entangled qubit - resonator system is described by the density operator ^q & r(t)= _ j , j=0,1 _ jj(t ) \|jj\| \|_j ( t)_j(t)\| , [ rho - qr - def]where @xmath270 and @xmath157 are given by eqs . ( [ alpha0-dot ] ) and ( [ alpha1-dot ] ) . we emphasize that the matrix elements @xmath271 describe the entangled qubit - resonator state , not only the qubit state . ( note that using the form ( [ rho - qr - def ] ) for the qubit - resonator state is equivalent to the polaron - frame approximation used in the theory of quantum trajectories @xcite . ) in the ideal case the evolution of the matrix elements @xmath272 can be obtained by converting eqs . ( [ c0 - 2])([delta - phi-2 ] ) into the language of density matrices , @xmath273 , \qquad \label{rho - diag}\\ & & \frac{\rho_{10}(t+\delta t)}{\rho_{10}(t ) } = \frac{\sqrt{\rho_{11}(t+\delta t)\,\rho_{00}(t+\delta t)}}{\sqrt{\rho_{11}(t)\,\rho_{00}(t)}}\ , e^{-i\delta\varphi } , \label{rho - offdiag}\end{aligned}\ ] ] where the phase shift @xmath245 is still given by eq . ( [ delta - phi-2 ] ) . another , more intuitive way to describe the evolution of the diagonal elements is by using the uncentered signal @xmath70 as in eq . ( [ c0 - 1 ] ) : _ jj(t+t)= . [ rho - jj ] note that these evolution equations for @xmath272 are exactly the same as eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) of the quantum bayesian formalism in the `` bad cavity '' limit , except the ac stark shift @xmath120 is now @xmath257 , dephasing is so far absent ( @xmath274 ) , and , most importantly , eqs . ( [ rho - diag ] ) and ( [ rho - offdiag ] ) describe the entangled qubit - resonator state ( [ rho - qr - def ] ) and are capable of describing transient evolution . during transients there is significant time - dependence in @xmath275 and @xmath276 , which also leads to time - dependence in @xmath277 , the quadrature phase difference @xmath278 , the response @xmath279 , and the middle point @xmath280 . therefore , during transients the time step @xmath8 in eqs . ( [ rho - diag ] ) and ( [ rho - offdiag ] ) should be much smaller than @xmath139 , in contrast to arbitrary @xmath68 in eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) . imperfect quantum efficiency @xmath84 of the measurement ( @xmath281 ) , similar to the case discussed in sec . [ sec : ph - sens - bc ] , mainly originates from two mechanisms : imperfect collection efficiency and imperfect amplifier efficiency . first , a fraction of the field leaked from the resonator is lost before reaching the amplifier . second , the amplifier produces more output noise than the quantum limitation . therefore , we can define the total quantum efficiency @xmath84 as = _ col_amp , _ col = , _ amp = , where @xmath282 is the ratio of the `` collected '' microwave energy , which reaches amplifier , to the total energy lost by the resonator ( @xmath283 ) , and @xmath92 is the spectral density of the output noise if a quantum - limited phase - preserving amplifier were used instead of the actual amplifier , which produces a larger noise @xmath81 . let us now discuss the effects produced by imperfect @xmath87 and @xmath86 . an imperfect collection efficiency @xmath87 can be modelled by adding an asymmetric beam splitter on the path of the leaked field , which splits each piece of the `` history tail '' into two pieces , @xmath284 , so that the first piece is measured , while the second one remains unmeasured . since no information can be obtained from the unmeasured piece , we need to trace over it , as in the calculation of the ensemble - averaged evolution , while for the measured piece we use the same procedure as above . the tracing over the unmeasured piece does not change the diagonal matrix elements in eq . ( [ rho - qr - def ] ) ; therefore , the total change of @xmath285 and @xmath286 is still given by eq . ( [ rho - diag ] ) . note , however , that imperfect collection efficiency reduces the response , @xmath287 and changes the central point @xmath73 , while the variance @xmath259 ( determined by the amplifier noise ) remains unchanged . for the off - diagonal matrix element @xmath109 , the tracing over the unmeasured piece @xmath288 produces the factor @xmath289 [ see eqs . ( [ rho-10-qr-1])([delta - omega-12 ] ) in sec . [ sec : ensemble - aver ] ] , while the measured piece gives the evolution described by eq . ( [ rho - offdiag ] ) , with @xmath290 in eq . ( [ delta - phi-2 ] ) multiplied by @xmath87 . therefore , the total evolution of @xmath109 is described by eq . ( [ rho - offdiag ] ) with the extra factor @xmath291 , while the phase @xmath245 is still given by eq . ( [ delta - phi-2 ] ) [ note again that @xmath79 and @xmath73 are affected by @xmath87 , but still correspond to the experimentally measured values ] . thus , the only effect of an imperfect collection efficiency @xmath87 on the system evolution ( [ rho - diag])([rho - offdiag ] ) is the extra factor @xmath291 in eq . ( [ rho - offdiag ] ) , where @xmath292 is given by eq . ( [ gamma - d ] ) . imperfect quantum efficiency @xmath86 of the amplifier produces additional noise at the output , so that the response @xmath79 and the middle point @xmath73 do not change , while the variance @xmath259 given by eq . ( [ d - def ] ) increases because of the increased noise spectral density @xmath81 . to take into account the extra noise , for a given measured output value @xmath72 , we need to guess what was the `` actual '' value @xmath293 ( the probability distribution is given by the classical bayesian analysis ) , then apply the evolution ( [ rho - diag])([rho - offdiag ] ) using the value @xmath293 , and then average over all possible values of @xmath293 . this is exactly what was done in ref . @xcite for a qubit measurement by qpc or set . since the evolution equations ( [ rho - diag ] ) , ( [ rho - offdiag ] ) , and ( [ delta - phi-2 ] ) have exactly the same form as what was considered in ref . @xcite , we can simply use the obtained result : the evolution is still given by eqs . ( [ rho - diag ] ) , ( [ rho - offdiag ] ) , and ( [ delta - phi-2 ] ) with two changes . first , the variance @xmath259 is the actual ( increased ) variance ; second , there is an extra dephasing factor in eq . ( [ rho - offdiag ] ) , which can be found from comparison with ensemble - averaged evolution . even though the formal derivation of this result is rather lengthy @xcite , it is easy to understand it . the evolution of the diagonal elements of the density matrix is the evolution of probabilities , and therefore must obey the classical bayes formula , which directly gives eq . ( [ rho - diag ] ) . the extra dephasing in eq . ( [ rho - offdiag ] ) comes from uncertainty of the phase @xmath294 due to uncertainty of the unknown `` actual '' value @xmath293 . the reduced proportionality factor between @xmath294 and the ( centered ) measurement result @xmath72 in eq . ( [ delta - phi-2 ] ) due to increased value of @xmath259 can be understood from the fact that for uncorrelated gaussian - distributed zero - mean random numbers @xmath295 and @xmath296 , the averaging of @xmath295 for a fixed sum @xmath297 gives the smaller value , @xmath298 $ ] . thus , combining both imperfection mechanisms of the quantum efficiency @xmath84 , we can describe the evolution of the qubit - resonator system ( [ rho - qr - def ] ) , measured using a phase - sensitive amplifier . the resulting equations are very similar to eqs . ( [ ph - sens - diag - bc ] ) and ( [ ph - sens - off - bc ] ) for the `` bad cavity '' case , @xmath299 , \qquad \label{rho - diag-2}\\ & & \hspace{-0.2 cm } \frac{\rho_{10}(t+\delta t)}{\rho_{10}(t ) } = \frac{\sqrt{\rho_{11}(t+\delta t)\,\rho_{00}(t+\delta t)}}{\sqrt{\rho_{11}(t)\,\rho_{00}(t)}}\ , \nonumber \\ & & \hspace{1.5 cm } \times \exp(-ik \tilde{i}_{\rm m } \ , \delta t ) \,\ , e^{-\gamma \delta t } \ , e^{-i\delta\omega_{\rm q , s}\delta t } , \qquad \label{rho - offdiag-2}\end{aligned}\ ] ] where @xmath300 and we repeated several previous formulas here for convenience . note that most parameters in these equations depend on time during transients , and therefore the time step @xmath8 should be sufficiently small . recall that @xmath81 is the single - sided spectral density of the output noise and @xmath301 is the corresponding noise variance of @xmath72 . equations ( [ rho - diag-2])([stark-2 ] ) are the _ main result _ of this paper . if instead of using experimental output signal @xmath0 we want to simulate the process , we can pick @xmath302 from the probability distribution @xmath303 } { \sqrt{2\pi d } } \nonumber \\ & & \hspace{1.0 cm } + \rho_{11}(t)\ , \frac{\exp [ - ( \tilde i_{\rm m}- \delta i/2)^2/2d ] } { \sqrt{2\pi d } } . \qquad \label{tilde - i - distr}\end{aligned}\ ] ] for an infinitesimally small @xmath8 this is equivalent to using i(t)= + [ _ 11 ( t)-_00(t ) ] + _ i(t ) , [ ph - sens - i(t)-2]where @xmath304 is a white noise with spectral density @xmath81 . it is easy to check that averaging of @xmath305 given by eqs . ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) over @xmath72 with probability distribution ( [ tilde - i - distr ] ) produces the expected ensemble - averaged equations @xmath306 similar to what was mentioned in sec . [ sec : ph - pres - bc ] , the averaging over the fluctuating phase @xmath307 in eq . ( [ rho - offdiag-2 ] ) produces an ensemble dephasing rate @xmath308 , while the averaging of the first term in eq . ( [ rho - offdiag-2 ] ) produces an ensemble dephasing rate @xmath309 . both dephasing rates depend on the amplified quadrature via the angle @xmath77 , but their sum , @xmath310 , does not depend on @xmath77 . if there is extra ( not measurement - related ) dephasing rate @xmath311 of the qubit - resonator system , e.g. , due to intrinsic pure dephasing of the qubit , then it can be easily included into eq . ( [ rho - offdiag-2 ] ) by adding the factor @xmath312 . alternatively , we can include @xmath311 into the ensemble - averaged dephasing , @xmath313 , so that evolution equations ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) remain unchanged , but now @xmath314 . in this case the overall quantum efficiency includes the extra dephasing , @xmath315 , with @xmath316 @xcite . note that for the evolution equations discussed in this section the initial state should not necessarily be pure , so eq . ( [ psi - in ] ) for the initial state can be replaced with eq . ( [ rho - in ] ) . moreover , it is sufficient to have an initial state of the form ( [ rho - qr - def ] ) ; the only necessary condition is that each qubit state @xmath99 corresponds to a certain coherent state @xmath317 . so far we considered the measurement using a phase - sensitive amplifier . in this section we use the results of the previous section to describe the case when a phase - preserving amplifier is used . we will do it in two ways , which give the same result . first , let us model the measurement using a phase - preserving amplifier in the following way . let us pass each piece of the `` history tail '' through a symmetric beam splitter @xmath318 , amplify orthogonal quadratures in these two parts , measure as discussed in sec . [ sec : main - idea ] , and output the results for both quadratures . from the structure of eqs . ( [ rho - diag-2])([tilde - i - distr ] ) it is easy to see that it does not matter how these two orthogonal quadratures are chosen if the amplification conditions in both channels are the same ( the same @xmath259 and @xmath79 , which also means the same quantum efficiency ) . note that for both channels we should simultaneously use either the first or the second gaussian in eq . ( [ tilde - i - distr ] ) , though no correlation is needed in the infinitesimal limit ( [ ph - sens - i(t)-2 ] ) . it is natural to choose one quadrature ( we call it @xmath319 ) along the informational direction @xmath320 , while the other quadrature ( we call it @xmath321 ) is shifted by @xmath48 , so that @xmath108 for the @xmath319-quadrature and @xmath126 for the @xmath321-quadrature . thus , the directions of the @xmath319 and @xmath321 quadratures are changing in time , but they are practically constant during the time step @xmath8 . note that if @xmath259 for both quadratures is kept the same as for a phase - sensitive amplifier , then the response @xmath124 is a factor @xmath322 smaller than @xmath79 for the phase - sensitive case ( because of the beam splitter ) . equivalently , if @xmath124 is kept the same as in the phase - sensitive case ( e.g. , by an additional classical amplification by the factor @xmath322 ) , then @xmath259 for both quadratures is twice larger than for the phase - sensitive case ( in the ideal case this corresponds to the fact that the noise of a phase - preserving amplifier is twice as large as for a phase - sensitive amplifier ) . therefore , for the phase - preserving case we can simply use eqs . ( [ rho - diag-2])([tilde - i - distr ] ) twice , for the optimal quadrature ( @xmath108 ) and for the orthogonal quadrature ( @xmath126 ) , assuming the same output noise , @xmath323 , for both output quadratures @xmath0 and @xmath4 . the @xmath319-quadrature has the maximum response , @xmath324 , while the @xmath321-quadrature has no response , @xmath129 . thus , after the time step @xmath8 the density matrix ( [ rho - qr - def ] ) of the qubit - resonator system changes as @xmath299 , \qquad \label{rho - diag-3}\\ & & \hspace{-0.2 cm } \frac{\rho_{10}(t+\delta t)}{\rho_{10}(t ) } = \frac{\sqrt{\rho_{11}(t+\delta t)\,\rho_{00}(t+\delta t)}}{\sqrt{\rho_{11}(t)\,\rho_{00}(t)}}\ , \nonumber \\ & & \hspace{1.0 cm } \times \exp ( -i\tilde{q}_{\rm m } \delta i/2d ) \,\ , e^{-\gamma \delta t } \ , e^{-i\delta\omega_{\rm q , s}\delta t } , \qquad \label{rho - offdiag-3 } \end{aligned}\ ] ] where @xmath110 is given by eq . ( [ tilde - i - m ] ) , while @xmath325 the ensemble - averaged dephasing @xmath292 is given by eq . ( [ gamma - d ] ) , and the ac stark shift @xmath269 is given by eq . ( [ stark-2 ] ) or eqs . ( [ delta - omega-1])([delta - omega-12 ] ) . an equivalent form for eq . ( [ rho - diag-3 ] ) in terms of the non - centered signal @xmath70 is given by eq . ( [ rho - jj ] ) . the factor of 2 in eq . ( [ gamma-3 ] ) appears because averaging over the result in each channel produces the contribution @xmath309 into the total ensemble dephasing @xmath292 . another way to derived eqs . ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) from eqs . ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) is to assume a slightly shifted pump frequency for a phase - sensitive amplifier , so that the angle @xmath77 rotates sufficiently fast , and for both quadratures @xmath0 and @xmath4 we collect only the values averaged over @xmath77 . then we have a natural formation of two quadratures in eqs . ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) : @xmath326 and @xmath327 , where the superscripts indicate the phase - sensitive ( ps ) or phase - preserving ( pp ) case . the variance of the noise in each quadrature is @xmath328 ( because @xmath329 ) and the response in the information - carrying quadrature is @xmath330 ( because the phase - sensitive response @xmath331 should be multiplied by @xmath332 to project onto the proper quadrature ) . therefore @xmath333 , and eqs . ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) directly transform into eqs . ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) . for the dephasing we get @xmath334 since @xmath335 and @xmath330 , thus reproducing eq . ( [ gamma-3 ] ) . in numerical simulations the probability distribution for @xmath72 is still given by eq . ( [ tilde - i - distr ] ) , while for @xmath336 it is p(q_m ) = . [ tilde - q - distr-2 ] for infinitesimal @xmath8 these distributions are equivalent to using eq . ( [ ph - sens - i(t)-2 ] ) for @xmath0 and q(t)=q_0 + _ q ( t ) , s__q= s_q= s_i , for @xmath4 , with uncorrelated white noises in the two channels . averaging of @xmath305 in eqs . ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) over random @xmath72 and @xmath336 using the probability distributions ( [ tilde - i - distr ] ) and ( [ tilde - q - distr-2 ] ) produces the ensemble - averaged evolution equations ( [ rho - ea - diag ] ) and ( [ rho - ea - offdiag ] ) . the ensemble - averaged evolution should remain the same as in the phase - sensitive case because of causality . similar to eq . ( [ eta - phase - pres - def ] ) , the quantum efficiency for a phase - preserving measurement can be defined in two ways , = 1- /_d , = /2 , where the first definition is based on the comparison with ideal phase - preserving measurement , while in the second definition we compare the information in @xmath319-channel only with the ideal phase - sensitive case . we emphasize that monitoring of a pure quantum state is still possible with a phase - preserving amplifier if @xmath337 , in spite of the fundamental limitation @xmath338 . we intentionally wrote the evolution equations ( [ rho - diag-2 ] ) , ( [ rho - offdiag-2 ] ) , ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) for a finite @xmath8 because this form is more transparent physically , suitable for numerical simulations , and also unambiguous . the differential form for an infinitesimal @xmath8 is significantly more ambiguous because it depends on a chosen definition of the derivative ( as should be for nonlinear stochastic differential equations @xcite ) . if we define the derivative in the symmetric way @xmath339/\delta t$ ] ( the so - called stratonovich form ) , then the standard calculus rules apply , and the differential equations for the evolution can be derived from eqs . ( [ rho - diag-2 ] ) , ( [ rho - offdiag-2 ] ) , ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) in a straightforward way ( keeping linear order in @xmath8 ) . thus , for the phase - sensitive measurement we obtain the stratonovich - form evolution as ( see @xcite ) @xmath340 , \qquad \label{dot - rho-11 - 1}\\ & & \dot{\rho}_{10}= -(\rho_{11}-\rho_{00})\ , \frac{\cos{\phi_{\rm d}}\delta i_{\rm max}}{s_i } \left [ i(t)-\frac{i_0+i_1}{2 } \right ] \nonumber \\ & & \hspace{0.9 cm } -i \,\frac{\sin{\phi_{\rm d}}\delta i_{\rm max}}{s_i } \left [ i(t)- \frac{i_0+i_1}{2 } \right ] \rho_{10 } \nonumber \\ & & \hspace{0.9 cm } -\gamma \rho_{10 } -i\,\delta\omega_{\rm q , s}\rho_{10 } , \quad \label{dot - rho-10 - 1}\\ & & \gamma = \gamma_{\rm d } -(\delta i_{\rm max})^2/4s_i , \\ & & i(t ) = \rho_{00}(t ) \ , i_0+\rho_{11}(t ) \ , i_1+\xi_i(t ) , \,\,\ , s_{\xi_i}=s_i , \label{i(t)}\end{aligned}\ ] ] where for convenience we repeated equations for @xmath83 and @xmath0 . we emphasize that @xmath74 , @xmath75 , @xmath79 , and @xmath77 may significantly depend on time during transients . for the phase - preserving measurement we similarly obtain the stratonovich - form equations @xmath341 , \qquad \label{dot - rho-11 - 2}\\ & & \dot{\rho}_{10}= -(\rho_{11}-\rho_{00})\ , \frac{\delta i}{s_i } \left [ i(t)-\frac{i_0+i_1}{2 } \right ] \nonumber \\ & & \hspace{0.9 cm } -i \,\frac{\delta i}{s_i } \ , [ q(t)- q_0 ] \,\rho_{10 } \nonumber \\ & & \hspace{0.9 cm } -\gamma \rho_{10 } -i\,\delta\omega_{\rm q , s}\rho_{10 } , \quad \label{dot - rho-10 - 2 } \\ & & \gamma = \gamma_{\rm d } -2\times ( \delta i)^2/4s_i , \\ & & i(t ) = \rho_{00}(t ) \ , i_0+\rho_{11}(t ) \ , i_1+\xi_i(t ) , \label{i(t)-2}\\ & & q(t ) = q_0 + \xi_q(t ) , \,\,\,\,\ , s_{\xi_q}=s_{\xi_i}=s_i . \label{q(t)}\end{aligned}\ ] ] if we define the derivative in the `` forward '' way , @xmath342/\delta t$ ] ( the so - called it form ) , then the usual calculus rules are no longer correct , and the derivation of the differential equations from eqs . ( [ rho - diag-2 ] ) , ( [ rho - offdiag-2 ] ) , ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) should retain the second order in @xmath8 . alternatively , we can use the standard rules of the transformation from the stratonovich form into the it form @xcite , applied to eqs . ( [ dot - rho-11 - 1])([q(t ) ] ) . the resulting it - form equations for the phase - sensitive measurement are @xmath343 , \qquad \ , \label{dot - rho-11 - 3}\\ & & \hspace{-0.4 cm } \dot{\rho}_{10}= -(\rho_{11}-\rho_{00})\ , \frac{\cos{\phi_{\rm d}}\delta i_{\rm max}}{s_i } \ , [ i(t)- ( \rho_{00}i_0+\rho_{11}i_1 ) ] \nonumber \\ & & \hspace{-0.4 cm } \hspace{0.9 cm } -i \,\frac{\sin{\phi_{\rm d}}\delta i_{\rm max}}{s_i } \ , [ i(t)- ( \rho_{00}i_0+\rho_{11}i_1)]\ , \rho_{10 } \nonumber \\ & & \hspace{-0.4 cm } \hspace{0.9 cm } -\gamma_{\rm d } \rho_{10 } -i\,\delta\omega_{\rm q , s}\rho_{10 } , \end{aligned}\ ] ] and the it - form equations for the phase - preserving measurement are @xmath344 , \qquad \\ & & \dot{\rho}_{10}= -(\rho_{11}-\rho_{00})\ , \frac{\delta i}{s_i } [ i(t)- ( \rho_{00}i_0+\rho_{11}i_1 ) ] \nonumber \\ & & \hspace{0.9 cm } -i \,\frac{\delta i}{s_i } \ , [ q(t)- q_0 ] \,\rho_{10 } \nonumber \\ & & \hspace{0.9 cm } -\gamma_{\rm d } \rho_{10 } -i\,\delta\omega_{\rm q , s}\rho_{10 } , \qquad \label{dot - rho-10 - 4 } \end{aligned}\ ] ] while @xmath0 and @xmath4 for numerical simulations are still given by eqs . ( [ i(t ) ] ) , ( [ i(t)-2 ] ) , and ( [ q(t ) ] ) . the it - form equations ( [ dot - rho-11 - 3])([dot - rho-10 - 4 ] ) have two differences compared with the stratonovich equations : ( i ) the combination @xmath345 is replaced with the `` pure noise '' combination @xmath346 and ( ii ) the dephasing rate @xmath83 is replaced with ensemble dephasing @xmath292 . note that it and stratonovich equations have identical solutions when the corresponding definitions of the derivative are used . the drawback of the it form is the loss of intuition based on the standard calculus , because the standard calculus rules are not valid in the it form . however , the advantage is that the ensemble - averaged equations can be obtained by simply replacing the noises @xmath100 and @xmath347 with zero . the quantum trajectory formalism @xcite is based on the it form , while the quantum bayesian formalism @xcite usually uses the stratonovich form ( some formalisms use both forms @xcite ) . we emphasize that while the evolution equations in the differential form are useful in analytical analysis , for numerical calculations a relatively large time step @xmath8 is often preferable . for finite time steps , the formalism discussed in secs . [ sec : imperfect ] and [ sec : ph - pres ] is more useful than the differential equations . the use of non - infinitesimal @xmath8 also avoids possible confusion between stratonovich and it forms . now let us discuss evolution of the qubit - resonator system for an arbitrarily long duration @xmath68 . as in the previous sections , we assume that the qubit does not evolve due to rabi oscillations , energy relaxation , etc . it is not obvious what the solution of the differential equations discussed in sec . [ sec : differential ] is . however , the structure of equations for a small time step @xmath8 derived in secs . [ sec : imperfect ] and [ sec : ph - pres ] permits very simple integration for an arbitrary @xmath68 . this simple solution is also expected from the picture of the `` history tail '' in fig . [ fig : tail ] . the evolution equations ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) for the qubit - resonator system ( [ rho - qr - def ] ) can be easily integrated within the time interval @xmath348 $ ] , @xmath349 , \qquad \,\,\ , \label{rho - offdiag - tau}\end{aligned}\ ] ] where @xmath350}{s_i } \,\ , dt ' \\ & & \tilde{i}(t ' ) = i(t ' ) -\frac{i_0(t')+i_1(t')}{2 } , \label{tilde - i(t')}\end{aligned}\ ] ] and the time - dependent dephasing @xmath351 and ac stark shift @xmath352 are given by eqs . ( [ gamma-2])([stark-2 ] ) . note that because parameters are time - dependent , there is no simple relation between the effective measurement results @xmath353 and @xmath354 , which produce `` spooky '' and phase back - actions . the choice of notations @xmath355 and @xmath356 relate to quadratures that are parallel or perpendicular to the informational quadrature . if we need to generate measurement results numerically , then @xmath353 can be picked from the probability distribution @xmath357 , which consists of two gaussians , as usual in the bayesian formalism , @xmath358 , \qquad \label{p(r)-2}\\ & & \bar{r}_{\rm 1}^\parallel = - \bar{r}_{\rm 0}^\parallel = \int_{t}^{t+\tau } \frac{[\delta i(t')]^2}{s_i } \ , dt ' , \,\,\ , d_r^\parallel = 2 \bar{r}_{\rm 1}^\parallel . \label{p(r)-3}\end{aligned}\ ] ] the validity of this formula can be checked by analyzing a composition of two evolutions for @xmath359 and @xmath360 , and by checking consistency with formulas in sec . [ sec : imperfect ] for small @xmath68 . from eqs . ( [ p(r)-1])([p(r)-3 ] ) we see that the qubit will eventually be collapsed onto the state @xmath6 or @xmath7 ( unless @xmath361 ) , as expected for a measured qubit with no additional evolution . note that eqs . ( [ p(r)-1])([p(r)-3 ] ) for @xmath357 can be written in this simple way because eq . ( [ rho - diag - tau ] ) is essentially the classical bayes rule . unfortunately , @xmath354 can not be generated in a similar way . therefore , we need to numerically generate the whole record @xmath362 . the output realization @xmath362 within the interval @xmath348 $ ] can be generated by dividing @xmath68 into small pieces @xmath8 and using eq . ( [ tilde - i - distr ] ) . the probability of a realization @xmath362 will then be @xmath363 ^ 2}{s_i } \ , dt ' \bigg ] \nonumber \\ & & \hspace{0.5 cm } + \rho_{11}(t ) \ , \exp \bigg [ - \int_t^{t+\tau } \frac{[i(t')-i_1(t')]^2}{s_i } \ , dt ' \bigg ] , \qquad \label{i(t')-1}\end{aligned}\ ] ] with an appropriate overall normalization . alternatively , the probability distribution can be obtained by applying eq . ( [ ph - sens - i(t)-2 ] ) , i.e. , taking into account the randomness `` locally '' instead of `` globally '' , which produces @xmath364 ^ 2}{s_i } \ , dt ' \bigg ] , \qquad \label{i(t')-2}\\ & & \hspace{0 cm } i_{\rm av}(t')= \rho_{00}(t ' ) \ , i_0(t ' ) + \rho_{11}(t ' ) \ , i_1(t ' ) , \qquad \label{i(t')-3}\end{aligned}\ ] ] where @xmath365 and @xmath366 should be calculated using eq . ( [ rho - diag - tau ] ) for the previous period @xmath367 $ ] . even though this gives the same probability distribution , it is easier to use the `` global '' method ( [ i(t)-1 ] ) . integrating eqs . ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) , we obtain the evolution during the time interval @xmath348 $ ] , @xmath368 , \qquad \,\,\ , \label{rho - offdiag - tau-2}\end{aligned}\ ] ] where @xmath369 @xmath370 is given by eq . ( [ tilde - i(t ) ] ) , @xmath371 is given by eq . ( [ gamma-3 ] ) , and @xmath352 is given by eq . ( [ stark-2 ] ) . we emphasize that the outputs @xmath0 and @xmath4 correspond to the informational and non - informational quadratures , which change in time . in terms of the `` fixed '' experimental quadratures @xmath372 and @xmath373 from the iq mixer they are @xmath374 + q^{\rm fix } ( t ) \sin [ \phi_{\rm opt}(t ) ] , \qquad \\ & & q(t)= q^{\rm fix } ( t ) \cos [ \phi_{\rm opt}(t ) ] - i^{\rm fix } ( t ) \sin [ \phi_{\rm opt}(t ) ] , \end{aligned}\ ] ] where @xmath375 $ ] corresponds to the informational quadrature . if the measurement results are not taken from an experiment , but have to be generated numerically , then it is always possible to generate @xmath376 and @xmath377 without explicitly generating the signals @xmath0 and q(t ) . for @xmath376 we can still use eqs ( [ p(r)-1])([p(r)-3 ] ) , just replacing the superscript @xmath355 with @xmath319 . the probability distribution for @xmath378 is the zero - mean gaussian , @xmath379 , \qquad \label{p(r)-4}\\ & & d_r^q = d_r^i = 2 \int_{t}^{t+\tau } \frac{[\delta i(t')]^2}{s_i } \ , dt ' . \label{p(r)-5}\end{aligned}\ ] ] the probability distribution for a realization of @xmath362 is still given by eqs . ( [ i(t)-1 ] ) or ( [ i(t)-2 ] ) , while the similar probability distribution for @xmath380 is p\{q(t ) ] } . the evolution equations derived in this paper describe the evolution of an entangled qubit - resonator state ( [ rho - qr - def ] ) . however , there is an important special case when we can discuss the state of the qubit alone . if the measurement is of a relatively short duration and the microwave drive is switched off after that , then several decay times @xmath139 later ( or after the rapid driven reset procedure @xcite ) the resonator field is practically vacuum for both qubit states . in this case our formulas give the resulting qubit state , unentangled from the resonator state . in this paper we have developed a simple quantum bayesian formalism for the qubit measurement in the circuit qed setup with a moderate bandwidth of the measurement resonator , so that transients are important . the simplification comes from three assumptions : ( i ) we assume that the qubit evolves only due to measurement ( in particular , there are no rabi oscillations or qubit energy relaxation ) , ( ii ) we assume that the measurement resonator is driven by a classical , i.e. , coherent field ( in particular , no squeezed fields are applied ) , and ( iii ) the resonator is initially in a coherent state ( e.g. , vacuum ) . in this case the entangled qubit - resonator state developing in the process of measurement can be described as the density operator , eq . ( [ rho - qr - def ] ) , in which each of the two qubit states corresponds to its own coherent state of the resonator . therefore , the entangled qubit - resonator state at any moment of time is fully characterized by only 4 numbers : @xmath285 , @xmath286 , @xmath109 , @xmath93 , and two field amplitudes of the resonator : @xmath50 and @xmath51 . the field amplitudes evolve according to the classical equations ( [ alpha0-dot ] ) and ( [ alpha1-dot ] ) . the elements of the @xmath381 matrix @xmath382 evolve according to eqs . ( [ rho - diag-2 ] ) and ( [ rho - offdiag-2 ] ) if a phase - sensitive amplifier is used in the measurement or according to eqs . ( [ rho - diag-3 ] ) and ( [ rho - offdiag-3 ] ) if a phase - preserving amplifier is used . these evolution equations in differential form ( in both stratonovich and it forms ) are presented in sec . [ sec : differential ] . integrated equations for an arbitrary long evolution are discussed in sec . [ sec : arb - duration ] . the equations depend on parameters that are directly measurable in an experiment . the evolution equations for @xmath382 [ eqs . ( [ rho - diag-2 ] ) , ( [ rho - offdiag-2 ] ) , ( [ rho - diag-3 ] ) , ( [ rho - offdiag-3 ] ) ] have exactly the same form as in the `` bad cavity '' limit @xcite and have a simple physical meaning . we see that the diagonal elements @xmath285 and @xmath286 evolve as probabilities , i.e. , they follow the classical bayes rule , which updates the probabilities according to the information on the qubit state acquired from the measurement result . therefore , this `` spooky '' back - action is sensitive to the `` informational '' quadrature of the microwave field . the evolution of @xmath109 ( and @xmath383 ) necessarily depends on the evolution of @xmath285 and @xmath286 ( at least because @xmath384 ) . besides that , there are three more effects producing evolution of @xmath109 : ( i ) phase back - action , which depends on the measurement result sensitive to the `` non - informational '' quadrature of the microwave field , ( ii ) dephasing due to non - ideality of the measurement ( essentially loss of potential information ) , and ( iii ) ac stark shift of the qubit frequency . as discussed in appendix b , the phase back - action can be physically interpreted as being due to fluctuations of the ac stark shift because of a fluctuating number of photons in the resonator . even though the evolution equations ( [ rho - diag-2 ] ) , ( [ rho - offdiag-2 ] ) , ( [ rho - diag-3 ] ) , and ( [ rho - offdiag-3 ] ) are the same as in the `` bad cavity '' regime @xcite , the time step @xmath8 is no longer arbitrary , since the parameters entering the equations ( response @xmath124 , amplified phase difference @xmath77 , ensemble dephasing @xmath292 , etc . ) change during the transients , and therefore @xmath8 should be smaller than the time scale of this change . we emphasize that in the case of non - changing parameters these equations are exact for an arbitrary long @xmath8 . this may be beneficial for numerical simulations in comparison with the quantum trajectory formalism @xcite based on a wiener process , which assumes infinitesimal @xmath8 . in particular , our evolution equations can be easily integrated for an arbitrarily long duration [ eqs . ( [ rho - diag - tau ] ) , ( [ rho - offdiag - tau ] ) , ( [ rho - diag - tau-2 ] ) , and ( [ rho - offdiag - tau-2 ] ) ] . we note that the evolution equations in the phase - sensitive case are also exactly the same as for a qubit measurement by qpc or set @xcite , except now we consider a significantly entangled qubit - resonator state , with classically evolving resonator fields . the case of a phase - preserving amplifier is different because there are two output signals , @xmath0 and @xmath4 , instead of only one signal @xmath0 . nevertheless , the evolution equations are almost the same , and the only significant difference is that the phase back - action is governed by the non - informational quadrature @xmath4 , while the `` spooky '' back - action ( evolution of @xmath285 and @xmath286 ) is governed by the informational quadrature @xmath0 . the derivation in this paper has been based on elementary quantum mechanics and basic facts related to coherent states . in general , the idea is similar to the idea of `` microscopic '' derivation @xcite used to describe a qubit measurement by qpc or set . we solve exactly the quantum evolution due to interaction between the qubit and resonator ( which is very simple because the qubit does not evolve by itself and measurement is of the qnd type ) , and then apply the textbook collapse postulate to the pieces of microwave field , leaking from the resonator . the formalism developed in this paper is equivalent to the `` polaron frame approximation '' used in the quantum trajectory formalism @xcite , even though our language is significantly different . we hope that our derivation is physically transparent and therefore more easily understandable . also , as mentioned above , our formalism may have advantages in numerical calculations . for an evolving qubit ( e.g. , due to rabi oscillations ) it is tempting to simply include additional evolution into the differential equations for evolution derived in sec . [ sec : differential ] . however , this is formally incorrect because in this case the approach based on coherent states is no longer applicable ( though this is still possible in the `` bad cavity '' limit @xcite ) . the reason is the following . when the additional evolution of the qubit is comparable to or faster than @xmath5 , the resonator state @xmath385 or @xmath386 may correspond to the `` wrong '' qubit state produced by this evolution . since for a resonator the evolution of a superposition of coherent states ( a `` cat state '' ) can not be easily described with coherent states , the simple approach based on coherent states fails . therefore , for measurement of an evolving qubit the simple formalism discussed in this paper is not applicable and should be replaced with a more complicated formalism . the derivation of the quantum bayesian formalism for measurement of an evolving qubit is similar ideologically ( using the measurement of the `` history tail '' ) , but much more cumbersome technically . the result is equivalent to `` full '' quantum trajectory formalism @xcite , but uses an explicit fock - space evolution in the schrdinger picture instead of the language of superoperators . we will discuss this formalism in another paper . the formalism developed in this paper can be easily generalized to measurement of a multi - level transmon or measurement of several qubits , which evolve only due to measurement . such a generalization is useful to describe the process of entanglement of superconducting qubits by measurement @xcite . for @xmath387 qubits the state of the system can be described in the way similar to eq . ( [ rho - qr - def ] ) , so that each of @xmath388 qubit basis states corresponds to particular coherent states of the resonators , obtained via the classical field evolution . therefore , we only need to describe the evolution of @xmath389 matrix of coefficients , for which we can easily use the quantum bayesian approach to update the coefficients , depending on the measurement results . this will also be the subject of a future publication . the author thanks justin dressel , eyob sete , mark dykman , farid khalili , and konstantin likharev for useful discussions . the author also thanks justin dressel and juan atalaya for critical reading of the manuscript . the work was supported by aro grant no . w911nf-15 - 1 - 0496 . in this appendix we review basic facts related to coherent states . most of them are very well known in the quantum optics community . however , some of these facts [ e.g. , eq . ( [ phi - dot-2 ] ) ] are usually not discussed in optical textbooks . in contrast to the notation used in the main text , in this appendix we will use hat symbols for operators . as known from undergraduate quantum mechanics , for an oscillator with frequency @xmath12 and mass @xmath148 , the ground state in the @xmath390-representation is \|0= _ gr(x ) = ( ) ^1/4 ( - x^2 ) . if we want to describe the classical state of this oscillator with coordinate @xmath391 and momentum @xmath392 ( still taking into account the uncertainty of the ground state ) , we need to shift the ground - state wavefunction by @xmath391 , producing @xmath393 , and also apply the momentum shift by adding the factor @xmath394 . this produces the so - called `` coherent state '' @xmath395 , which is widely used in optics : @xmath396 where @xmath397 and @xmath398 are the ground - state uncertainties . the normalization by doubled uncertanties @xmath399 and @xmath400 in eq . ( [ re - im ] ) as well as the overall phase factor @xmath401 in eq . ( [ alpha - def ] ) are to some extent arbitrary , but this conventional choice simplifies most of the formulas discussed below . note that the phase @xmath402 is exactly in between what we would obtain by first shifting @xmath390 , and then @xmath403 [ in this case we would obtain @xmath404 and , instead , first shifting @xmath403 and then @xmath390 [ in this case we would obtain @xmath405 . equation ( [ alpha - def ] ) can be rewritten in a more standard form @xcite @xmath406 where @xmath407 is the raising ( creation ) operator , @xmath408 ) satisfies the relations @xmath409\ , \|\alpha\rangle$ ] and @xmath410\ , confusion between the notations for the stationary states @xmath198 and the coherent state @xmath411 ( for example , @xmath412 with @xmath413 is not the first excited level @xmath7 ) ; to avoid the confusion , we can use greek letters for coherent states and roman letters or integer numbers for the stationary states ( fock states ) . for the ground state the notations coincide , @xmath414 . if the oscillator state rotates with frequency @xmath2 ( for example , due to drive with this frequency ) , @xmath415 , @xmath416 , then from eq . ( [ re - im ] ) we find @xmath417 . in this case it is useful to introduce the rotating frame by defining @xmath418 , so that @xmath419 does not change in time . in the general case @xmath419 changes with time slowly , while @xmath420 rapidly oscillates . the rotating frame frequency @xmath2 can be chosen arbitrarily ; in the case with a drive , the most natural choice is the drive frequency @xmath22 ( because then @xmath419 does not change in the steady state ) ; in the absence of the drive , a natural choice is the oscillator frequency @xmath12 . note that the time dependence for the stationary states is @xmath421 ( counting the energy from the ground state energy ) , so for a `` non - evolving '' oscillator ( i.e. , evolving only naturally ) , from eq . ( [ alpha - def-2 ] ) we find @xmath422 . note that in the main text we always use the rotating frame based on the drive frequency @xmath22 and omit the tilde sign in the notation of the rotating - frame @xmath44 . in contrast , in this appendix we explicitly write @xmath423 for the rotating frame . so far we considered a textbook mechanical oscillator . if we consider a microwave resonator , then the role of @xmath390 and @xmath403 is played by properly normalized voltage and current ( at some point in the resonator ) or by flux and charge ; the effective mass @xmath148 can also be appropriately introduced . the formalism does not change . in quantum optics it is often preferred not to introduce coordinates and effective mass explicitly , and instead to start with the commutation relation @xmath424=1 $ ] , then producing fock states @xmath198 from vacuum @xmath6 with the creation operator . 1 . * from eq . ( [ alpha - def-2 ] ) it is easy to see that \|= \| , [ prop-1]since @xmath425 for the lowering ( annihilation ) operator @xmath426 . the property ( [ prop-1 ] ) is sometimes used as a definition of the coherent state @xmath395 . note , however , that it does not specify the overall phase and normalization , while the overall phase if often important in analysis ( when more than one coherent state is involved ) . also note that @xmath427 does not have a simple formula , though @xmath428 from conjugation of @xmath429 . * 2 . * from eq . ( [ alpha - def-2 ] ) , the probability to measure @xmath26 photons in the state @xmath412 is p(n ) = e^-\|\|^2\|^2\|^n / n ! , [ prop-2 - 1]which is the poissonian distribution with average @xmath430 . this proves that the wavefunction ( [ alpha - def-2 ] ) is normalized and shows that the mean photon number is * 3 . * the inner product of two coherent states @xmath395 and @xmath431 can be easily calculated using eq.([alpha - def-2 ] ) , giving the result @xcite \|=e^-(\|\|^2+\|\|^2 ) e^^ * = e^-\|-\|^2 e^ -iim(^ * ) . [ inner - product]note that a shift of the coherent states by the same value changes the inner product , @xmath432 , since this changes the phase factor . it is useful to introduce the ( unitary ) displacement operator @xmath433 @xcite , ( ) ( ^-^ * ) , ( ) \|0= \| . a composition of two displacement operators has a phase factor @xcite similar to the phase factor in eq . ( [ inner - product ] ) , ( ) ( ) = ( + ) , [ displacement - comp]as follows from the baker - campbell - hausdorff formula @xmath434 for @xmath435=c$ ] . also note the useful relations @xmath436 * 5 . * let us introduce the ( hermitian ) quadrature operators @xmath437 and @xmath438 as @xcite _ q== , _ q== , [ _ q , _ q]=. [ quadrature - def]note that the quadrature operators are often defined as @xmath439 and @xmath440 ; then their commutator is @xmath441 ; another possible definition @xcite is @xmath442 and @xmath443 ; then the commutator is @xmath444 . the definition ( [ quadrature - def ] ) gives simpler formulas for the average values for the coherent states , \|_q\|= re ( ) , \|_q\|= i m ( ) , which follow from the relation @xmath445 . the variance in this case is \|_q^2\|- \|_q\|^2 = \|_q^2\|- \|_q\|^2 = . the quadrature operator at an angle @xmath446 can be defined as _ _ q + _ q . * 6 . * an important property of a coherent state is that it splits into two _ unentangled _ coherent states after passing through a beam splitter , in full analogy with a classical optical wave or microwave . actually , so far we defined a coherent state only for a resonator , and it is not obvious how to introduce it for a propagating wave . we will not discuss how to do it rigorously @xcite , just implying that a piece of propagating wave can be described in a way , similar to a resonator description . there is a rather simple rigorous way to describe transformation of an arbitrary quantum state passing through a beam splitter ( see , e.g. , @xcite ) . the idea is essentially to write classical field relations , but for the annihilation operators ( conjugated relations are for the creation operators ) , then express the initial state via vacuum and creation operators of the input arms , and then substitute these input - arms operators with their expressions via output - arms operators . this gives the resulting output state . applying this procedure to a beam splitter with transmission and reflection amplitudes @xmath447 and input state @xmath448 , we obtain the output state @xmath449 , exactly as we would expect for a classical field . technically , this follows from the formula @xmath450 [ see eq . ( [ alpha - def-3 ] ) ] and relation @xmath451 , with commuting output - arms operators @xmath452 and @xmath453 , so that @xmath454 . note that if we apply coherent fields to both input arms , @xmath455 , then the resulting output state is also an unentangled product of classically - expected coherent states , @xmath456 , without an overall phase . we can think about field leakage from a microwave resonator to a transmission line through a `` mirror '' ( coupler ) as transmission through a beam splitter . therefore , from the discussed above property , if the initial state in the resonator is a coherent state @xmath395 , then it remains a coherent state @xmath457 , with @xmath35 given by the classical field evolution , ( t)=(0 ) e^-i_r t e^-t/2 , [ kappa - only]where @xmath5 is the energy dissipation rate and @xmath12 is the resonator frequency . we emphasize that this property is highly unusual for a quantum system ( thus indicating that coherent states are classical to a significant extent ) . dissipation usually leads to decoherence , so that an initially pure quantum state becomes a mixed state . in this case we have an exception : a pure state remains pure during the whole evolution . this makes quantum analysis very simple for an evolution involving coherent states . note that eq . ( [ kappa - only ] ) is still applicable when the energy loss rate @xmath5 has a contribution from intrinsic energy relaxation ( at zero temperature ) . now let us for a moment neglect the energy relaxation , and instead consider a classical drive with frequency @xmath22 and ( complex ) amplitude @xmath23 ( in some normalization ) . this is usually described by the hamiltonian = _ r ^ + e^-i_d t ^ + ^ * e^i_d t , which already assumes rotating wave approximation , requiring @xmath458 and sufficiently slowly changing drive @xmath23 . using this hamiltonian , we can find the evolution of an arbitrary quantum state of the resonator @xmath459 via the schrdinger equation @xmath460 . it is easy to see by solving this equation that if the initial state is a coherent state , then it remains a coherent state , though with a _ nontrivial overall phase _ @xmath461 , \|(t)= e^-i(t ) \|(t ) , so that the evolution is described by two equations , @xmath462 now let us combine the drive @xmath60 and dissipation @xmath5 . since both of them keep the state coherent ( with an overall phase ) , their combination will also keep it coherent ( with an overall phase ) . introducing the rotating frame based on the drive frequency , ( t ) e^i_d t ( t ) , from eqs . ( [ kappa - only ] ) , ( [ alpha - dot ] ) , and ( [ phi - dot-1 ] ) we obtain @xmath463 equation ( [ tilde - alpha - dot ] ) is the standard result for the evolution of a resonator under the drive and dissipation , while eq.([phi - dot-2 ] ) is usually not discussed in quantum optics , even though it is very important for quantum dynamics involving more than one coherent state ( for example , for measurement of a qubit in the circuit qed setup ) . note that eqs . ( [ tilde - alpha - dot])([phi - dot-2 ] ) rely on the fact that for coherent states the dissipation @xmath5 does not introduce decoherence and only brings the term @xmath464 into eq.([tilde - alpha - dot ] ) . we have derived this fact by considering the problem of a coherent state passing through a beam splitter . another ( lengthier ) way to prove it , is to consider the lindblad equation for the density matrix and to show that ( surprisingly ) a pure initial state remains pure if initially it was a coherent state . one of the ways to show it , is to separate the lindblad evolution into `` jump '' and `` no jump '' scenarios ( e.g. , @xcite ) . then the `` jump '' scenario ( application of operator @xmath465 ) brings no evolution because of eq . ( [ prop-1 ] ) , so all the evolution comes from the `` no jump '' scenario ( essentially the bayesian update ) , which keeps a coherent state coherent , with decreasing @xmath35 . this is why _ there is no randomness _ @xcite , normally leading to decoherence . note that the derivation via the lindblad equation can not easily reproduce important equation ( [ phi - dot-2 ] ) , because the overall phase is lost in the density matrix language . in this appendix we derive the results for _ phase back - action _ in the process of qubit measurement using the picture of vacuum noise , which is incident on the resonator from the transmission line ( fig . [ fig : vacuum ] ) . we assume the `` bad cavity '' limit and phase - sensitive amplification . the vacuum noise is treated in a simple classical way . let us start with assuming for simplicity that the resonator damping @xmath5 is only due to coupling with the transmission line carrying the outgoing wave , @xmath142 ; in particular , this requires @xmath143 ( later this assumption will be removed ) . then the vacuum noise enters the resonator only from the output line ( fig . [ fig : vacuum ] ) , and the wave equations for the resonator field @xmath44 and the outgoing field @xmath40 in the rotating frame based on the drive frequency @xmath22 are @xmath466 where @xmath467 is the vacuum noise , which is normalized in the same way as @xmath40 . in this normalization @xmath468 is the average number of photons in the resonator , while @xmath43 is the average number of propagating photons per second . note that the reflection coefficient in eq . ( [ appb - f ] ) is @xmath469 , while the transmission through the `` mirror '' is characterized by the coupling @xmath470 @xcite , as well as in eq . ( [ appb - alpha - dot ] ) . the drive term @xmath471 can also be written via the properly normalized incoming field @xmath472 as @xmath473 . also note that for the two qubit states we have slightly different resonator frequencies , @xmath474 ; however , in this appendix we will mostly use notation @xmath12 for brevity and because the resonator frequency shift is not important for the phase back - action , which is our focus here . in quantum optics the vacuum noise is treated as an operator @xcite with correlator @xmath475 , and eqs . ( [ appb - alpha - dot ] ) and ( [ appb - f ] ) are written for annihilation operators in the heisenberg representation . however , in our simple derivation we will treat the noise @xmath467 classically ( i.e. , as a complex number ) and consider evolution of classical fields ( which corresponds to the schrdinger picture ) . it is simple to see that the photon shot noise is properly reproduced if we assume that _ for any quadrature _ ( so that @xmath476 is real ) v_qu(t ) v(t)_qu= ( t - t ) , [ appb - v - qu - corr]which is equivalent to v(t ) v(t)^= ( t - t ) , v(t ) v(t)=0 , [ appb - v - corr]if @xmath477 is treated as a complex number , describing both quadrature components ( obviously , @xmath478 ) . for example , this relation can be obtained by considering a propagating wave @xmath479 with constant @xmath480 . then the fluctuating photon number @xmath481 within duration @xmath482 should have the same variance @xmath483\ , dt ' \|^2 \rangle$ ] as the mean @xmath484 . therefore , \| _ 0^t v_qu(t ) dt \|^2 = [ appb - v - qu - int]for the quadrature @xmath476 along @xmath480 , and eq . ( [ appb - v - qu - corr ] ) follows from ( [ appb - v - qu - int ] ) . note that eq . ( [ appb - v - qu - corr ] ) can be interpreted as following from the standard operator correlator , using the correspondence @xmath485 . as another check of this noise formalism , let us derive the correlator for the fluctuating number of photons in the resonator from eq . ( [ appb - v - qu - corr ] ) . using eq . ( [ appb - alpha - dot ] ) , we find the fluctuation ( t)= _ -^t e^-[/2+i(_r-_d)](t - t ) v(t ) dt , [ appb - dalpha]due to the noise @xmath467 . for a fixed stationary value @xmath486 , this leads to photon number fluctuation @xmath487 . then using eq . ( [ appb - dalpha ] ) , performing the double - integration using eq . ( [ appb - v - corr ] ) , and denoting @xmath488 , we find n(t ) n(t+)= \|n ( - \|\| ) , [ appb - dn - corr]which is the standard result for the photon number correlator @xcite . note that the photon number fluctuation decays with the rate @xmath25 instead of naively expected @xmath5 . it is also interesting to note that at time @xmath489 only the quadrature @xmath476 along @xmath490 with the fluctuations ( [ appb - v - qu - corr ] ) contributes to the correlator ( [ appb - dn - corr ] ) , while the orthogonal quadrature does not contribute . it is equally possible to say that the contribution comes only from the quadrature @xmath476 along @xmath491 , while the orthogonal quadrature does not contribute . also note that from eq . ( [ appb - dalpha ] ) we obtain \|\|^2= 1/2 , corresponding to the variance of @xmath492 for any quadrature . now let us consider the qubit measurement , assuming the `` bad cavity '' regime , as in sec . the fluctuation @xmath467 leads to the fluctuating ac stark shift _ q ( t ) = 2n = 4re [ _ st^ ( t ) ] with @xmath493 given by eq . ( [ appb - dalpha ] ) , and to the fluctuating outgoing field f(t ) = -v(t)+ ( t ) . by integrating these effects over the time period @xmath348 $ ] with @xmath494 , so that the exponential dependence in eq . ( [ appb - dalpha ] ) has sufficient time to fully decay , we find @xmath495 , \label{appb - dwq - int}\\ & & \int_t^{t+\tau } \delta f ( t')\ , dt ' = \frac{\kappa/2 -i(\omega_{\rm r}-\omega_{\rm d})}{\kappa /2 + i(\omega_{\rm r}-\omega_{\rm d } ) } \nonumber \\ & & \hspace{2.3 cm } \times \ , \int_t^{t+\tau } v(t')\ , dt ' . \label{appb - df - int}\end{aligned}\ ] ] we see that these fluctuating integrals are _ proportional _ to each other . obviously , the first integral determines the phase back - action on the qubit state , while the second integral is related to the measurement result . this is how we can relate the phase back - action to the measurement result . using eq . ( [ alpha - st ] ) for the steady - state values @xmath496 and @xmath181 corresponding to the qubit states @xmath6 and @xmath7 , and assuming @xmath497 , we find _ 1,st-_0 , st= _ st , and therefore from eqs . ( [ appb - dwq - int ] ) and ( [ appb - df - int ] ) we obtain @xmath498 . \label{appb - dwq - int2}\end{aligned}\ ] ] this relation shows that the phase back - action is determined by the output quadrature which is _ orthogonal _ to the informational quadrature along @xmath499 . note that the vacuum fluctuations @xmath467 , which produce the output fluctuations along the informational quadrature , do not affect the qubit state , so the corresponding evolution ( [ ph - sens - diag - bc ] ) of the qubit state ( diagonal matrix elements ) is only due to `` spooky '' back - action and can not be explained as an effect of @xmath467 . let us first consider an ideal phase - sensitive amplification of the `` orthogonal '' ( non - informational ) quadrature , so that @xmath126 [ see eq . ( [ phi - d - def ] ) ] . in this case we need to associate the output noise with the effect of @xmath467 fluctuations ( no added noise due to amplifier ) , and therefore _ t^t+ f_qu(t ) dt = , [ appb - dfqu - int]where @xmath110 is the measurement result [ eq . ( [ ph - sens - im - bc ] ) ] , @xmath259 is its variance , and real @xmath500 is the fluctuation along the measured quadrature . note that the left hand side of this relation is for a particular realization of the noise @xmath500 , while the last term in the right hand side assumes averaging over all noise realizations . since @xmath501 should have the usual vacuum noise statistics , we can use eq . ( [ appb - v - qu - int ] ) , which gives @xmath502 ^2 \rangle=\tau /4 $ ] . following eq . ( [ appb - dfqu - int ] ) , we can do the similar conversion for the response , \|_1,st -_0,st \| = . [ appb - response]finally , multiplying eqs . ( [ appb - dfqu - int ] ) and ( [ appb - response ] ) and noticing that this product corresponds to the right hand side of eq . ( [ appb - dwq - int2 ] ) multiplied by @xmath503 , we obtain _ t^t+ _ q ( t ) dt = , [ appb - dwq - int3]which is exactly the result for phase back - action @xcite presented in sec . [ sec : ph - sens - bc ] , when @xmath504 see eqs . ( [ ph - sens - off - bc ] ) , ( [ d - def - bc ] ) , and ( [ ph - sens - k - bc ] ) . the non - fluctuating part of the ac stark shift can be simply added . if we consider an ideal phase - sensitive amplification of an arbitrary quadrature , @xmath505 , then the derivation for the fluctuating phase shift @xmath506 is similar , except the amplified quadrature @xmath501 is no longer along @xmath507 , and therefore from eq . ( [ appb - dwq - int2 ] ) we obtain an extra factor @xmath508 , which appears in eq . ( [ ph - sens - k - bc ] ) but is absorbed by @xmath124 in eq . ( [ appb - dwq - int3 ] ) . however , it is not obvious if @xmath72 in eq . ( [ appb - dfqu - int ] ) should be counted from @xmath73 or from @xmath509 , and correspondingly if the phase back - action term in eq . ( [ ph - sens - off - bc ] ) should be @xmath510 or @xmath511\tau \}$ ] . we can find the answer by requiring that the phase shift due to the phase back - action term in eq . ( [ ph - sens - off - bc ] ) is zero on average . counterintuitively , the phase shift of the averaged @xmath512 in eq . ( [ ph - sens - off - bc ] ) is zero when the phase back - action term is @xmath510 , even though @xmath513 obviously has a non - zero phase if @xmath514 . this occurs due to a compensating effect from the first term in eq . ( [ ph - sens - off - bc ] ) , which contains @xmath285 and @xmath286 : for example , if @xmath515 , then a positive @xmath110 occurs more often , but produces smaller @xmath516 than for a negative @xmath110 . ( this somewhat counterintuitive compensation is related to the difference between the it and stratonovich approaches . ) thus , using the approach of the vacuum noise we derived the phase back - action term in eq . ( [ ph - sens - off - bc ] ) in the case of ideal phase - sensitive measurement . let us briefly discuss how in this approach we can take into account non - ideality due to additional resonator damping ( e.g. , because of coupling to other transmission lines ) and the loss of the microwave signal before it reaches amplifier ( which is still ideal ) . then eqs . ( [ appb - alpha - dot ] ) and ( [ appb - f ] ) can be replaced with @xmath517 \nonumber \\ & & \hspace{0.7 cm } + \sqrt{1-\kappa_{\rm col}/\kappa_{\rm out}}\ , v_{\rm add,2 } , \label{appb - f-2}\end{aligned}\ ] ] where @xmath518 is the vacuum noise entering the resonator from other transmission lines , the ratio @xmath90 characterizes the energy loss between the resonator and amplifier ( which can be modeled via a beam splitter ) , and because of this loss ( at zero temperature ) an additional vacuum noise @xmath519 contributes to the field @xmath40 , which reaches the amplifier . the noises @xmath520 , @xmath521 , and @xmath519 are uncorrelated and all satisfy eq . ( [ appb - v - corr ] ) ; then the noise of @xmath40 has the same statistics . the calculation becomes more complicated , but it still can be done explicitly . it shows that the correlation ( [ appb - dwq - int3 ] ) between the ac stark shift and the measurement result fluctuations is reduced by the factor @xmath522 , which is the same factor as for the reduction of @xmath124 . therefore , eq . ( [ appb - dwq - int3 ] ) and the corresponding eq . ( [ ph - sens - k - bc ] ) remain valid . analysis of imperfection due to a non - ideal amplifier can be performed as in ref . @xcite ; in this case eqs . ( [ ph - sens - off - bc ] ) and ( [ ph - sens - k - bc ] ) still remain valid . note that even though this approach based on vacuum noise gives a natural description of the physical mechanism responsible for the phase back - action , it still can not explain why in the ideal case with @xmath108 there are no fluctuations of the photon number in the resonator . the fact that in the ideal case only the observed quadrature fluctuates ( and the orthogonal quadrature does not fluctuate ) is a `` spooky '' property of quantum measurement and can not have a realistic interpretation . derivation of the phase back - action coefficient for the phase - preserving measurement can be done in a similar way . alternatively , as discussed in sec . [ sec : ph - pres - bc ] , the results for the phase - preserving case can be obtained from the results for the phase - sensitive measurement . j. y. mutus , t. c. white , e. jeffrey , d. sank , r. barends , j. bochmann , y. chen , z. chen , b. chiaro , a. dunsworth , j. kelly , a. megrant , c. neill , p. j. j. omalley , p. roushan , a. vainsencher , j. wenner , i. siddiqi , r. vijay , a. n. cleland , and j. m. martinis , appl . . lett . * 103 * , 122602 ( 2013 ) . e. jeffrey , d. sank , j. y. mutus , t. c.white , j. kelly , r. barends , y. chen , z. chen , b. chiaro , a. dunsworth , a. megrant , p. j. j. omalley , c. neill , p. roushan , a. vainsencher , j. wenner , a. n. cleland , and j. m. martinis , phys . * 112 * , 190504 ( 2014 ) .	we consider continuous quantum measurement of a superconducting qubit in the circuit qed setup with a moderate bandwidth of the measurement resonator , i.e. , when the `` bad cavity '' limit is not applicable . the goal is a simple description of the quantum evolution due to measurement , i.e. , the measurement back - action . extending the quantum bayesian approach previously developed for the `` bad cavity '' regime , we show that the evolution equations remain the same , but now they should be applied to the entangled qubit - resonator state , instead of the qubit state alone . the derivation uses only elementary quantum mechanics and basic properties of coherent states , thus being accessible to non - experts .
in many applications and implementations of quantum information processing , one has to compare two different quantum states . in this context , the quantum fidelity @xcite is a very useful tool to measure the `` closeness '' between two states in the hilbert space of a quantum system . for two arbitrary states @xmath0 and @xmath1 , it is defined as @xmath2 for any pair of pure states @xmath3 and @xmath4 , the quantum fidelity reduces to their ( squared ) overlap , @xmath5 . although the fidelity does not define a metric on the state space , it is the core ingredient for several of them , like for instance the bures distance @xmath6^{1/2}$ ] @xcite . the fidelity is also widely used to define various entanglement monotones . the bures distance @xmath7 is itself such an example @xcite . for multipartite pure states , the geometric measure of entanglement @xmath8 with @xmath9 is another example that exploits the maximal fidelity between the state @xmath3 to characterize and the set @xmath10 of all fully separable states @xmath11 @xcite . the question arises as to how the supremum in eq . ( [ fs ] ) can be computed . in general , it is known to be an np - hard task @xcite . for multipartite states @xmath12 that are symmetric with respect to the permutations of the parties it has been shown @xcite that this supremum is realized among the symmetric separable states @xmath13 only , @xmath14 in fact it can even be proven that for three or more particles the state maximizing the overlap in the definition of @xmath15 is necessarily symmetric @xcite . this nice property considerably simplifies the calculation of the geometric measure of entanglement for symmetric states . the maximization of the fidelity over sets other than the separable states @xmath10 has proven to be very useful for discrimination strategies of inequivalent classes of multipartite entangled states with witnesses @xcite or other methods @xcite . in this case , the maximization is typically to be performed on sets of states equivalent through either local unitary operations ( lu ) or stochastic local operations assisted by classical communication ( slocc ) . one needs to evaluate the maximal fidelity @xmath16 with @xmath17 any considered lu or slocc class of states . mathematically , these classes are defined as follows : the lu equivalence class of a pure state @xmath18 is given by all states of the form @xmath19 , where the @xmath20 are unitary matrices acting on the @xmath21-th party . the slocc equivalence class of @xmath22 is given by normalized states of the form @xmath23 , where the @xmath24 are invertible matrices @xcite . the lu and slocc equivalence classes of states never coincide , except for the fully separable states that are all both lu and slocc equivalent . in eq . ( [ fc ] ) , if @xmath17 contains symmetric states , the question naturally arises whether the simplification given by eq . ( [ fss ] ) for the particular case of the slocc ( and lu ) class @xmath10 of separable states generalizes similarly . in other words , do we have for any symmetric state @xmath12 and any lu or slocc classes @xmath17 containing symmetric states @xmath25 this paper provides answers to this question for multiqubit systems . first , in the case of lu classes , the answer is positive and this is formally proven in sec . [ luclasssec ] . second , for the case of slocc classes , the answer is surprisingly negative and spectacular violations of eq . ( [ fcc ] ) will be given in sec . [ sloccclasssec ] . in sec . [ conclusionsection ] we summarize and discuss further open problems . when considering lu equivalence classes @xmath17 and multiqubit systems in eq . ( [ fcc ] ) and since any two lu - equivalent symmetric states can be transformed into each other with the same local unitary acting on each party @xcite , the question can be rephrased as follows : do we have , for any @xmath26-qubit symmetric states @xmath12 and @xmath27 , @xmath28 where @xmath29 is the group of unitary matrices of dimension @xmath30 ? since only the absolute value of the overlap matters , we can choose the phases of the @xmath20 as we like . so , it suffices to take matrices with determinant @xmath31 and consider @xmath32 with @xmath33 the group of unitary matrices of determinant 1 . the question ( [ supslu2 ] ) can thus be rephrased as @xmath34 to tackle this problem , we note that an arbitrary @xmath33 matrix can be written as @xmath35 we then define the function @xmath36 with @xmath37 where @xmath38 and @xmath39 are considered as _ real _ hilbert spaces . the function @xmath40 is symmetric with respect to the permutations of the variables , because @xmath12 and @xmath27 are symmetric states . furthermore , it is @xmath41-multilinear in the coefficients @xmath42 , that is , @xmath43 the multilinearity is only ensured for real @xmath44 and @xmath45 and this is the reason why we have to consider the _ hilbert spaces @xmath46 and @xmath47 . under these conditions , hrmander s theorem 4 of ref . @xcite and its extension to the case of real hilbert spaces of any dimension @xcite can be applied and we have @xmath48 where @xmath49 . because of the @xmath41-multilinearity of @xmath40 , @xmath50 with @xmath51 and thus @xmath52 equation ( [ horm ] ) then yields @xmath53 that is to say , @xmath54 so , the answer to question posed in eq . ( [ supslsu2 ] ) and in eq . ( [ supslu2 ] ) is clearly positive . this finishes the proof . now we turn to the case when slocc classes are considered in eq . ( [ fcc ] ) . since any two slocc - equivalent symmetric states can be transformed into each other with the same invertible local operation ( ilo ) acting on each party @xcite , the question addressed here for multiqubit systems is the following : do we have , for any @xmath26-qubit symmetric states @xmath12 and @xmath27 , @xmath55 where @xmath56 is the group of invertible matrices of dimension @xmath30 ? here , the expression to be maximized contains a normalization constant that also depends on the state @xmath27 . contrary to the lu case , the left - hand side term of eq . ( [ supslocc ] ) can not be cast in a multilinear form divided by a product of norms like the left - hand side term of eq . ( [ horm ] ) . hrmander s theorem therefore can not be exploited to tentatively prove eq . ( [ supslocc ] ) . actually , the counterexamples to this equation identified hereafter prove that any attempt in this direction would be doomed to failure . as the first observation , for three - qubit systems , extensive numerical simulations showed no violation of eq . ( [ fcc ] ) . similarly , numerical simulations for @xmath26 up to 8 gave indications that eq . ( [ fcc ] ) seems also to hold for the classes of states slocc - equivalent to the @xmath31-excitation dicke states @xmath57 @xcite , hereafter denoted by the classes @xmath58 . these classes gather both symmetric and nonsymmetric states . when restricted to the symmetric subspace , they merely identify to the @xmath59 families of symmetric states of ref . this brings us to conjecture that the equality @xmath60 actually holds for any @xmath26 and any symmetric state @xmath12 . the generalization of eq . ( [ conj ] ) to arbitrary slocc classes containing symmetric states , however , is not correct . spectacular violations are obtained when considering the classes of states slocc - equivalent to the @xmath21-excitation dicke states @xmath61 @xcite , hereafter denoted by the classes @xmath62 , for @xmath63 and @xmath64 . all these classes gather both symmetric and nonsymmetric states . when restricted to the symmetric subspace , they identify to the @xmath65 families of symmetric states of ref . @xcite . for the values of @xmath26 and @xmath21 considered , there are symmetric states @xmath12 for which @xmath66 to prove this result , we first consider the state @xmath67 . for all aforementioned @xmath26 and @xmath21 , @xmath68 @xcite and one gets ( see appendix [ apa ] for a detailed calculation ) @xmath69}}_{y \in [ -1,1 ] } f(x , x',y)\ ] ] with @xmath70^{j}}.\ ] ] equation ( [ rsnk ] ) can not be reduced analytically . it can , however , be straightforwardly evaluated numerically . we illustrate it in fig . [ fsdn1dnmkk ] for @xmath71 and @xmath72 . the figure clearly shows that @xmath73 remains significantly below one with an asymptotic behavior as @xmath26 tends to infinity for fixed @xmath21 . for fixed @xmath26 , the considered fidelities decrease with increasing @xmath21 . the largest ones are obtained for @xmath74 with an asymptotic value for large @xmath26 of @xmath75 . by contrast , surprisingly we have @xmath76 which means that the dicke state @xmath57 can be approached as closely as desired by non - symmetric slocc inequivalent @xmath62 states . we thus clearly have @xmath77 and this is a neat violation of eq . ( [ fcc ] ) . to prove eq . ( [ maxrnk ] ) , we define for any @xmath78 the non - symmetric @xmath62 state @xmath79 with @xmath80 the @xmath26-qubit non - symmetric local operation @xmath81 where , in the computational basis @xmath82 , @xmath83 the state @xmath84 belongs to the @xmath62 slocc class since for any @xmath78 @xmath85 is an invertible local operation : @xmath86 the state @xmath84 is also non - symmetric since a non - symmetric ilo acting on @xmath61 always yields a non - symmetric state for @xmath87 @xcite . a detailed calculation then gives , for any @xmath78 @xcite , ( see appendix [ apb ] ) @xmath88 where @xmath89 and @xmath90 we have @xmath91 and thus @xmath92 this implies that @xmath93 and this proves eq . ( [ maxrnk ] ) . these results show that for any @xmath63 and @xmath94 , a non - symmetric ilo of the form @xmath95 can transform the dicke state @xmath61 into a non - symmetric state located as closely as desired to the dicke state @xmath57 , even though @xmath61 and @xmath57 are slocc inequivalent . this result can not be achieved when only symmetric slocc operations are considered . as an aside and out of curiosity , it is interesting to study the inverse operation @xmath96 acting on the @xmath57 state . while one obviously has @xmath97 the state @xmath98 reads , up to a normalization constant , @xmath99 , \nonumber\end{aligned}\ ] ] this state is totally independent of @xmath100 and differs significantly from @xmath61 . it has only components onto states with at least @xmath101 excitations . therefore , except for @xmath102 , @xmath103 has no overlap with @xmath61 . for @xmath102 , the fidelity with the state @xmath61 amounts only to @xmath104 . @xmath105 \hline \hline 4 & 2 & 0.5 & 7 & 2 & 0.457 \\ 5 & 2 & 0.477 & & 3 & 0.383 \\ 6 & 2 & 0.465 & 8 & 2 & 0.451 \\ & 3 & 0.4 & & 3 & 0.372 \\ & & & & 4 & 0.344\\ \hline \end{array}\ ] ] equation ( [ maxrnk ] ) can even be generalized . any state @xmath106 of the @xmath58 slocc class satisfies @xmath107 indeed , for such states an ilo @xmath108 can be found such that @xmath109 we then define for any @xmath78 the @xmath62 states @xmath110 with @xmath111 as defined by eq . ( [ psink2 ] ) . we get trivially , up to a normalization constant , @xmath112 and thus @xmath113 it follows that @xmath114 and this implies eq . ( [ genn1 ] ) . this result shows that all states of the @xmath58 slocc class can be approached as closely as desired by non - symmetric states of any of the @xmath62 slocc classes , for @xmath115 . topologically , this means that the @xmath58 slocc class of states lies at the boundary of the non - symmetric side of any of the @xmath62 classes . this result sheds an additional light on the general topology of the slocc classes of multiqubit systems , whose restriction on the only symmetric subspace was established in ref . @xcite . the converse of the previous result is by far not true : the states of the @xmath62 slocc classes can not be approached as closely as desired by @xmath58 states , even if they are non - symmetric . as an example , a detailed calculation yields ( see appendix [ apa ] ) @xmath116^{n - k-1 } \left[\tilde{k } + ( 1 - 2\tilde{k } ) \tilde{k}_r \right ] \ ] ] with @xmath117 and @xmath118 and extensive numerical simulations [ see our conjecture in eq . ( [ conj ] ) ] showed that this should also correspond to @xmath119 . we exemplify some values of @xmath120 in table [ tabi ] . it is noteworthy to mention that these fidelities decrease with increasing @xmath26 and @xmath21 . for fixed @xmath21 , we have @xmath121 explicitly , for @xmath122 , this limit reads @xmath123 , @xmath124 , @xmath125 , @xmath126 , @xmath127 , @xmath128 , and @xmath129 , respectively . all these results clearly show that while the @xmath57 state can be approached as closely as desired by non - symmetric states that are slocc equivalent to any of the @xmath61 states ( @xmath64 ) , none of these latter can be closely approached by any state slocc equivalent to the dicke state @xmath57 . in this paper we have analyzed the maximization of the quantum fidelity between symmetric multiqubit states and sets of lu- or slocc - equivalent states that contain symmetric states . we have shown that the open question in eq . ( [ fcc ] ) admits a positive answer when lu classes of qubit states are considered , while the answer is negative when turning to slocc classes of states . in the case of lu sets , the positive answer simplifies considerably the calculation of the desired maximal overlap . for slocc classes of states , we have shown significant violations of eq . ( [ fcc ] ) where for some states @xmath12 and classes @xmath17 the fidelity @xmath130 takes the maximal possible value 1 , while the symmetric restriction @xmath131 has only significantly much lower values . this is in particular the case when considering any states @xmath12 of the @xmath58 slocc class in combination with any of the @xmath62 slocc classes , for @xmath63 and @xmath64 . finally , extensive numerical simulations have also lead us to conjecture that eq . ( [ fcc ] ) seems to be correct when the considered slocc class @xmath17 identifies to @xmath58 , whatever the state @xmath12 . there are several directions in which our work can be generalized . first , concerning lu equivalence classes , it would be highly desirable to prove our result also for higher - dimensional systems and not only for qubits . in our proof , we made use of the simple parametrization of @xmath33 matrices , which is not so simple in higher dimensional systems . second , for slocc equivalence classes it would be very useful to find out under which additional conditions the optimization over symmetric states is enough . based on numerical evidence we identified some examples where this seems to be the case , but so far no clear understanding has been reached . from a more general perspective , our work presents examples where symmetries can help to solve optimization problems related to the numerical range @xcite . understanding further the role of symmetry in such problems is clearly a challenging task , nevertheless it will have a significant impact on various problems in quantum information theory . a.n . acknowledges a fria grant and the belgian f.r.s .- fnrs for financial support . o.g . acknowledges financial support from the fqxi fund ( silicon valley community foundation ) , the dfg , and the erc ( consolidator grant no . 683107/tempoq ) . tb acknowledges financial support from the belgian f.r.s .- fnrs through iisn grant no . 4.4512.08 . @xmath136}}_{y \in [ -1,1 ] } \frac{\sum_{j=0}^{k ' } \left[(2 - \delta_{j,0 } ) \sum_{j ' = j}^{k ' } c_{j'}(x , x ' ) c_{j'-j}(x , x')\right ] t_j(y)}{\sum_{j=0}^{k ' } c_{k'}^j c_{n - k'}^j \left [ x^2 x'^2 + ( 1-x^2)(1-x'^2 ) + 2 y x x ' \sqrt{(1-x^2)(1-x'^2)})\right]^{j}},\ ] ] where @xmath137 is the binomial coefficient between @xmath138 and @xmath139 , with the usual convention @xmath140 if @xmath141 or @xmath142 , @xmath143 denotes the kronecker delta , @xmath144 is the @xmath145-th degree chebyshev polynomial of the first kind , and , for @xmath146 , @xmath147 in particular , for @xmath148 , one gets the well known result @xcite @xmath149 } x^{2k } ( 1-x^2)^{n - k } \nonumber \\ & = c_n^k \tilde{k}^k \left(1-\tilde{k}\right)^{n - k},\end{aligned}\ ] ] with @xmath117 the fractional excitation of the dicke state @xmath61 . for @xmath150 , one gets @xmath151}}_{y \in [ -1,1 ] } \frac{c_0(x , x')^2 + c_1(x , x')^2 + 2 c_0(x , x ' ) c_1(x , x ' ) y}{1 + ( n-1)\left[x^2 x'^2 + ( 1-x^2)(1-x'^2 ) + 2 y x x ' \sqrt{(1-x^2)(1-x'^2)}\right ] } \nonumber \\ & = c_n^k \tilde{k}_r^{k-1 } \left ( 1-\tilde{k}_r \right)^{n - k-1 } \left[\tilde{k } + ( 1 - 2\tilde{k } ) \tilde{k}_r \right],\end{aligned}\ ] ] with @xmath152 for @xmath153 , eq . ( [ rsnkkp ] ) can not be reduced analytically and it must be evaluated numerically . we noticed that for all tested cases the supremum was systematically obtained for @xmath154 . to prove eq . ( [ rsnkkp ] ) , we first observe that in the computational basis the dicke state @xmath61 merely reads @xmath155 , where the sum runs over all multiqubit states with any @xmath21 qubits in the @xmath156 state and the remaining @xmath157 qubits in the @xmath158 state . we then note that any symmetric state of the @xmath159 slocc class , i.e. , states of the @xmath160 family , can be written in the form @xcite @xmath161 , where @xmath162 is a normalization constant and the sum runs over all multiqubit states with any @xmath163 qubits in an @xmath164 state and the remaining @xmath165 qubits in a distinct @xmath166 state . writing the single qubit states @xmath164 and @xmath167 in the form @xmath168 and @xmath169 ( @xmath170 , \phi,\phi ' \in [ 0,2\pi[$ ] ) , respectively , yields @xmath171 with @xmath172 , @xmath173 and @xmath174 . we then get @xmath175 } } _ { \delta \phi \in [ 0,2\pi [ } \mathcal{n}^2 c_n^k \left\|\sum_{j=0}^{k ' } c_j(x , x ' ) e^{i j \delta \phi}\right\|^2,\ ] ] with @xmath176 as given by eq . ( [ cjxxp ] ) . finally , using the identity @xmath177 and setting @xmath178 yields eq . ( [ rsnkkp ] ) . we then define the unnormalized dicke states @xmath182 , which satisfy @xmath183 . \label{eq : decomposition_gd}\end{aligned}\ ] ] inserting eq . ( [ bnkeps ] ) in eq . ( [ eq : decomposition_gd ] ) and observing that for any @xmath26 and @xmath21 the symmetric state @xmath184 reads in the dicke state basis @xmath185 yields straightforwardly @xmath186 with @xmath187 , @xmath188 , and @xmath189 as given by eq . ( [ ajbjcj ] ) and @xmath190 . in the sum over @xmath145 in eq . ( [ gnkeps ] ) , the first term @xmath191 merely yields the state @xmath192 , which is nothing but the dicke state @xmath57 [ see eq . ( [ unkrec ] ) ] . the rest of the sum from @xmath193 to @xmath194 is by definition the state @xmath195 [ eq . ( [ psieps ] ) ] . we thus get @xmath196 from which eq . ( [ psink2 ] ) immediately follows for any @xmath78 . for @xmath197 , @xmath198 and the normalized state @xmath199 is not defined . c. d. cenci , d. w. lyons , s. n. walck , arxiv:1011.5229 ; _ theory of quantum computation , communication and cryptography _ , edited by d. bacon , m. martin - delgado , and m. roetteler , lecture notes in computer science , vol . 6745 ( springer , berlin , 2014 ) , p. 198 . we recall that the @xmath21-excitation dicke states ( @xmath200 ) are defined as @xmath201 , where @xmath202 denotes the binomial coefficient @xmath203 , the multiqubit states in the sum contain @xmath21 qubits in state @xmath156 , and @xmath204 denotes all permutations of the qubits leading to different terms in the sum . all @xmath205 dicke states ( @xmath200 ) are symmetric and they form an orthonormal basis in the symmetric subspace of the multiqubit system . for @xmath206 , where @xmath207 denotes the floor function , the dicke states @xmath61 are slocc inequivalent between each others @xcite . in contrast , the @xmath208 and @xmath61 states are lu equivalent . although the @xmath26-qubit local operation @xmath80 remains defined for @xmath197 , though not invertible in this case , the normalized state @xmath199 is in contrast not defined in this case since @xmath209 ( see appendix [ apb ] ) .	for two symmetric quantum states one may be interested in maximizing the overlap under local operations applied to one of them . the question arises whether the maximal overlap can be obtained by applying the same local operation to each party . we show that for two symmetric multiqubit states and local unitary transformations this is the case ; the maximal overlap can be reached by applying the same unitary matrix everywhere . for local invertible operations ( stochastic local operations assisted by classical communication equivalence ) , however , we present counterexamples , demonstrating that considering the same operation everywhere is not enough .
research interest in the origins of the long - range bidirectional movement of particles ( organelles , vesicles , nutrients ) driven by molecular motors is motivated by fundamental questions concerning the nature of interactions between motors and their cargos as transport processes take place . a current explanation for the phenomenon relies on the idea that motors of different polarities act coordinately on the same particle at different times . if , however , they act in parallel , the bidirectional movement would reflect dominance of one or another kind of motor achieved by a _ tug - of - war _ mechanism @xcite , @xcite , @xcite , @xcite , @xcite . an important question that remains in this context concerns the mechanisms that would promote such coordination @xcite . alternatives to the coordination or _ tug - of - war _ models in the literature arise from the possibility of attributing the phenomenon to a dynamic role of the microtubules @xcite or to a mechanical coupling between different motors @xcite . a general difficulty encountered within any of these views is related to the presence of other particles ( including other motors ) on the microtubule at a given time that are not directly involved with the transfer process . these other particles are expected to impose restrictions on motility and performance of the motors that are directly interacting with cargo at that time @xcite . contrarily to these expectations , however , data from observations of beads driven by kinesins in steady - state conditions indicate that the number of long length runs of such beads increases significantly as the density of motors at the microtubule increases , although their velocities remain essentially unaltered within a wide range of motor concentrations @xcite , @xcite . thus , the reality of traffic jam in crowded microtubules still challenges the current view of long - range cargo transport that presupposes an effective and controllable movement of the motor(s ) arranged into a motor - cargo complex . this , of course , requires a certain degree of stability of motor - cargo interactions and motor processivity . our intention here is to discuss these problems from a different perspective by bringing into this scenario the model introduced in @xcite to examine cargo transport as a _ hopping _ process . according to that , motors and cargos would not assemble into complexes to put transport into effect . on the contrary , each motor would function as an active overpass for cargo to step over to a neighboring motor . in this case , the long - range movement of cargo is envisaged as a sequence of these elementary ( short - range ) steps either forwards or backwards . in @xcite we examined the conditions under which this may happen , accounting for the fact that motor motility is affected by the interactions with other motors and with cargos on the microtubule . there , we considered the presence of a collection of interacting motors , all of them presenting the same polarity ( kinesins may be thought of as prototypes ) and a single cargo . here , we examine whether it is possible to explain in a similar context the origin of the observed bidirectional movement displayed by cargos . the particular mechanism we propose to substantiate the hopping differs from that suggested in @xcite . it keeps , however , the same general ideas of the original . as it will be explained below , we view the hopping of cargo between motors as an effect of thermal fluctuations undergone by motor tails . the flexibility of the tails may promote contact and , eventually , exchange of cargo between neighboring motors . as in @xcite , the model dynamics is mapped into an asymmetric simple exclusion process ( asep ) @xcite , @xcite , @xcite whose stationary properties are resolved explicitly in the limit of very large systems . other asep models have already been considered in the literature to study the conditions for motor jamming in the absence of cargo @xcite , @xcite , @xcite . our model is conceived to account explicitly for changes in the dynamics of the motors that at a certain instant of time are interacting with cargos . the model is reviewed here in order to include a second cargo in the system , still keeping the presence of motors of a single polarity . we believe that this approaches more realistic situations in which the simultaneous presence of many cargos and motors on the same microtubule must be the prevailing situation @xcite . we show that under these conditions , a cargo may be able to execute long - range bidirectional movement as it moves over clusters of motors assembled either at its back end or at the back end of the cargo in front . one may recognize in this a possibility for explaining the origins of self - regulation in intracellular transport since it has been suggested in the last few years that signaling pathways involved in intracellular traffic regulation can be performed simply by the presence of cargos at the microtubule @xcite . we then speculate that the passage of cargos on microtubules does not get blocked by motor jamming . on the contrary , jamming operates as an allied process to promote long runs of cargos across motor clusters . in this case , the density of motors on the microtubule can be identified as an element of control in intracellular transport since it directly affects the conditions for jamming . it is worth mentioning that the model developed here does not rule out other possibilities , such as the _ tug - of - war _ or competition models . what we suggest is that the presence of motors of different polarities may not be essential to explain the origin of the bidirectional movement . the hopping mechanism is presented in sec.2 . the kinetic properties of the extended version are developed in sec.3 , considering the presence of two cargos . in sec.4 we present our results . additional remarks and conclusions are in sec.5 . the stochastic model in @xcite formulated in a lattice describes the dynamics of motors and cargos accounting for ( i ) steric interactions among different particles moving on the same microtubule ; ( ii ) the presence of motors of a single polarity ; ( iii ) the fact that cargos do not move if not driven by motors . the crucial point is in item ( iii ) because it requires a specific model for motor - cargo dynamics as transportation takes place . we offer here a slightly different view from that in @xcite keeping however the reliance on the ability of motors to transfer cargo . one way by which this may be achieved is sketched in _ figure ( 1)_. the stepping of cargo would be accomplished as it is released from a motor to which it is attached at a certain instant of time and then get attached to another ( neighboring ) motor either at the left or at the right see _ figure ( 1a)_. this process , like the one discussed in @xcite , relies strongly on the flexibility of the motor 's tail . the idea was inspired by experimental results suggesting a dynamic role of the kinesin 's coiled - coil segment in the process @xcite , @xcite and also by data indicating that under load , kinesin motors display an oscillatory movement @xcite @xcite . here , we think of these oscillations as signaling fluctuations in the position of motor 's tail , not necessarily being correlated with displacement of its center of mass . if this is the case , such oscillations would promote contact between neighboring motors favoring cargo exchange . accordingly , long - range displacements of cargo would reflect a hopping process extended over many neighboring motors which may be accomplished if these motors get jammed into clusters for sufficient long periods of time.notice that the whole mechanism does not require special stability of cargo - motor binding . on the contrary , the transfer of cargo by a motor would be ease by a loose attachment between them . the model in @xcite was resolved explicitly considering the presence of a single cargo in the system . the averages for the quantities of interest were determined in steady - state conditions . we showed there that the long - range displacements of the cargo would occur predominantly in the backwards direction , i.e. in opposition to the direction of the movement of the considered motors . we shall show here that the same dynamics may lead cargo to display bidirectional movement if the system contains at least one more cargo interfering with the movement of the motors . to examine the properties of the system containing two cargos and an arbitrary number of motors , we map it into the same asep as in goldman1 whose dynamics can also reproduce the scheme in _ figure(1)_. we consider a one - dimensional lattice with @xmath0 sites , representing the microtubule with periodic boundary conditions . this system contains @xmath1 motors and a number @xmath2 of other particles - the cargos - that interact with motors in order to move . each site can be occupied by a motor or by a motor attached to a cargo _ ( see fig.1(a ) ) _ , otherwise it is empty . the total number of sites that remain unoccupied is @xmath3 . here , we analyze the long - time behavior of this system for @xmath4 and determine the average cargo velocity as a function of the parameters . the results indicate conditions for cargos to perform a type of long - range movement that share the characteristics of the observed bidirectional movement . the map of the dynamics shown in _ figure.1 _ into the considered asep is carried out as follows . _ _ _ _ first , each site is identified by its position @xmath5 at the lattice . then , to each of these sites is associated a variable @xmath6 that assumes integer values @xmath7 or @xmath8 such that @xmath9 if the site @xmath10 is empty , @xmath11 if it is occupied by a motor ; or @xmath12 if it is occupied by a motor attached to a cargo . with these , a configuration @xmath13 of the lattice is specified by the set @xmath14 @xmath15 the dynamics of the asep that reproduces the elementary steps in _ fig.1 _ can now be defined . for this , consider that at each time interval @xmath16 a pair of consecutive sites , say @xmath10 and @xmath17 are selected at random . the occupancy of these two sites is then switched according to the following rules @xmath18where the pair @xmath19 is represented by the values of the corresponding site variables @xmath20the parameters @xmath21 and @xmath22 are the assigned probabilities per unit time ( rates ) for occurrence of the processes indicated . process @xmath23 describes the possibility for a motor ( kinesin ) that carries no cargo to step forward to a neighboring empty site _ ( figure 1b)_. processes @xmath24 and @xmath25 account for the switching of the cargo between two neighboring motors . this accounts either for backward @xmath24 or for forward steps notice that the dynamics conserves the number of particles of type-1 as well as those of type-2 . in order to investigate the long - time dynamics of a cargo resulting from these elementary steps we use the _ matrix _ _ ansatz _ introduced by derrida @xcite , @xcite . the idea is to represent the probability @xmath26 of a configuration @xmath13 of the system with @xmath1 sites and @xmath0 particles of type-1 as a trace over a product of @xmath1 non - commuting matrices , each specifying the corresponding site occupancy:@xmath27where @xmath28is the normalization . the sum runs over all configurations for which @xmath29 and @xmath30 . in this product , a site @xmath31 is represented by a matrix @xmath32 if it is occupied by a motor ( @xmath33 ) or by a matrix @xmath34 if occupied by a motor with a cargo ( @xmath35 ) ; if the site is empty it is represented by a matrix @xmath36 ( @xmath37 ) . in order to calculate averages over these configurations in the stationary state , it is necessary at first to find the _ algebra _ that must be satisfied by these matrices such that the probabilities defined in ( [ prob do estado ] ) satisfy the stationary conditions @xcite , @xmath38 where the sum extends over all configurations of @xmath0 motors distributed over @xmath39 lattice sites . observe that the nonzero terms on the lhs of the above equation are those for which configurations @xmath13 and @xmath40 differ from each other at most by the positions of a pair of consecutive sites , which can be reversed by any of the elementary processes defined by the dynamics in ( [ dinamica ] ) . in this case , each factor @xmath41 ( or @xmath42 ) must be replaced by the rate @xmath43 , @xmath44 or @xmath22 for the corresponding elementary process that brings @xmath13 back from @xmath40 ( or @xmath40 from @xmath13 ) . the algebra corresponding to the asep defined by the dynamics ( [ dinamica ] ) has been presented in ( @xcite ) for @xmath45 @xmath46 with @xmath47here , we shall use this same algebra to evaluate the traces over products of matrices @xmath48 @xmath34 and @xmath36 that appear in calculating averages over the quantities that characterize the movement of a cargo . before proceeding , however , a few remarks are in order . the model with two cargos is not ergodic . the dynamics preserves the number of empty spaces in each of the two partitions defined by the initial positions of the two cargos in the system with periodic boundary conditions . in this case , all configurations @xmath13 and @xmath49 that satisfy equation ( [ mestra ] ) must share the number of empty spaces in each of the partitions . moreover , configurations in which the empty spaces are all concentrated in one of the two partitions must be excluded , for these do not satisfy ( [ mestra ] ) with the algebra ( [ algebra ] ) . we treat the initial conditions ( _ ic _ ) , namely the number of empty spaces - @xmath50 in one of the partitions and @xmath51 in the other , as _ random variables_. this artifact shall account for the uncertainty one has in experimental data regarding the relative positions of the particles , and also for effects of random processes that are not explicitly described by the present model such as motor binding and unbinding at the microtubule . for computing averages , we shall account first for all possible configurations at fixed @xmath50 and then average the results over @xmath50 . the procedure is further specified observing that ( a ) because there is no reason to favor any initial configuration , we may consider that @xmath50 is uniformly distributed and ( b ) in analogy with a situation of equilibrium , we take the average _ annealing _ as the averages over particle configurations are performed in parallel with average over @xmath52 the measure @xmath53 of a configuration @xmath54 of a _ subset_-@xmath50 is written as the trace over a product of matrices @xmath55 and @xmath36 that satisfy the algebra in ( [ algebra ] ) : @xmath56 in the expression above , each @xmath57 is a binary variable such that @xmath58 @xmath59 and @xmath60 satisfying @xmath61 the normalization @xmath62is conveniently expressed in terms of the weights @xmath63 . these are defined as the sum over the traces corresponding to the configurations that belong to the subset @xmath50 for which the occupation of the sites ... @xmath64 ... and ... @xmath65 ... in the _ n - tuples _ are _ _ _ _ fixed and _ _ _ _ specified by the values of the corresponding site variables @xmath66 and @xmath67 respectively . the @xmath53 defined above must satisfy the stationary conditions @xmath68where the sum extends over all configurations @xmath69 that belong to the subset @xmath50 . the transition rates @xmath70 lead configurations @xmath69 into configurations @xmath54 . consistently with the above definitions we represent the average value@xmath71 of the velocity of any of the two cargos at fixed @xmath50 as @xmath72the configurations associated with @xmath73 are such that the specified cargo has one motor at its right side that allows it to move one step to the right at a rate @xmath22 . similarly , in the configurations associated to @xmath74 there is a motor at the left side of this cargo that allows it to move one step to the left at rate @xmath43 . in both types of configurations the neighborhood of the other cargo is not specified . it shall be convenient to subdivide the above sum into sums over configurations having the same trace . this is achieved by specifying in ( vmedio(a)2h ) the occupation of the sites that precede both cargos . for this , @xmath75 is rewritten as @xmath76 \right . \notag \\ & & \left . -w\left [ w_{\left ( 12\right ) -\left ( 12\right ) } ^{(h)}+w_{\left ( 12\right ) -\left ( 02\right ) } ^{(h)}\right ] \right\ } \label{vmedio(b)2h}\end{aligned}\ ] ] to proceed in the evaluation of ( [ vmedio(b)2h ] ) it is also convenient to replace the site variables @xmath77 by block variables @xmath78 and @xmath79 @xmath80 , that assume integer values to represent , respectively , sequences of motors and empty sites in a configuration @xmath81 with these , the sum over configurations that contribute to @xmath82 for example , can be expressed as @xmath83with @xmath84 . here , @xmath85 @xmath86or @xmath87 indicates a product of @xmath88 @xmath86or @xmath89 matrices @xmath32 @xmath86or @xmath90 . the symbol on the summation signals indicates that these are restricted to the configurations that satisfy the constraints in ( [ vinculo ] ) for @xmath91 . all the traces in the rhs of ( [ vmedio(b)2h ] ) can now be reduced with the aid of the algebra in ( [ algebra ] ) . for this , we use the identity @xmath92that also follows directly from ( [ algebra ] ) @xcite . the results are quoted as follows@xmath93 the above expressions are independent of @xmath50 . the only dependence on @xmath50 in the evaluation of the weights comes from the multiplicity of the configurations . configurations for which @xmath50 * * * * @xmath94 * * or * * @xmath95 * * * * contribute with a factor @xmath96 with respect to the contributions from all other configurations that result in the same trace . we now take the average of @xmath75 over all realizations of @xmath50 that assumes an integer value within the interval @xmath97 $ ] for @xmath98 , with equal probability . this is performed here as@xmath99that corresponds to the average _ annealing _ in analogy to a situation of equilibrium . the averaged quantities are indicated by the bars over the corresponding symbols representing the weights @xmath100 and normalization @xmath101 . now , observe that because the traces do not depend on @xmath50 , then the sums in the above expression , both in the numerator and in the denominator , account for all possible configurations of arbitrary sequences of empty and occupied sites , keeping fixed just the _ n - tuples _ indicated in each term . with this , the restrictions imposed on the sums in ( [ tracos tabela1 ] ) are removed . we estimate the number of configurations that contribute to @xmath102 for a given _ n - tuple _ @xmath103 by fixing the relative position of the cargos @xmath104 and counting for all possible sequences of @xmath105 and @xmath106 . we then sum over @xmath104 observing the invariance of the trace under cyclic transformations . the results are compiled below . @xmath107 tr(e ) \\ & & & \\ ( d ) & \overline{w}_{(021)-(12 ) } & \simeq & \displaystyle\sum\limits_{m_{i}=1}^{m-2}\dbinom{n - m_{i}-5}{m - m_{i}-1}\left [ ( n - m_{i}-4)x^{m_{i}}\right ] tr(e ) \\ & & & \\ ( e ) & \overline{w}_{(121)-(02 ) } & \simeq & \displaystyle\sum\limits_{m_{i}=1}^{m-1}\dbinom{n - m_{i}-5}{m - m_{i}-1}\left [ ( n - m_{i}-4)x^{m_{i}}\right ] tr(e ) \\ & & & \\ ( f ) & \overline{w}_{(021)-(02 ) } & \simeq & \dbinom{n-5}{m-1}\left [ ( n-4)\right ] tr(e ) \\ & & & \\ ( g ) & \overline{w}_{(02)-(02 ) } & \simeq & \frac{1}{2}\dbinom{n-4}{m}tr(e)\end{array } \label{config}\ ] ] .... .... variables @xmath88 and @xmath108 indicate the number of possible consecutive motors at the left of the cargos in each of these configurations contributing to a given @xmath109 . the sums over @xmath88 and @xmath108 are estimated here in the limit of very large systems for which @xmath110 and @xmath111 keeping the motor density @xmath112 finite within the range @xmath113 in this limit , the sums converge to integrals and these integrals can be evaluated using laplace 's asymptotic method . consider , for instance , the sum @xmath23 in ( [ config ] ) . we use stirling 's formula @xmath114 to approximate the factorials involving the variables @xmath1 and @xmath0 and define the new variables marchetti @xmath115 @xmath116 and @xmath117 assume continuous values in this limit so that the referred sum converges to the integral @xmath118the function @xmath119 in the expression above depends only on the sum @xmath120 @xmath121\ln [ 1-\left ( y+z\right ) ] -[\rho -\left ( y+z\right ) ] \ln [ \rho -\left ( y+z\right ) ] . \label{h}\ ] ] thus , by defining @xmath122 ( [ w1212 - 1 ] ) can be rewritten as @xmath123where@xmath124 in order to apply laplace 's method for estimating the above integral , it is convenient to change the order of the integration observing that @xmath125with this change , the integral in @xmath117 becomes trivial and the double integral in ( [ somaw12 - 12 ( 3 ) ] ) reduces to @xmath126which can be estimated by its maximum at large @xmath1 . for this , notice that @xmath127 has a maximum at@xmath128thus , if @xmath129 the condition @xmath130 can not be satisfied and the maximum contribution to the integral in ( [ if2 ] ) comes from the extremum of the interval at @xmath131 which is a local maximum . the result is @xcite , @xmath132 \frac{1}{n^{2}}\frac{1}{\sqrt{\rho } } \frac{\exp [ -n\rho \ln \rho ] } { \left [ \ln ( x\rho ) \right ] ^{2}}. \label{if2 estimativa a e c}\]]if however , @xmath133 then @xmath134 is localized inside the integration interval so that the integral in ( [ if2 ] ) is estimated as @xcite @xmath135 \frac{(x-1)^{2}}{\left ( x\right ) ^{3/2}}\frac{1}{(1-\rho ) ^{5/2}}\frac{(1-x\rho ) } { 1-x}\sqrt{\frac{2\pi } { n}}\exp \left\ { n\left [ \ln x-(1-\rho ) \ln \tfrac{(x-1)}{(1-\rho ) } \right ] \right\ } . \label{if2 estimativa b}\ ] ] we use this same procedure to estimate all the remaining terms in expression ( [ < v(2 ) > ] ) . we merely quote the results below , making some extra comments when necessary . for estimating the sum indicated as @xmath24 in ( [ config ] ) we notice that the difference @xmath136 between the expression in the rhs of @xmath24 and the sum in @xmath23 is@xmath137applying laplace 's method to the resulting integrals after taking the thermodynamic limit , it gives @xmath138 \notag \\ & & \times \left\ { \begin{array}{l } \dfrac{1}{n^{2}}\dfrac{1}{\sqrt{\rho } } \dfrac{\exp \left ( -n\rho \ln \rho \right ) } { \left ( \ln x\rho \right ) ^{2}},\quad \qquad x\rho < 1 \\ \text{or } \\ \dfrac{-\sqrt{\rho } } { \left ( x\right ) ^{5/2}}(1-x\rho ) \dfrac{(1-x)^{2}}{(1-\rho ) ^{7/2}}\sqrt{\dfrac{2\pi } { n}}\exp \left\ { n\left [ \ln x-(1-\rho ) \ln \tfrac{(x-1)}{(1-\rho ) } \right ] \right\ } , \text { } \\ \hspace{2.5in}\text{\quad\ \ \ \ \ \ \ \ } x\rho > 1\ \hspace{0.1in}\text { } \end{array}\right.\end{aligned}\ ] ] the sum over configurations that contribute to @xmath139 in ( [ config ] ) - @xmath25 is estimated through the asymptotic behavior of a single integral , which gives@xmath140 \right\ } \sqrt{\dfrac{2\pi } { n}},\qquad \text { } x\rho > 1\ \end{array}\right . \notag\end{aligned}\ ] ] analogous procedures are used to estimate the sums over configurations of the kind @xmath141 and @xmath142 in ( [ config ] ) - @xmath143 and @xmath144which coincide in this limit : @xmath145 \right\ } \sqrt{\dfrac{2\pi } { n}},\quad x\rho > 1\end{array}\right . \notag\end{aligned}\ ] ] for the remaining sums indicated in ( [ config ] ) - @xmath146 and @xmath147 it is sufficient to estimate the relevant contributions to @xmath148 which are @xmath149 } \label{w02 - 02 comb(1)}\]]and @xmath150 } \label{somaw021 - 02 ( 0)}\ ] ] in the following , we analyze the results for the average velocity of a cargo obtained in this limit using the estimates above . the average velocity of a cargo in the system of interacting motors and cargos that obey the asep dynamics set in ( [ dinamica ] ) can now be analyzed observing the differences in the expressions obtained above for the integrals in each of the asymptotic regions limited by the range of the product @xmath151 of the two variables @xmath152 and @xmath153 such differences lead to distinct behaviors for @xmath154 characterizing different phases of the system that , in turn , reflect the differences in the distribution of motors along the considered microtubule . a a fixed value of @xmath155 such that @xmath156 and for @xmath157 @xmath154 is obtained from the behavior of the integrals for @xmath158 resulting in@xmath159 -p(1-\rho ) } { 1 + 4\left\vert \ln \rho x\right\vert + \left [ \ln \rho x\right ] ^{2 } } \label{<v(2)>(a)}\ ] ] within the complementary region in which @xmath160 @xmath154 is determined from the behavior of the integrals for @xmath161 we find @xmath162 notice that for any @xmath163 the condition ( a ) @xmath164 is always satisfied so that the results for @xmath154 are given in this case by ( [ < v(2)>(a ) ] ) within the entire interval @xmath113 thus , for small values of @xmath155 the system does not exhibits phase transitions . the behavior of @xmath154 is shown in _ fig.2 _ for the whole range of motor density @xmath152 , at fixed @xmath165 and @xmath166 and at various values of @xmath155 ( x ) . we notice in these results that for varying @xmath152 and for @xmath155 slightly above @xmath167 the average velocity of the cargo changes sign . this means that at steady state , which may be achieved at sufficiently short times after an eventual change in motor density at the microtubule , cargos may adjust and change their direction of propagation moving across motor clusters . the mechanism for cargo transfer envisaged here is equivalent to a hopping process in which the associated rates depend on site occupation . because motors move and their movement is affected by the presence of the cargos and all other motors on the microtubule , the long - time dynamics of the system must be examined globally . our results indicate that the existence of mutual interactions and the fact that many cargos are allowed to coexist at the microtubule are determinant for reproducing in this context the characteristics of the bidirectional movement . we show that within a certain range of motor density a cargo in this system executes long - range displacements in both directions . we may argue then that long - range cargo transfer is facilitated by traffic and specifically , by the assembly of motors into clusters , which characterizes traffic jam . the presence of the other cargos in the system is essential for this to occur as they function as additional obstacles that interfere in the motor density profile . each cargo induces aggregation of motors at its back end . in turn , this provides the conditions for cargo to execute long - range displacements either backwards , over the aggregate assembled at its back end , as well as forwards , over the aggregate assembled at the back end of the cargo in front . as it was originally formulated the model does not account for the possibility that a motor with one or more attached cargos may move as well . in fact , this is the only mechanism that is usually employed to describe cargo transport and it is the basis for _ coordination _ or _ tug - of - war _ models . also , for simplicity we have considered interactions of a cargo with a single motor at a time . should a set of motors be allowed to interact with the cargo to participate in the transfer process then the map into the asep would need to be modified accordingly . we are currently working on these possibilities by including into the dynamics ( [ dinamica ] ) a process of the kind @xmath168 that recovers ergodicity of the model @xcite we should emphasize that the occurrence of the long - range bidirectional movement as a consequence of the hopping processes devised here may happen by the action of motors of just one kind possessing a well defined polarity . changes in motor density and related traffic profile suffice as a mechanism to control cargo direction and the size of the runs determined essentially by the extent of motor clusters at jamming conditions . this offers a rather straightforward explanation for the data mentioned above suggesting that the number of long run - lengths performed by the observed beads increases significantly as the density of motors at the microtubule increases beeg . in addition , the results presented here indicate that , for sufficiently high values of the motor density for which @xmath169 at the point where the model displays a phase transition , cargos would perform a uniform movement ( on average ) since their velocities become independent of @xmath170 such behavior has also been observed in the same set of experiments . as noticed by ma and chisholm @xcite , `` little is known regarding motor traffic and how it correlates with the movement of cargo '' . here , we offer a possibility based on the idea that the transport does not require the action of an external agent to coordinate the process , or a _ tug - of - war _ mechanism or even the existence of a mechanical coupling between two kinds of motors as proposed more recently @xcite . instead , it suggests that such coordination can be achieved by collective effects on the course of the dynamics as the system `` self - organizes '' so that it presents characteristics that reflect an internal ( and global ) order that does not have its origin in the characteristics of the external medium . it is then possible that the necessary transport in cells is accomplished just by adjusting the density of motors at the microtubule . accordingly , the presence of processive motors of different polarities that are normally required to explain the movement of a putative motor - cargo complex would not be necessary . transportation here is based on a mechanism that requires formation of clusters of motors , not necessarily on their ability to travel along long distances . i would like to thank domingos h.u . marchetti for very helpful and enthusiastic discussions regarding the conditions imposed by fubini 's theorem and the procedure used here to estimate the double integrals ; and also elisa t. sena for pointing to me many of the difficulties in the initial developments of this work . i thank scott hines for kindly editing a preliminary version of the text . this work had integral support from fundao de amparo a pesquisa do estado de so paulo ( fapesp ) - brazil . i. m. kulic , a. e. x. brown , h. kim , c. kural , b. blehm , p. r. selvin , p. c. nelson , v. i. gelfand , the role of microtubule movement in bidirectional organelle transport , proc . usa 105 , 10011 - 10016 ( 2008 ) . b. derrida , m.r . evans , the asymmetric exclusion model : exact results through a matrix approach , nonequilibrium statistical mechanics in one dimension , university press , uk , chapt . 14 , 277 - 304 ( 1997 ) ; b. derrida , m.r . evans , v. hakim , v. pasquier , an exact solution of a 1d asymmetric exclusion model using a matrix formulation , j. phys.a26 , 1493 - 1517 ( 1993 ) . * figure 1 - * dynamics of motors and cargos . ( a ) * cargo transfer . * it happens here through a mechanism of hopping between neighbor motors . due to the flexibility of the tail , the attached cargo may display small oscillations leading to the possibility of it being caught either by the motor at its left or by the motor at its right . the corresponding processes 12 @xmath171 21 or 21 @xmath171 12 are represented in the figure . ( b ) * the step of a motor*. the time spent by the motor with the two heads attached to the microtubule is much larger than the time it spends with just one of the heads attached @xcite , as a part of the `` hand - over - hand '' mechanism proposed to explain the kinetics of two - headed motor proteins @xcite . occupation of a site by a motor occurs here whenever it is occupied by the two heads of a motor . the motor step is then represented as @xmath172 which is indicated in the figure .	most models designed to study the bidirectional movement of cargos as they are driven by molecular motors rely on the idea that motors of different polarities can be coordinated by external agents if arranged into a motor - cargo complex to perform the necessary work @xcite . although these models have provided us with important insights into these phenomena , there are still many unanswered questions regarding the mechanisms through which the movement of the complex takes place on crowded microtubules . for example ( i ) how does cargo - binding affect motor motility ? and in connection with that - ( ii ) how does the presence of other motors ( and also other cargos ) on the microtubule affect the motility of the motor - cargo complex ? we discuss these questions from a different perspective . the movement of a cargo is conceived here as a _ hopping process _ resulting from the transference of cargo between neighboring motors . in the light of this , we examine the conditions under which cargo might display bidirectional movement even if directed by motors of a single polarity . the global properties of the model in the long - time regime are obtained by mapping the dynamics of the collection of interacting motors and cargos into an asymmetric simple exclusion process ( asep ) which can be resolved using the matrix _ ansatz _ introduced by derrida @xcite . _ keywords - _ intracellular transport by molecular motors ; bidirectional movement of cargo , traffic jam on microtubules ; asep models .
classical supersymmetry sets out to extend the unified treatment of bosonic and fermionic quantities in the usual qft framework to the classical level . normally , in semiclassical treatments the fermionic variables are set to zero as soon as the supersymmetric theory has been constructed . the usual argument goes that since we can not find classical fermions in nature , fermionic quantities should be omitted altogether at the classical level . however , this is far from necessary . in fact , a consistent approach to classical supersymmetry has long been available for a review see e.g. the book by de witt @xcite . fermionic quantities are then treated as anticommuting variables taking values in a grassmann algebra @xmath1 . grassmann - valued mechanics has been analysed in the works of berezin and marinov @xcite and casalbuoni @xcite and later by junker and matthiesen @xcite . a main difference to our work is that both @xcite and @xcite do not distinguish clearly between generators of the algebra and dynamical quantities and thus define the grassmann algebra @xmath1 rather implicitly . the fact that the bosonic variables take values in the even part of the same algebra @xmath1 is not apparent in these works , although both recognize that the bosonic variables can not be real functions anymore without , however , elaborating on this fact . a central aim of this paper is therefore to make sense of the _ general _ grassmann - valued equations of motion , including the fermionic ones , and to find ways to their solution , which is done in @xcite and @xcite only in very special cases . junker and matthiesen , who investigate a similar mechanical model , achieve a more general solution than in @xcite and @xcite , but again under the ( implicit ) assumption that the grassmann algebra is spanned by only two generators identified with the fermionic dynamical variables . we can confirm most of their results ( in different form , though , due to a different choice of variables ) as special cases of our solutions . however , we disagree about some details , in particular , concerning the case of zero energy . the mechanical model that we study here is the supersymmetric motion of a particle in a one - dimensional potential , derived by dimensional reduction from the usual @xmath2 supersymmetric @xmath3-dimensional field theory with yukawa interaction . a slightly different version of this model was investigated in @xcite , where a different concept of reality was used that led to a negative potential in the lagrangian . the approach taken here stays closer to the usual case with the positive potential . an important result of @xcite was that a complete solution for the particle motion could be found on the assumption that the underlying algebra @xmath1 has only two generators . this led to relatively simple results , however is unnecessarily restrictive . here we show first that for a large class of potentials the solution to the equations of motion can be found for any @xmath1 and depends only on a small number of @xmath1-valued constants of integration , one of which is a grassmann energy @xmath4 . to deal with essentially arbitrary potentials we adopt a second method which is closer to that of @xcite , although we need not restrict ourselves to two generators : choosing the grassmann algebra to be finitely generated , with @xmath5 generators , we split all dynamical quantities and equations into their real components , named according to the number of generators involved in the corresponding monomial . then , beginning from the zeroth order equation , which can be seen as a form of newton s equation , we subsequently work our way up to higher and higher orders , utilizing the solutions already found for the lower levels . this layer - by - layer strategy allows us to solve the equations of motion for any potential with reasonable mathematical properties . the existence of a complete solution to the coupled system of equations of motion looks surprising in view of the increasingly large number of equations involved for large @xmath5 . however , on second thoughts it is not so unexpected : due to our first solution method we know that a full grassmann solution can be found in many cases , the decomposition of which should give us exactly the component solutions obtained by the second method which it does indeed as we shall demonstrate . a final word has to be said about the assumption of only a finite number of generators since it has been claimed that this must necessarily lead to contradictions : emphasizing that our paper deals with the _ classical _ theory we do not find this to be true . we begin our analysis in section [ susymechanics ] by presenting the lagrangian and the equations of motion that we will be concerned with in this paper . essential for solving these equations are the symmetries and associated noether charges of the lagrangian which we therefore examine in section [ grassmannsymmetries ] . for a certain class of potential functions , namely those for which a particular integral can be calculated analytically , we describe in section [ solfixpot ] how the equations of motion can be solved completely and illustrate this method for two exemplary potentials , the harmonic potential @xmath6 and the hyperbolic potential @xmath7 . sections [ finite algebra ] and [ solarbpot ] are devoted to the description of our layer - by - layer method which is then explicitly carried out up to fourth order , and illustrated by the harmonic oscillator case in section [ harmpot ] . next , we investigate the symmetries in component form in section [ symmetries ] : while all component charges can be simply derived by decomposing the original charges , they also reflect a huge number of symmetries of the highest order component lagrangian by which they can be found using noether s procedure . in addition to the symmetries known from the original lagrangian there appear extra classes of symmetries which can not be related to the former . it is possible to trace them back to invariance under certain changes of base in the grassmann algebra , viewed as a @xmath8-dimensional vector space . one question that immediately arises from the decomposition of the physical quantities into components is addressed in section [ interpret ] : how must we interpret the numerous new functions that arise from this procedure ? utilizing the new symmetries found in section [ symmetries ] we offer a geometric interpretation for the lowest order bosonic and fermionic functions and describe the three basic types of motion possible for them . the higher order quantities are interpreted here from a different point of view : we see them as variations of the lower order quantities with respect to the integration constants involved in these functions , namely an initial time @xmath9 and the in section [ specialtypesofmotion ] we apply our results to study general oscillatory motion , by which we mean that the lowest order bosonic function is periodic . the characteristic appearance of linear - periodic terms is explained from both a physical and mathematical perspective employing results of floquet theory . again , we choose a particular potential to illustrate this in section [ hyperbolpot ] , which allows us also to show that the two solution methods presented in this paper coincide . a common restriction on both methods is that the energy @xmath10 of the system must be positive , therefore the zero energy case has to be discussed separately . we do this in section [ zeroenergy ] . some ideas for further generalization and analysis conclude this paper . we start our discussion with the standard lagrangian density for @xmath3-dimensional supersymmetric field theory with yukawa interaction : @xmath11 containing a real bosonic scalar field @xmath12 , a real two - component fermionic spinor field @xmath13 and a potential function @xmath14 ; @xmath15 are the light cone derivatives @xmath16 . we assume that @xmath12 and the two components of @xmath13 take their values in the real even and odd part , respectively , of an arbitrary grassmann algebra @xmath1 , and that the potential function @xmath14 can be expanded into a power series in @xmath12 with real coefficients . complex conjugation is defined such that @xmath17 in accordance with the conventions in @xcite so that the lagrangian density is a real function despite the @xmath18-factors that occur in front of the fermion terms . as we are dealing with the classical case we assume further that all fields and their derivatives commute or anticommute , depending in the usual way on the bosonic or fermionic nature of the fields . in this paper we shall be interested only in spatially independent fields , i.e. @xmath19 . as the fields are then functions of time only this leads us directly from field theory to mechanics . it is therefore sensible to think of the bosonic field @xmath12 as describing the one - dimensional motion of a particle in the potential @xmath20 . to support this notion we will change the variable @xmath12 to @xmath21 for the rest of this paper . we then obtain the following lagrange function : @xmath22 where the dot denotes a time derivative and @xmath23 means the derivative of @xmath24 with respect to @xmath21 . performing a formal variation of this lagrangian with respect to the variables @xmath21 , @xmath25 and @xmath26 and neglecting total time derivatives we derive the equations of motion for the system : @xmath27 there is a slight ambiguity in these equations if the grassmann algebra is finitely generated and has an odd number of generators . then the two equations for the fermion variables will be determined only up to an arbitrary function of highest order in the grassmann algebra . we can think of this as a gauge degree of freedom and will come back to this point later . in order to solve the equations it is advisable to understand the symmetries of the system first . this we will do in the next section . the first thing to notice is that the lagrangian has no explicit time dependence . we therefore have invariance under time translation , leading to a conserved hamiltonian as the corresponding noether charge . this we calculate to be @xmath28 its conserved value we call the energy , denoted by @xmath4 . note , however , that this hamiltonian is an even grassmann - valued function and that therefore the conserved energy @xmath4 is an even element of @xmath1 . the operator @xmath29 generates time translations , and acts on the dynamical variables in the obvious way : @xmath30 where @xmath31 is an infinitesimal even grassmann parameter . in addition to time translation invariance there are two further independent symmetries of the lagrangian , relating fermions and bosons . these can be written in infinitesimal form as : @xmath32 where @xmath33 is an arbitrary infinitesimal odd grassmann parameter . these transformations lead only to a change in the lagrangian by a total time derivative . we can therefore apply noether s procedure and find the following charges : @xmath34 the conservation of which can be easily shown using the equations of motion . note that the charges are odd elements of the grassmann algebra . as for time translation invariance we now define two operators @xmath35and @xmath36 generating the two symmetry transformations . from ( [ susy transformation ] ) we can read off the action of these operators on the dynamical variables : @xmath37 using the action of the operators on @xmath21 , @xmath25 and @xmath26 we find that @xmath35 , @xmath36 and @xmath29 form a closed algebra with the relations @xmath38 as long as the equations of motion are satisfied . notice that @xmath29 commutes with everything . this is the usual @xmath2 supersymmetry algebra in @xmath39 space dimensions , and with ( [ susy representation ] ) we have an _ _ representation of this algebra . thus we will from now on speak of supersymmetry transformations and call the associated charges supercharges . beside time - translation invariance and the two supersymmetries there exists still a further invariance , this time for the fermionic functions only . in infinitesimal form it is given by the transformation @xmath40 where @xmath31 again denotes an infinitesimal even grassmann parameter . we can think of this as an internal rotation of the fermionic variables . this invariance leads to a further conserved noether charge : @xmath41 comparison with ( [ lagrangian ] ) shows that the fermionic functions enter the interaction term in the lagrangian only via @xmath42 . this leads directly to the fact that the @xmath21-motion depends on the fermionic functions only through this one constant . in addition , the hamiltonian ( [ hamiltonian ] ) reveals that the only fermion contribution to the total energy @xmath4 is through @xmath42 . this is an important simplification , and it will allow us to solve the equations of motion completely for a number of potential functions @xmath24 without restriction on the nature of the grassmann algebra @xmath1 . we have already mentioned that the @xmath21-equation of motion ( [ x - equation ] ) nearly decouples from the other two equations because the fermion functions in the coupling term @xmath43 form a grassmann constant . it is therefore sensible to begin with this equation . from the conserved hamiltonian we know that @xmath44 for the next step we have to use that every grassmann number @xmath45 can be split into two parts , its body and its soul : @xmath46 . the body is just the real number content of @xmath45 , the soul is the remaining linear combination of products of ( odd ) grassmann generators and will be nilpotent if the grassmann algebra is finitely generated . note , however , that we do not need such a restriction for body and soul to be well - defined . a square root for a grassmann quantity can be defined by its power series as long as the body is positive . since @xmath42 is the product of two odd grassmann terms , its body and therefore the body of the whole third term on the right hand side of ( [ energy ] ) is zero , leaving the restriction @xmath47 if @xmath48 is to be well - defined . this just means that the kinetic energy of the classical particle moving in the potential @xmath49 has to be positive . the resulting first order differential equation for @xmath21 is @xmath50 provided @xmath1 is finitely generated , we may regard @xmath21 as lying in the vector space @xmath1 and apply the standard theory of systems of ode s to show that this equation has a unique solution for any given initial data @xmath51 . the equation can formally be solved by separating variables : @xmath52 note that while the left - hand side is just a real expression , the right - hand side is a grassmann integral of a grassmann integrand . such an integral is defined as a line integral in the grassmann algebra , thought of as a finite or infinite - dimensional vector space spanned by products of generators . as is shown in @xcite such an integral is independent of the actual path relating start- and endpoint of the integral since the integrand is grassmann - even . now the integrand comes from an ordinary real function extended into the full grassmann algebra using its power series . we can therefore sensibly ask whether there is an indefinite integral @xmath53 to this _ real _ function . if so , we can extend _ this _ function back into the grassmann algebra thus gaining an indefinite integral for the grassmann integrand . because the integrand is even it then follows @xcite that the integral is given by the difference of @xmath54 evaluated at the start- and endpoints of the path : @xmath55 if the function @xmath54 has an inverse , we finally get @xmath21 as a function of @xmath56 and therefore the solution of ( [ x - equation ] ) . we will illustrate this method below for two potential functions @xmath24 . the solution to the fermion equations ( [ fermion - equation 1 ] ) and ( [ fermion - equation 2 ] ) is now easy . from the solution for @xmath57 we can immediately calculate @xmath58 and @xmath59 . using the explicit formulae ( [ q ] ) and ( [ qt ] ) for the two conserved supercharges @xmath60 and @xmath61 we can evaluate the linear combinations @xmath62 where we have used that @xmath63 which follows from ( [ r ] ) . since the hamiltonian @xmath64 has the constant value @xmath4 we deduce that @xmath65 this effectively solves ( [ fermion - equation 1 ] ) and ( [ fermion - equation 2 ] ) , as the values of @xmath60 and @xmath61 are determined by the data at the initial time @xmath9 . there is one subtle point here , however , since the division by @xmath66 is only possible as long as @xmath67 , the body of @xmath4 , which can be interpreted as the classical energy of the particle , is non - zero . indeed , when we analyse the solution for finitely generated grassmann algebras later , we will find again that the @xmath68-case is special . a second point which has to be mentioned here is that we have treated @xmath42 as an independent parameter . however , @xmath42 is _ determined _ by the two fermionic quantities @xmath25 and @xmath26 and , in effect , by the two supercharges @xmath60 and @xmath61 : @xmath69 which can be verified using the definition of @xmath42 in ( [ r ] ) . this does not render our solution invalid but it shows that @xmath42 is not a parameter that can be chosen independently . but let us now turn to some examples : we start with one of the easiest problems , the harmonic oscillator with potential real . here the integrand in ( [ grassmann integral ] ) is given by : @xmath70 where @xmath4 is the constant grassmann energy . note that due to the special nature of the potential we can combine the @xmath71-term @xmath72 with the energy @xmath4 into one overall constant . if we treat the integrand as an ordinary real function it has an indefinite integral : @xmath73 so that we can write @xmath74^{x(t)}_{x(t_{0})}.\ ] ] this formula can be inverted to yield the solution @xmath75 where we have denoted the constant @xmath76 by @xmath77 . this looks formally like the usual harmonic oscillator solution , and it is periodic with period @xmath78 , but all terms are grassmann - valued and not just real functions . for the two fermion terms we need only calculate @xmath48 and @xmath79 and insert them into equations ( [ fermion1 ] ) and ( [ fermion2 ] ) , which leaves us with : @xmath80 where @xmath81 with ( [ hx - solution ] ) , ( [ hpsi1-solution ] ) and ( [ hpsi2-solution ] ) we have found the complete solution to the equations of motion for @xmath82 , independent of the nature of the grassmann algebra @xmath1 . as a second example we choose the hyperbolic potential @xmath84 , with @xmath85 and @xmath86 both real . the integrand in ( [ grassmann integral ] ) is @xmath87 and viewed as a real function has an indefinite integral , the form of which depends on the constants involved : @xmath88 because the grassmann - valued function is completely determined by the corresponding real function we have to look at the body of @xmath89 to decide which integral to use . since @xmath42 as product of two odd grassmann constants has no body , the crucial term is @xmath90 . all three integrals ( [ integrals ] ) can be inverted to give @xmath21 as a function of @xmath56 : @xmath91 where @xmath92 from equations ( [ fermion1 ] ) and ( [ fermion2 ] ) we can calculate the fermion solutions : @xmath93 with @xmath94 , etc . again , we have found the complete solution for the system without using any information about the grassmann algebra @xmath1 itself . we will return to this example later , after we have investigated a special class of grassmann algebras . we will now make a choice for the underlying grassmann algebra @xmath1 of the system , namely that it is generated by a finite number of elements with the property @xmath95 every element @xmath45 of @xmath1 can be written in the form @xmath96 with complex coefficients @xmath97 . @xmath98 is the body of @xmath45 . the requirement for @xmath45 to be real ( in the sense explained earlier ) fixes each coefficient to be either purely real or purely imaginary , depending on the number of generators involved . for practical reasons we will usually restrict ourselves to the case of four generators , although the general aspects of our discussion should be true for an arbitrary number @xmath5 . in the case of four generators it is useful to define the following real monomials : @xmath99 note that due to antisymmetry there are only six linearly independent monomials involving two generators , four monomials involving three generators , and just one monomial of highest order , i.e. involving all four generators . we can decompose the dynamical quantities as follows : @xmath100 where it is implied from here on that the summation over indices is with @xmath101 . in total there are 24 independent real functions which we call of zeroth , first , second , third or fourth order according to the number of generators involved in the corresponding monomials . every analytic function involving @xmath21 , @xmath25 or @xmath26 can be decomposed as well using its taylor expansion . applied to the potential function @xmath24 this yields : @xmath102}\right)\xi_{1234},\ ] ] where the brackets denote antisymmetrization . for example @xmath103}= x_{12}x_{34}-x_{13}x_{24}+x_{14}x_{23}+x_{23}x_{14}-x_{24}x_{13}+x_{34}x_{12}.\ ] ] note that @xmath104 and its derivatives are ordinary real functions . inserting these results into ( [ lagrangian ] ) yields eight component lagrangians of order zero , two and four , respectively : @xmath105}+\dot{{\ensuremath{x_{0}}}}\dot{x}_{1234}- \frac{1}{2}\left(uu'\right)'x_{[12}x_{34]}-uu'x_{1234}\label{l1234}\\ & + & \lambda_{[123}\dot{\lambda}_{4]}\!+\!\rho_{[123}\dot{\rho}_{4]}+\!u ' \left(\lambda_{[123}\rho_{4]}\!-\!\rho_{[123}\lambda_{4]}\right)+u''x_{[12 } \left(\lambda_{3}\rho_{4}\!-\!\lambda_{4}\rho_{3}\right ) _ { ] } \nonumber\end{aligned}\ ] ] where the argument of @xmath24 , @xmath23 and @xmath106 is always @xmath107 , here and below . from the highest order lagrangian @xmath108 , which is a functional of 24 generally different component functions , we get the following set of euler - lagrange equations : @xmath109 } - ( uu')'x_{1234}\\ + u''\left(\lambda_{[123}\rho_{4]}-\rho_{[123}\lambda_{4]}\right ) + u'''x_{[12}\left(\lambda_{3}\rho_{4}-\lambda_{4}\rho_{3}\right ) _ { ] } \end{array}\label{x1234 } \\ \left . \begin{array}{l } \dot{\lambda}_{i}\\ \dot{\rho}_{i } \end{array } \right\}&= & \left\ { \begin{array}{l } - u ' { \ensuremath{\rho_{i}}}\\ \;\;\,u ' { \ensuremath{\lambda_{i}}}\end{array } \right.\label{liri}\\ \left . \begin{array}{l } \dot{\lambda}_{ijk}\\ \dot{\rho}_{ijk } \end{array } \right\ } & = & \left\ { \begin{array}{l } - u ' { \ensuremath{\rho_{ijk}}}- u''x_{[ij}\rho_{k]}\\ \;\;\ , u ' { \ensuremath{\lambda_{ijk}}}+ u''x_{[ij}\lambda_{k ] } \end{array } \right.\label{lijkrijk}\end{aligned}\ ] ] the same equations can also be obtained by splitting the original equations of motion ( [ x - equation])([fermion - equation 2 ] ) into their components . it is remarkable that all equations can be derived from just the one lagrangian . the other seven component lagrangians are completely redundant and yield no new equations : we can only obtain equation ( [ x0 ] ) whereas from the lagrangians of type ( [ lij ] ) we can derive equations ( [ x0 ] ) , ( [ xij ] ) and ( [ liri ] ) . now we can also see where the ambiguity mentioned earlier comes from when there is an odd number @xmath5 of generators : because the lagrangian is an even functional , none of its components can contain any function of highest order @xmath5 , so these functions can not be governed by any equations of motion . e.g. for three generators the functions @xmath110 and @xmath111 will stay completely undetermined . formally , we can derive evolution equations similar to ( [ lijkrijk ] ) by decomposing ( [ fermion - equation 1 ] ) and ( [ fermion - equation 2 ] ) into components but they are meaningless since an arbitrary function can be added to both sides . thus we can regard these highest order functions as non - physical and treat them as trivially separated gauge degrees of freedom . however , if there is an even number of generators this problem can not occur , since there will always be one lagrangian of highest order containing component functions of all orders which are thus determined by equations of motion . we now proceed to derive the solutions to the equations of motion ( [ x0])([lijkrijk ] ) for an arbitrary , sufficiently differentiable potential function @xmath24 adopting a layer - by - layer strategy . that means that we will start with the lowest order or body equation ( [ x0 ] ) and then use its solution to work our way up to the higher order and more complex equations , including also the fermionic ones . we do this here up to fourth order , although there is no obstruction in principle to continue the strategy for an algebra with more than four generators . in fact , in every layer the equations and their solutions are the same irrespective of how many generators there are ; there will only be a different number of them . before we start with the bosonic equations , it is useful to look at the decomposition of the hamiltonian @xmath112 and of the noether charge @xmath113 ( note that @xmath42 does not have a body ) : @xmath114}+\dot{x}_{0}\dot{x}_{1234}+ \frac{1}{2}\left(uu'\right)'x_{[12}x_{34]}+uu'x_{1234}\\ & & -u'{\ensuremath{r_{1234}}}- u''x_{[12}r_{34]}\nonumber\\ { \ensuremath{r_{ij}}}&=&{\ensuremath{\lambda_{i}}}\rho_{j}-{\ensuremath{\rho_{i}}}\lambda_{j}\label{rij}\\ { \ensuremath{r_{1234}}}&=&\lambda_{[123}\rho_{4]}-\rho_{[123}\lambda_{4]}.\label{r1234}\end{aligned}\ ] ] since these are components of grassmann conserved quantities , they are conserved too , as can be checked easily using the equations of motion . all components can be derived as noether charges from the highest order lagrangian @xmath108 and we will do this in section [ symmetries ] but for now we only need that they are constant in time . the respective values of the hamiltonian functions @xmath115 , @xmath116 and @xmath117 are denoted in the following by @xmath10 , @xmath118 and @xmath119 . we notice first that ( [ x0 ] ) is just the standard newtonian equation of motion for a particle moving in a potential @xmath120 . this is the bottom layer of the system . the solution to ( [ x0 ] ) using the constancy of the hamiltonian @xmath115 is well - known : @xmath121 the sign has to be chosen carefully to comply with the direction of motion of the particle ; if the particle motion changes direction , the integral will be multi - valued and has to be glued together from pieces with unique sign to ensure that the overall result for @xmath122 is monotonically growing . the implicit function theorem then allows us to locally invert @xmath122 and obtain the required @xmath107 . since ( [ x0 ] ) is a second order equation there are two constants of integration , @xmath123 ( or equivalently @xmath124 ) and the energy @xmath10 . whereas ( [ x0 ] ) is a non - linear equation , ( [ xij ] ) is an inhomogeneous _ linear _ equation ( which can be simplified using ( [ rij ] ) ) . it is now convenient to use the hamiltonian ( [ hij ] ) : solving the equation @xmath125 for @xmath126 we can reduce the problem to a first order differential equation . this can be solved by standard methods to yield : @xmath127 where @xmath128 is an arbitrary integration constant related to the initial value of @xmath129 . the second integration constant of ( [ xij ] ) is the energy variable @xmath118 . this result was also found in @xcite with @xmath130 set to @xmath131 . however , using equation ( [ x0 ] ) and @xmath132 , the last term on the right hand side can be rewritten : @xmath133 so that we end up with the simpler expression @xmath134 all three functions which occur in ( [ solutxij ] ) can be calculated directly from @xmath107 . the same procedure can be applied to solve ( [ x1234 ] ) which simplifies when we use ( [ rij ] ) . we rearrange the equation @xmath135 to isolate @xmath136 and then solve this first order equation . the result can be written as : @xmath137}-\dot{x}_{[12}\dot{x}_{34]}}{\dot{x}_{0}^2}dt ' + \frac{1}{2}\dot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{u''x_{[12}r_{34]}}{\dot{x}_{0}^2}dt ' \nonumber.\end{aligned}\ ] ] again , there are two additional integration constants , @xmath138 and @xmath119 . note that the solution depends only on functions that we already know from the lower layers , namely @xmath139 and @xmath129 . in fact , the solution can be expressed as a function of @xmath139 and its derivatives only by inserting ( [ solutxij ] ) into ( [ solutx1234 ] ) : @xmath140}\ddot{x}_{0}+c_{[12}e_{34]}\left ( \ddot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1 } { \dot{x}_{0}^2}dt'+\frac{1}{\dot{x}_{0}}\right)\nonumber\\ & & + \frac{1}{2}e_{[12}e_{34]}\left[\ddot{x}_{0}\left(\int_{{\ensuremath{t_{0}}}}^t\frac{1 } { \dot{x}_{0}^2}dt'\right)^2+\frac{2}{\dot{x}_{0}}\int_{{\ensuremath{t_{0}}}}^t\frac{1 } { \dot{x}_{0}^2}dt'-3\dot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1}{\dot{x}_{0}^4}dt ' \right]\nonumber\\ & & + \frac{r_{[12}c_{34]}}{2e_{0}}u'\dot{x}_{0}+\frac{r_{[12}e_{34]}}{2e_{0 } } \left(u'\dot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1}{\dot{x}_{0}^2}dt ' -\frac{u}{e_{0}}\right ) \nonumber.\end{aligned}\ ] ] this complicated - looking expansion is particularly useful when we come to the interpretation of the solution in section [ interpret ] . there remain the fermion equations ( [ liri ] ) and ( [ lijkrijk ] ) to be solved . both are systems of two homogeneous first order equations and therefore have two independent solutions . just as we used the decomposition of @xmath64 to solve the bosonic equations , so it is now appropriate to decompose the conserved supercharges @xmath141 and @xmath142 to solve the fermionic equations . we find : @xmath143 } + \rho_{[i}u'x_{jk]},\\ \tilde{q}_{ijk}&=&{\ensuremath{\rho_{ijk}}}\dot{x}_{0}-{\ensuremath{\lambda_{ijk}}}u+\rho_{[i}\dot{x}_{jk ] } -\lambda_{[i}u'x_{jk]}. \end{array}\ ] ] using suitable linear combinations of @xmath144 and @xmath24 with coefficients @xmath145 and @xmath146 and applying @xmath132 we find that @xmath147 where @xmath148 and @xmath149 ; hence the two integration constants of the solution are basically given by the first order supercharges . a similar result was obtained in @xcite , although the restrictions placed there on the shape of the potential @xmath24 are unnecessary . note , however , that this solution does not work if the particle energy @xmath10 equals zero . this has to do with the fact that both @xmath144 and @xmath24 must vanish in this case . we will return to this in section [ zeroenergy ] . the same procedure can be used to find the solution to equation ( [ lijkrijk ] ) and we get the result : @xmath150}}{2e_{0 } } + \frac{u\dot{x}_{[ij}\rho_{k]}-u'\dot{x}_{0}x_{[ij}\rho_{k]}}{2e_{0 } } \label{solutlijkrijk1}\\ { \ensuremath{\rho_{ijk}}}&=&r_{ijk}\dot{x}_{0}+l_{ijk}u-\frac{e_{[ij}\rho_{k]}}{2e_{0 } } -\frac{u\dot{x}_{[ij}\lambda_{k]}-u'\dot{x}_{0}x_{[ij}\lambda_{k]}}{2e_{0 } } , \label{solutlijkrijk2}\end{aligned}\ ] ] where we have used the equations @xmath132 and @xmath125 and the identities @xmath151}=$ ] . the new integration constants @xmath152 and @xmath153 are related to the third order supercharges @xmath154 and @xmath155 in the same way as @xmath156 and @xmath157 are to @xmath145 and @xmath146 . as was the case for the second and fourth order bosonic solutions we find that @xmath158 and @xmath159 depend only on functions we know already from the lower order layers , namely , @xmath139 , @xmath160 , @xmath161 and @xmath129 , and again we can simplify the solution even further , inserting ( [ solutxij ] ) and ( [ solutliri ] ) into equations ( [ solutlijkrijk1 ] ) and ( [ solutlijkrijk2 ] ) : @xmath162}\ddot{x}_{0}+e_{[ij}l_{k ] } \left(\ddot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1}{\dot{x}_{0}^2}dt'+\frac{1}{\dot{x}_{0}}- \frac{\dot{x}_{0}}{e_{0}}\right)\label{3rdvar1}\\ & & -r_{ijk}u - c_{[ij}r_{k]}u'\dot{x}_{0}-e_{[ij}r_{k ] } \left(u'\dot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1}{\dot{x}_{0}^2}dt'- \frac{u}{e_{0}}\right)\nonumber\\ { \ensuremath{\rho_{ijk}}}&=&r_{ijk}\dot{x}_{0}+c_{[ij}r_{k]}\ddot{x}_{0}+e_{[ij}r_{k ] } \left(\ddot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1}{\dot{x}_{0}^2}dt'+\frac{1}{\dot{x}_{0}}- \frac{\dot{x}_{0}}{e_{0}}\right)\label{3rdvar2}\\ & & + l_{ijk}u+c_{[ij}l_{k]}u'\dot{x}_{0}+e_{[ij}l_{k ] } \left(u'\dot{x}_{0}\int_{{\ensuremath{t_{0}}}}^t\frac{1}{\dot{x}_{0}^2}dt'- \frac{u}{e_{0}}\right)\nonumber.\end{aligned}\ ] ] we have now derived all solutions to equations ( [ x0])([lijkrijk ] ) , i.e. we have explicitly solved the equations of motion up to fourth order for an arbitrary potential function @xmath24 . they are all functions of only four quantities and their time derivatives : @xmath163 to illustrate the method we choose the harmonic oscillator as our first example . this has the advantage that we can compare the solution with our earlier result ( [ hx - solution])([hpsi2-solution ] ) . the solution for @xmath139 could be found using ( [ solutxn ] ) but here it is easy to solve equation ( [ x0 ] ) directly , with the familiar result : @xmath164 the functions @xmath144 and @xmath24 can be calculated easily and we find : @xmath165 where @xmath166 and @xmath167 are the constants @xmath168 next we have to calculate @xmath129 and thus we need @xmath169 on using ( [ solutxij ] ) we find that @xmath170 for @xmath158 and @xmath159 we need not calculate any new terms except @xmath171 . after collecting various constants into @xmath172 and @xmath173 we end up with @xmath174 finally we have to calculate @xmath175 . therefore we need the integral @xmath176_{{\ensuremath{x_{0}}}({\ensuremath{t_{0}}})}^{{\ensuremath{x_{0}}}(t)}.\ ] ] there seems to be a singularity when @xmath144 approaches zero but fortunately this term has to be multiplied by @xmath144 and combined with two other terms which share the same coefficient @xmath177}$ ] . one finds that all terms involving @xmath144 in the denominator cancel each other , so that the overall result is just a linear combination of @xmath139 and @xmath144 , i.e. of @xmath178- and @xmath179-terms . this holds also for all other contributions to @xmath175 , so that we can write : @xmath180 where @xmath181 and @xmath182 are rather complicated functions of @xmath119 , @xmath138 and all lower order integration constants . the most interesting observation is that all bosonic _ and _ fermionic functions are linear combinations of @xmath178- and @xmath179-terms only . as we will see later , this similarity between the solutions of different orders is the exception rather than the rule and a special feature of the harmonic oscillator . we also find that the fermionic functions oscillate with the same period as the bosonic functions and differ only in amplitude and phase . once the motion of the lowest order bosonic function @xmath139 is fixed , the principal motion for @xmath160 , @xmath161 is almost completely determined . this is a general feature of our system , and it can be fully understood by investigating the symmetries of our model . finally , it is easy to verify that the complete component solution is compatible with the general solution found in section [ harmonicpotential ] . we have already dealt with the decomposition of the hamiltonian @xmath64 , the constant @xmath42 and the two supercharges @xmath60 and @xmath61 . now we want to explain the origin of the component charges from the underlying component symmetries . the lagrangian @xmath108 ( [ l1234 ] ) is all we need to look at , since the non - trivial symmetries of all lower order lagrangians form subgroups of the full symmetry group of @xmath108 . one would not guess the extraordinary number of symmetries of @xmath108 if one did not know their supersymmetric origin . to start with , ( [ l1234 ] ) is invariant under time translations , and this leads to the conservation of the highest order hamiltonian @xmath117 . this result can be generalized by looking at the full grassmann transformation ( [ timetranslation ] ) . decomposion into components gives us back the time translation symmetry mentioned above when @xmath183 is real ; the choices @xmath184 ( no summation ) or @xmath185 lead to seven extra charges which correspond to the hamiltonians @xmath116 and @xmath115 , respectively . apart from the hamiltonians , sixteen noether charges @xmath186 and @xmath155 derive from the supercharges @xmath60 and @xmath61 and belong to a set of symmetry transformations which can be obtained from ( [ susy transformation ] ) by the choices @xmath187 and @xmath188 ( no summation both times ) , respectively . as an example , @xmath145 and @xmath146 are the charges that belong to : @xmath189 where @xmath190 is a cyclic permutation of @xmath191 and no summation over @xmath18 is implied . the term supersymmetry transformations should be avoided here , though , since the dynamical quantities in component form are just real functions therefore they can be termed bosonic or fermionic only by convention . notwithstanding this fact the component transformations and the corresponding charges _ reflect _ the underlying supersymmetry . further component noether charges derived from the lagrangian ( [ l1234 ] ) are @xmath130 and @xmath192 . the transformations can be read off from ( [ fermion rotation ] ) choosing @xmath193 ( no summation ) and , respectively . in the latter case we get , for example @xmath194 so far all transformations have just resembled those found in section [ grassmannsymmetries ] . there are however further sets of symmetry transformations which only occur for the component lagrangian @xmath108 . the easiest of these is given by @xmath195 where again @xmath190 is a cyclic permutation of @xmath191 . this leads to four charges @xmath196 the next easiest is @xmath197 and leads to six independent charges @xmath198 we shall exploit these charges later but for now we are more interested in where they come from . it turns out that these and two other groups of transformations not mentioned yet reflect invariance of the original lagrangian under change of basis in the grassmann algebra . a grassmann algebra with @xmath5 generators can be viewed as a @xmath8-dimensional vector space in the obvious way , but the choice of the generators @xmath199 and their products as basis vectors is somewhat arbitrary . a change of basis , however , has to be compatible with the multiplicative structure of the algebra . this means firstly that we only have to consider transformations of the @xmath5 generators and secondly that we can change an ( odd ) generator only by another odd element of the algebra . the final constraint is that the highest order monomial , e.g. @xmath200 in the case of four generators , remains unchanged by the transformation so that the highest order lagrangian , here @xmath108 , remains invariant . in the four generator case , this leaves only a relatively small group of acceptable linear transformations . there are eight independent odd basis elements , namely @xmath199 and @xmath201 . correspondingly , there are the following independent infinitesimal transformations that fulfil our conditions : 1 . @xmath202[active3 ] 2 . @xmath203 an even permutation of @xmath191)[active1 ] 3 . @xmath204 : there are just three independent _ scaling _ transformations of this type . [ active4 ] 4 . @xmath205 an even permutation of @xmath191 ) : there are six independent _ grassmann scaling _ transformations of this type.[active2 ] in order for the grassmann quantities @xmath21 , @xmath25 and @xmath26 to be invariant , corresponding to these transformations of the _ basis vectors _ of @xmath1 there have to be transformations of the real _ components_. it is these component transformations that we have found in ( [ lintrafo1 ] ) and ( [ lintrafo2 ] ) , which belong to the basis transformations [ active1 ] . and [ active2 ] . , respectively . the component transformations belonging to [ active3 ] . and [ active4 ] . are also symmetries of the lagrangian , but since they are more complicated to write down and we will not use them later we refrain from giving them here . it should be mentioned though that they generate a full @xmath206 symmetry group . we finally come to the physical interpretation of the results we have so far obtained concerning the component dynamical variables . starting with the lowest order bosonic function @xmath139 we first recall that this always describes the motion of a particle in one dimension in the potential @xmath207 . there are three possible types of motion in one dimension : 1 . movement with no turning point : the particle velocity is always positive ( or negative ) , the motion can be bounded , approaching finite values of @xmath139 as @xmath56 goes to @xmath208 , or unbounded . example : flat potential function @xmath209 . 2 . movement with one turning point : the particle velocity changes sign once , when @xmath139 reaches a maximal ( minimal ) value . again , the motion can be bounded or unbounded . example : reciprocal potential @xmath210 . 3 . oscillatory motion : the particle velocity changes sign infinitely often , thus the motion is always restricted to a finite interval . example : harmonic oscillator @xmath6 . there is one conserved quantity , namely the lowest order hamiltonian @xmath115 , given in ( [ h0 ] ) . we can interpret this as half the squared length of the two - dimensional bosonic vector @xmath211 the conservation of @xmath115 means that the motion of this vector is restricted to a circle with squared radius @xmath212 . corresponding to the three types of motion for @xmath139 there are three types of motion for this vector : 1 . movement with no turning point : the vector moves on the right ( left ) semicircle in the @xmath144-@xmath24-plane . the number of direction changes depends on the number of potential extrema of @xmath24 . 2 . movement with one turning point : motion starts on the right ( left ) semicircle of the @xmath144-@xmath24-plane , then changes when the turning point is reached to the other semicircle where it then mirrors the previous motion . 3 . oscillatory motion : there are two subcases , depending on whether the number of sign changes of @xmath79 between the two extremal points @xmath213 and @xmath214 is odd or even . 1 . when there is an odd number the vector will continually wind around the circle ; direction changes depending on the number of extrema of @xmath24 can be superimposed . this is the only case where the motion covers the complete circle . the most important example is the harmonic oscillator where we can see directly the circular motion of the bosonic vector . when the number of sign changes of @xmath24 is even , the bosonic vector will oscillate on the circle between two points lying symmetrically on either side of the @xmath24-axis . an example of this case is the potential function @xmath215 with @xmath216 , which leads to a harmonic oscillator potential shifted upwards by @xmath85 . the bosonic vector is important because it largely determines the behaviour of the @xmath5 first order fermionic functions . to see this , we look at the @xmath5 two - dimensional vectors @xmath217 the charges @xmath218 introduced in ( [ lincharge1 ] ) guarantee that the lengths of these vectors are conserved , thus their motions are also restricted to circles . the squared radii are given by @xmath219 , providing us with a nice geometrical interpretation of these constants . furthermore , the charges @xmath220 , given by ( [ lincharge2 ] ) , can be interpreted as _ scalar products _ between the @xmath18th and @xmath221th vectors , thus effectively specifying the absolute value of the angle between them . hence we can see already that all @xmath5 vectors must corotate . this can be confirmed further by the charges @xmath130 calculated in ( [ rij ] ) . they can be seen as two - dimensional _ determinants _ , fixing the area of the parallelogram defined by the above - mentioned two vectors . it then follows that the sign of the angle between these is determined , too . the other conserved quantities that involve only terms of first or lower order are the supercharges @xmath145 and @xmath146 , given in ( [ qiqti ] ) . they couple the fermionic vectors with the bosonic vector : the charges @xmath145 fix the scalar products , the constants @xmath146 the determinants between @xmath222 and @xmath223 , so that all angles between these vectors , including their signs , are determined and constant in time . this means that once the bosonic motion is calculated and the fermionic initial data is given , we can read off the time evolution of all fermionic functions since their vectors are rigidly corotating with the bosonic one . having now dealt with the lowest order bosonic and fermionic functions we have to ask how to interpret the higher order functions . unfortunately , the geometric interpretation of these can not be seen as clearly . instead we will interpret the higher order functions as _ variations _ of those of lower order . to start with , @xmath129 as given in ( [ solutxij ] ) has two terms coming from the two homogeneous solutions of the @xmath129-equation with coefficients @xmath128 and @xmath118 and the term with coefficient @xmath224 which constitutes a particular solution . we will now show that the first two terms can be written as variations of the lowest order motion @xmath139 with respect to the two free parameters @xmath124 and @xmath10 of that motion . the variation of @xmath107 with @xmath124 is clearly proportional to the @xmath225 which gives us the term with coefficient @xmath128 . next we vary the equation @xmath132 : @xmath226 using equation ( [ x0 ] ) , dividing by @xmath227 and integrating between @xmath124 and @xmath56 we obtain @xmath228 hence @xmath229_{{\ensuremath{t_{0}}}}^t=\delta e_{0}\int_{{\ensuremath{t_{0}}}}^t \frac{1}{\dot{x}_{0}^{2}}dt'\ ] ] and finally @xmath230 which is exactly the term with coefficient @xmath118 in ( [ solutxij ] ) . so we end up with : @xmath231 in the same way we can analyse the third order fermion solutions ( [ 3rdvar1 ] ) and ( [ 3rdvar2 ] ) . both have parts proportional to @xmath144 and @xmath24 . but in addition there are two further terms which have to be interpreted as variations of these parts with @xmath124 and @xmath10 . the @xmath124-variations can be written down immediately : @xmath232 and yield the terms with coefficients @xmath233}$ ] and @xmath234}$ ] . the @xmath10-variations give : @xmath235 they explain the first part of the terms with coefficients @xmath236}$ ] and @xmath237}$ ] ; note that the second part can be absorbed into the @xmath152 and @xmath153-terms . so the vectors formed by @xmath158 and @xmath159 consist of three parts : a vector rigidly fixed to the familiar first order fermionic vectors and two parts which are proportional to the variation of these vectors with @xmath124 and @xmath10 . finally we want to explain the parts of the highest order bosonic function @xmath175 . the first three terms of ( [ 4var ] ) are familiar : they are the variations of @xmath139 with @xmath124 and @xmath10 and the inhomogeneity term proportional to @xmath24 . the next three terms can be explained as _ second _ variations . so @xmath171 gives the second variation of @xmath139 with respect to @xmath124 ; for the term with coefficient @xmath238}$ ] we have already shown that it is the time derivative of @xmath239 , hence it is proportional to @xmath240 . so we only have to deal with the @xmath177}$]-component . we find that @xmath241 the last term can be rewritten as @xmath242 combining this with the expanded first term @xmath243 we find exactly the component in ( [ 4var ] ) with coefficient @xmath177}$ ] . the last two terms in ( [ 4var ] ) are easy again : they are given by the two first variations of the inhomogeneity term @xmath244 with @xmath124 and @xmath10 , which we have already calculated above . note that the term @xmath245 with coefficient @xmath246}$ ] can be absorbed into the @xmath192 term . in summary the lowest order bosonic function @xmath139 describes the one - dimensional motion of a point particle with energy @xmath10 in the potential @xmath247 . it is completely unaffected by all higher order bosonic and all fermionic functions and not influenced by any higher order charges . there are two free parameters involved , an initial time @xmath124 and the energy @xmath10 . the first order fermionic functions comprise two - dimensional vectors @xmath222 . the lengths of these vectors and the angles between them are fixed by charges arising from symmetry under change of basis in the grassmann algebra and from the second order charge @xmath130 , and so are constant in time , i.e. all fermionic vectors rigidly corotate . their motion is in turn determined by the bosonic vector @xmath223 which has constant squared length @xmath212 and is rigidly coupled to the fermion vectors by the supercharges @xmath145 and @xmath146 . for every generator there is a pair of free parameters which specify the length and the phase of the corresponding vector . the second order bosonic function @xmath129 has terms proportional to the ( first ) variation of @xmath139 with @xmath124 and @xmath10 . there is one additional term proportional to @xmath24 stemming from the inhomogeneity of the @xmath129-equation of motion and ultimately from the yukawa interaction term . the third order fermionic functions again comprise two - dimensional vectors @xmath248 . they can be divided into three additive parts . the first part is rigidly rotating with the first order fermionic vectors , the second and third parts are proportional to the variations of this first order vectorial motion with initial time @xmath124 and @xmath10 . finally , the fourth order bosonic function involves all first and second order variations of @xmath139 with respect to @xmath124 and @xmath10 plus a further term proportional to @xmath24 and _ its _ first order variations . in this section we apply our results of the previous section to oscillatory motion , i.e. we assume that the lowest order bosonic function @xmath139 is periodic with period @xmath249 . we have found already that in this case the first order fermionic motion is described by two - dimensional vectors that either continuously wind around a circle or oscillate between two symmetrically lying points on it , depending on the number of sign changes of @xmath24 . when we look at the next bosonic level , i.e. the functions @xmath250 , we can see immediately that the first part of the solution ( [ solutxij ] ) , which is proportional to @xmath144 , is a periodic function with the same period @xmath249 . the last term proportional to @xmath251 is periodic , too , but it can have period @xmath252 if the potential function @xmath104 has reflection symmetry , i.e. if @xmath253 for some constant @xmath254 . let therefore @xmath255 , which is either @xmath249 or @xmath252 , denote the period of @xmath256 . the most interesting term is the remaining second one containing the integral . we will analyse this term from a mathematical point of view using floquet theory but before we do that we would like to understand it from a more physical perspective . from equation ( [ xijvar ] ) in the previous section we know that the integral term can be interpreted as the variation of the solution @xmath139 with respect to the energy @xmath10 . because the variation is infinitesimally small , we can make the assumption that the functional form of the motion remains unchanged . we treat the motion as determined by only two parameters , the period @xmath249 and a characteristic amplitude @xmath257 , defined e.g. as the distance between the two turning points of the motion . in the generic case @xmath249 and @xmath257 will be related by a non - trivial function @xmath258 , therefore a change in energy means not only a change in the amplitude @xmath257 but also in the period @xmath249 , so : @xmath259 since @xmath139 is a @xmath249-periodic function we can use its fourier series @xmath260 , where @xmath261 and the coefficients @xmath262 are regarded as functions of @xmath257 , to find : @xmath263 thus @xmath264 both @xmath265 and @xmath144 are @xmath249-periodic functions , but @xmath266 itself is clearly not since the second term is _ linear_-periodic , i.e. the product of the linear term @xmath56 with a periodic function . the extra linear - periodic term diverges with time , thus seemingly spoiling the oscillatory nature of the solution . however , this problem only arises because the period of the oscillation depends on the amplitude @xmath257 and thus ultimately on the energy @xmath10 . expressing a solution with slightly changed period @xmath267 in terms of @xmath249-periodic functions inevitably leads to non - periodic terms . the effect is well - known from celestial mechanics where linear - periodic functions appear as secular terms in the study of stability problems of planetary orbits ( see e.g. @xcite ) . there is one case where no linear - periodic term occurs , and this is when @xmath268 i.e. when the period is independent of the amplitude . this is true for the harmonic oscillator , and consequently there is no non - periodic term in equation ( [ harmxijsolut ] ) . the existence of non - periodic terms in the bosonic function @xmath250 can be derived in a mathematically more stringent way from floquet theory , the theory of linear differential equations with periodic coefficients . equation ( [ xij ] ) , with @xmath269 replaced by the constant @xmath130 according to ( [ rij ] ) , determines the motion of @xmath250 . because we know that the particular solution proportional to @xmath24 is periodic it suffices to treat the homogeneous equation which can be written in the form @xmath270 where @xmath271 is a @xmath255-periodic function . this allows us to apply floquet theory which states that @xcite : 1 . there exists a non - zero constant @xmath272 , called the characteristic multiplier , and a non - trivial solution @xmath273 such that @xmath274 from which one deduces 2 . there exist linearly independent solutions @xmath275 and @xmath276 to ( [ floquet ] ) , such that either @xmath277 or @xmath278 where in both cases @xmath279 are @xmath255-periodic functions and @xmath280 , @xmath281 , @xmath282 , called characteristic exponents , are not necessarily distinct constants . whether the solutions take the form ( [ floquet1 ] ) or ( [ floquet2 ] ) depends on whether there are two independent solutions of ( [ floquet ] ) with the property ( [ multiplier ] ) or just one . if we denote by @xmath283 the two linearly independent solutions of ( [ floquet ] ) with @xmath284 and by @xmath285 the matrix @xmath286 we get a solution of type ( [ floquet1 ] ) if @xmath287 is diagonalizable and a solution of type ( [ floquet2 ] ) if @xmath287 has jordan normal form . to determine the characteristic exponents @xmath288 it is useful to notice that our equation is not an arbitrary differential equation with periodic coefficients but an example of hill s equation which takes the form @xmath289 @xmath54 and @xmath290 being @xmath255-periodic functions . for equations of hill s type the floquet multipliers are completely determined by the trace @xmath291 of the fundamental matrix @xmath287 . this trace can be determined indirectly , using the fact that we know one periodic solution namely @xmath144 , which has either period @xmath255 or @xmath292 , depending on whether @xmath293 is reflection - symmetric or not . now we can use another theorem of floquet theory which states : 1 . hill s equation has non - trivial solutions with period @xmath255 if and only if @xmath294 . then either @xmath295 and @xmath296 or @xmath297 and @xmath298 2 . hill s equation has non - trivial solutions with period @xmath292 if and only if @xmath299 . then either @xmath300 and @xmath301 or @xmath302 and @xmath303 using these theorems we find that the second independent solution ( [ xijvar ] ) has to be either periodic or linear - periodic . the period is in both cases @xmath249 , the period of the first independent solution @xmath144 . this means that we can formally confirm our earlier result derived from physical arguments . additionally , we have that any solution must be bounded in finite intervals . this is not immediately obvious from the explicit form of ( [ xijvar ] ) , which could be singular when @xmath304 . before we proceed to analyse a particular potential function which admits oscillatory motion we want to comment briefly on the higher order bosonic and fermionic quantities @xmath305 and @xmath175 . from equations ( [ 3rdvar1 ] ) and ( [ 3rdvar2 ] ) we can see that all terms that appear in the third order fermionic solution are either periodic or linear - periodic . for the term @xmath306 this follows e.g. from the fact that the derivative of a linear - periodic term is again linear - periodic . we can expect this behaviour physically since @xmath307 and @xmath308 are given by the _ first _ variation of the motion of the first order fermionic vectors with energy and initial time . ( there are further periodic contributions to @xmath307 and @xmath308 with coefficients @xmath152 and @xmath153 . ) most terms in the fourth order bosonic solution @xmath175 can be easily seen to be periodic or linear - periodic in the same manner . however , there are the three terms which have been interpreted as the _ second _ variation of @xmath139 with energy @xmath10 . these should contain a quadratic term in @xmath56 , because @xmath309 \left(\frac{da}{de_{0}}\right)^{2}+\left[\frac{\partial { \ensuremath{x_{0}}}}{\partial a } + \frac{\partial { \ensuremath{x_{0}}}}{\partial t}\frac{dt}{da}\right ] \frac{d^{2}a}{de_{0}^{2}}\ ] ] and @xmath310 all other terms being ( linear-)periodic . again , the quadratic term does not occur when period @xmath249 and amplitude @xmath257 are independent , i.e. when @xmath311 . this is confirmed by the harmonic oscillator example where @xmath175 is completely periodic . we can also understand the existence of a quadratic - periodic term directly from the which contains the term @xmath312 we know already that the integral will yield a term proportional to @xmath56 ; this is squared to give a quadratic term in @xmath56 , and then multiplied by the periodic function @xmath171 . this quadratic term can not cancel out because the two remaining terms with coefficient @xmath177}$ ] in ( [ 4var ] ) are at most linear - periodic . we illustrate our results by the hyperbolic potential @xmath84 , already discussed in section [ grassmannhyperbol ] for general grassmann algebra . since we are only interested in oscillatory motion here , we make the assumption that @xmath313 . inverting the result obtained from ( [ solutxn ] ) we can calculate that @xmath314 with @xmath315 so @xmath139 is indeed a periodic function . for @xmath144 and @xmath24 we immediately find that @xmath316 where we have abbreviated @xmath317 to @xmath318 . therefore the motion of the bosonic is a _ non - uniform rotation _ , which confirms our comments in section [ interpret ] since the potential function @xmath24 changes sign only once . all fermionic vectors corotate with the bosonic one . to determine @xmath250 we need the integral @xmath319 we then obtain for the @xmath118-related part of @xmath250 : @xmath320 this result clearly shows a linear periodic term which means that the period - amplitude relation is non - trivial . in fact , period and amplitude of @xmath321 , the latter defined as distance between the two turning points , can be easily calculated using ( [ xnhyperbol ] ) : @xmath322 thus @xmath323 and @xmath324 . we point out that all terms in the solution are clearly bounded , which is not the case for the integral itself . figure [ diagram2 ] gives an example of both the bosonic quantities @xmath325 and @xmath250 and the fermionic quantities @xmath326 and @xmath327 . we do not present the higher order solutions in detail here since their explicit form , although not complicated to find , is lengthy and does not provide new insights . the only interesting exception is the part of @xmath175 which has been characterized as a second variation with energy . this part can be calculated to be @xmath328}\left [ \frac{(\alpha^{3}\!-\!3\alpha^{2}\!+\!6\alpha)\sin y\sec^{3}y + ( 2\alpha^{2}\!-\!4\alpha)\sin y \sec y } { k(c^{2}\!-\!2e_{0})^{2}\sqrt{(\alpha \sec^{2}y\!-\!1)^{3}}}\right.\\ & & \left.+(t-{\ensuremath{t_{0}}})\left(\frac{\left(-2\alpha^{2 } + \alpha\right)\sec^{2}y+1}{\sqrt{(c^{2}\!-\!2e_{0})^{3 } } \sqrt{(\alpha\sec^{2}y\!-\!1)^{3}}}\right ) + ( t-{\ensuremath{t_{0}}})^{2}\left(-\frac{k\alpha}{c^{2}\!-\!2e_{0 } } \frac{\sin y \sec^{3}y}{\sqrt{(\alpha \sec^{2}y\!-\!1)^{3 } } } \right ) \right],\end{aligned}\ ] ] where we have abbreviated @xmath329 to @xmath272 . again , there are two noteworthy points here : firstly , the occurence of a quadratic - periodic term , which has been predicted in section [ interpret ] and secondly the complete boundedness of the whole solution , which is not immediately obvious from the functional form ( [ 4var ] ) , especially regarding the term where the velocity appears in the denominator . next we want to show that the two solution methods presented in this paper deliver the same result . therefore we have to decompose our previous solution , the first of equations ( [ xnhyperbolgrassmann ] ) , from section [ grassmannhyperbol ] . recall that any real analytical function @xmath330 is extended to a grassmann - valued function by the formula @xmath331 where @xmath332 and @xmath333 are body and soul of the grassmann number @xmath45 , respectively @xcite . to keep the calculation short we will expand here only up to second order , i.e. the following equations have to be understood modulo a component proportional to @xmath200 . we begin with the frequency @xmath334 : @xmath335 then @xmath336 since @xmath337 we find @xmath338\xi_{ij},\end{aligned}\ ] ] where again @xmath339 . this result equals @xmath340 ; decomposition yields : @xmath341 thus by comparison @xmath342 which gives the same solution for @xmath139 as ( [ xnhyperbol ] ) . similarly , we can read off the result for @xmath343 and then calculate @xmath250 : @xmath344 the term with coefficient @xmath130 can be readily identified as the potential function @xmath345 , derived in ( [ hyperbolu ] ) . the second term with coefficient @xmath346 equals the velocity @xmath144 , which we have calculated in ( [ hyperbolxdot ] ) , when we make the identification . finally , the term proportional to @xmath118 is identical with our . so the solution found by using our layer - by - layer approach can also be obtained by the decomposition of the full grassmann solution calculated through direct integration in grassmann space . we have demonstrated this feature here using the bosonic quantity @xmath57 but we could equally well have chosen one of the fermionic quantities @xmath347 or @xmath348 . this means that our two solution methods are compatible with each other . we have already mentioned in sections [ solfixpot ] and [ solarbpot ] that the case where the energy @xmath10 , the body of the full grassmann energy @xmath4 , equals zero has to be treated differently . all the solutions to the equations of motion ( [ x0])([lijkrijk ] ) were given under the restriction @xmath349 so that we can not use our previous results . it will turn out , however , that for a finitely generated grassmann algebra we can derive all solutions in explicit functional form for arbitrary potential . starting with equation ( [ h0 ] ) we can see that if @xmath350 both @xmath144 and @xmath104 have to vanish : @xmath351 this means that the particle stays permanently at rest at a minimum @xmath352 of the one could assume at this point that all higher order bosonic and fermionic functions are trivial as well , but as we shall soon see this is far from being necessary , in contradiction to @xcite . because @xmath107 is constant , all the ( spatial ) derivatives of @xmath24 are constant functions of time too ; we denote their values by @xmath353 we assume in the following that @xmath354 . when we now look at the first order fermionic equations ( [ liri ] ) for @xmath326 and @xmath327 , we find that they are easily solved by the harmonic oscillator ansatz : @xmath355 unlike the true harmonic oscillator , the constants @xmath156 and @xmath157 are , however , not linked to the first order supercharges @xmath145 and @xmath146 which vanish completely as can be verified from ( [ qiqti ] ) . to understand the first order fermionic motion we assume that the energy @xmath10 is small but non - zero , restricting the particle motion to a small neighbourhood of the stability point @xmath352 where the potential function @xmath24 is approximately linear . the result is an almost harmonic oscillation with frequency @xmath356 , mirrored by the fermionic quantities as follows from ( [ fermionharm1 ] ) and ( [ fermionharm2 ] ) . when we now let @xmath10 approach zero , this fermionic motion seems to diverge as one can see from the formulae for the coefficients @xmath166 and @xmath167 in ( [ li ] ) and ( [ ri ] ) . however , there is a subtle point here : while the supercharges @xmath145 and @xmath146 are arbitrarily chosen constants for @xmath357 , they have to vanish for @xmath350 . thus to avoid any discontinuities they have to smoothly approach zero as the energy decreases . exactly how they approach zero finally determines which value the harmonic oscillator coefficients @xmath166 and @xmath167 take in the limit @xmath350 , which justifies the two remaining degrees of freedom in our solution , @xmath156 and @xmath157 . next we want to analyse the second order bosonic quantity @xmath250 . because @xmath23 and @xmath106 are constants the equation of motion ( [ xij ] ) describes formally a harmonic oscillator with frequency @xmath356 subject to a constant external force @xmath358 , so the solution ( for @xmath359 ) is @xmath360 where @xmath361 and @xmath362 are integration constants . to interpret this result we recall that in case @xmath357 the homogeneous part of @xmath250 consists of two terms , the two variational derivatives of @xmath139 with respect to @xmath10 and @xmath124 . when @xmath10 is zero , a small change in energy will result in oscillatory motion around the stability point @xmath352 . this explains the functional form and one of the free parameters of the solution ( [ e0xij ] ) . the second parameter , though , can not be connected to the variation with @xmath124 anymore since this variation is zero for a constant @xmath139 . an interesting observation is connected to the second order energy which can be calculated from ( [ hij ] ) to be @xmath363 . thus the second order energy is not independent any more but determined by the four first order constants @xmath156 , @xmath157 , @xmath364 , @xmath365 . note the fact that @xmath118 is _ not _ connected to the parameters @xmath361 and @xmath362 . the third order fermion equations ( [ lijkrijk ] ) are again equations describing a harmonic oscillator with a driving term . the homogeneous part of the solution looks therefore the same as ( [ e0li ] ) and ( [ e0ri ] ) with integration constants @xmath152 and @xmath153 . to find the particular solution we have to investigate the driving terms @xmath366}$ ] and @xmath367}$ ] further . using equations ( [ e0li ] ) , ( [ e0ri ] ) and ( [ e0xij ] ) we find @xmath368}&=&c_{ijk}+d_{1,ijk}\sin 2k_{1}(t-{\ensuremath{t_{0 } } } ) - d_{2,ijk } \cos 2k_{1 } ( t-{\ensuremath{t_{0}}})\\ x_{[ij}\lambda_{k]}&=&\tilde{c}_{ijk}+d_{2,ijk } \sin 2k_{1}(t-{\ensuremath{t_{0 } } } ) + d_{1,ijk } \cos 2k_{1}(t-{\ensuremath{t_{0}}}),\end{aligned}\ ] ] where @xmath369 , @xmath370 , @xmath371 and @xmath372 are constants built from @xmath156 , @xmath157 , @xmath361 and @xmath362 . so the driving terms are not constant but have an oscillating part that oscillates with _ double _ the basic frequency @xmath356 . the particular solutions are thus @xmath373 the fermionic vectors @xmath248 consist of three parts : a uniform rotation with natural frequency @xmath356 , a forced uniform rotation with double this frequency that we can interpret as the _ first excited _ mode and a constant shift away from the origin . note that there is no double frequency term if @xmath374 , and in particular for the harmonic oscillator potential . this agrees with our earlier result in section [ harmpot ] . calculating the supercharges @xmath154 and @xmath155 from ( [ qiqti ] ) we find that @xmath375 and @xmath376 which gives us a neat interpretation of the displacement constants @xmath369 and @xmath370 . it further means that the third order supercharges are not related to the free parameters @xmath152 and @xmath153 as is the case for @xmath377 but instead are completely determined by first and second order parameters . this is the same phenomenon already encountered for the second order energy @xmath118 and again it means that we should not think of @xmath154 and @xmath155 as independent constants when @xmath350 . finally we have to investigate the fourth order bosonic quantity @xmath175 . the equation of motion ( [ x1234 ] ) simplifies in the case @xmath350 to @xmath378}+ k_{2}r_{1234}+k_{3}x_{[12}r_{34]},\ ] ] so it describes a harmonic oscillator with three driving terms . the homogeneous part is analogous to ( [ e0xij ] ) with two integration constants @xmath138 and @xmath379 ; the comments made about @xmath250 apply here , too . analyzing the three driving terms we first note that @xmath380 specifies a constant external force , playing the same role as @xmath358 for the variables @xmath250 . the first and third terms can be calculated using the explicit formulae for @xmath250 , resulting in @xmath381}+k_{3}x_{[12}r_{34 ] } & = & -\frac{3}{4}k_{1}k_{2}\left(c_{1234}+d_{1234 } \cos2k_{1}(t\!-\!{\ensuremath{t_{0}}})+\tilde{d}_{1234}\sin 2k_{1}(t\!-\!{\ensuremath{t_{0}}})\right)\\ & & + \left(k_{3}\!-\!3\frac{k_{2}^{2}}{k_{1}}\right)\left(f_{1234 } \cos k_{1}(t\!-\!{\ensuremath{t_{0 } } } ) + \tilde{f}_{1234 } \sin k_{1}(t\!-\!{\ensuremath{t_{0}}})\right),\end{aligned}\ ] ] where @xmath382 , @xmath383 , @xmath384 , @xmath385 and @xmath386 are fixed constants built from @xmath361 , @xmath362 and @xmath130 . so besides a further constant term we find four oscillating ones , two with frequency @xmath356 and two with frequency @xmath387 . the latter ones will cause forced harmonic motion with double the natural frequency as we found for the third order fermionic terms . new , however , are the oscillating terms which have the _ same _ frequency as the homogenous solution . this means that the oscillator is driven in resonance . therefore the solution is not bounded anymore but includes linear - periodic terms like @xmath388 and @xmath389 . combining all contributions we find the particular solution @xmath390 thus the fourth order bosonic solution consists of harmonic oscillation with the natural frequency @xmath356 , an excited mode with twice this frequency , linear - periodic motion and a constant shift . especially interesting are the linear - periodic terms . they can be explained physically as in section [ specialtypesofmotion ] , although we have to see them as the _ second _ variation with energy here : when the particle stays at rest in a potential minimum the first variation gives rise to harmonic oscillation , independently of the shape of the potential function @xmath24 ( as long as @xmath354 ) . only when we look at the second order variation do non - harmonic terms come into play : the period of oscillation will in general depend on the amplitude and the energy , therefore a variation in energy results in a change of period ; expressing the new solution in terms of the old period then generally leads to secular , non - periodic terms . notice the fascinating fact that although the physical interpretation is similar , the way the linear - periodic terms arise mathematically is completely different : for non - zero energy oscillations they come as solutions of a _ homogenous _ differential equation with _ periodic _ coefficients ; in the zero energy case they are particular solutions to an _ inhomogenous _ differential equation with _ constant _ coefficients . in this work we have analysed a supersymmetric mechanical model from two different viewpoints : either we make no specification with regard to the nature of the underlying grassmann algebra @xmath1 . then for a range of potential functions @xmath24 the model can be explicitly solved . or we regard @xmath1 as finitely generated thus reducing all quantities to a set of real functions and their interrelationships . then we are able to solve the system completely , without any restrictions on @xmath24 . the methods have been shown to be compatible with each other here . an open question with our first point of view is what meaning we can give to the full , i.e. non - decomposed grassmann solutions @xmath391 and @xmath348 . interpretation may be aided by our second approach , where we have shown that the component solutions include the purely classical motion dependent on @xmath10 _ and _ all its variations with respect to this ( real ) constant . apparently , supersymmetric dynamics captures information about a whole range of energies of the mechanical system , so in a sense we can say that the grassmann energy @xmath4 corresponds to some fuzzy classical energy . this suggests possibly that supersymmetric classical dynamics is closely related to the quantum dynamics . a study of the quantized version of our model would therefore be an important step to complement this research . we also hope to apply our methods to the full , i.e. space- and time - dependent field theory . thanks p. bongaarts and r. casalbuoni for correspondence , and a. rogers for discussions . r.h . gratefully acknowledges the support of epsrc and the studienstiftung des deutschen volkes . 99 f. a. berezin and m. s. marinov , _ ann . phys . _ * 104 * ( 1977 ) , 336 . r. casalbuoni , _ nuovo cim . _ * 33a * ( 1976 ) , 389 . b. de witt , `` supermanifolds , '' 2nd ed . , cup , cambridge , 1992 . m. s. p. eastham , `` the spectral theory of periodic differential equations , '' scottish academic press , edinburgh , 1973 . h. goldstein , `` classical mechanics , '' 2nd ed . , addison - wesley , reading , 1980 ;	we analyse a supersymmetric mechanical model derived from @xmath0-dimensional field theory with yukawa interaction , assuming that all physical variables take their values in a grassmann algebra @xmath1 . utilizing the symmetries of the model we demonstrate how for a certain class of potentials the equations of motion can be solved completely for any @xmath1 . in a second approach we suppose that the grassmann algebra is finitely generated , decompose the dynamical variables into real components and devise a layer - by - layer strategy to solve the equations of motion for arbitrary potential . we examine the possible types of motion for both bosonic and fermionic quantities and show how symmetries relate the former to the latter in a geometrical way . in particular , we investigate oscillatory motion , applying results of floquet theory , in order to elucidate the role that energy variations of the lower order quantities play in determining the quantities of higher order in @xmath1 .
in @xcite we studied the problem of complete controllability of finite - level quantum systems with nearest - neighbour interactions . we showed that many quantum systems of physical interest are indeed completely controllable but that there are nevertheless systems with certain symmetries that are not completely controllable . this paper is devoted to identifying the dynamical lie algebras for the latter systems . as in the previous paper , we consider the case of a driven quantum system , for which the interaction with the control field is linear , i.e. , we assume that the total hamiltonian of the system is @xmath6 where @xmath7 is the internal system hamiltonian and @xmath8 represents the interaction of the system with the real control field @xmath9 . we assume that @xmath7 and @xmath8 are hermitian . for a finite - level quantum system there always exists a complete orthonormal set of energy eigenstates @xmath10 such that @xmath11 and thus the internal hamiltonian can be expanded in terms of the energy eigenfunctions @xmath10 , @xmath12 where @xmath13 is an @xmath14 matrix with elements @xmath15 and @xmath16 are the energy levels of the system . the @xmath16 are real since @xmath7 is hermitian . we shall assume that the energy levels are ordered in a non - decreasing sequence , i.e. , @xmath17 . hence , the frequencies for transitions @xmath18 are non - negative @xmath19 in the following it will be convenient to deal with trace - zero operators . thus , if @xmath7 has non - zero trace then we define the trace - zero operator @xmath20 i_n,\ ] ] which is equivalent to @xmath7 up to addition of a constant multiple of the identity matrix @xmath21 . expanding the interaction hamiltonian @xmath8 with respect to the complete set of orthonormal energy eigenstates @xmath10 leads to @xmath22 where the transition dipole moments @xmath23 , which we assume real , satisfy @xmath24 . in this paper we shall only be concerned with quantum systems for which the interaction with the control field is determined by transitions between adjacent energy levels , as is typical in the dipole approximation . it will also be assumed that there are no ` self - interactions ' , i.e. , that the diagonal elements @xmath25 are zero for all @xmath26 . thus , letting @xmath27 for @xmath28 we have @xmath29 the operators @xmath30 and @xmath31 generate a lie algebra @xmath32 called the dynamical lie algebra of the control system . this lie algebra is important since it determines the lie group @xmath33 on which the control system evolves @xcite . precisely speaking , the trajectories of the system subject to any control field are confined to the exponential image of the lie algebra @xmath32 . knowledge of the dynamical lie algebra thus enables us to determine the degree of controllability of a quantum system @xcite , to identify reachable or non - reachable target states @xcite , and to determine whether a kinematical bound for an observable is dynamically accessible @xcite . the dynamical lie algebra @xmath32 generated by the operators @xmath30 and @xmath31 defined in ( [ eq : hzero ] ) and ( [ eq : hone ] ) is a real lie algebra of @xmath14 skew - hermitian matrices , and the related lie algebra @xmath34 generated by @xmath35 and @xmath31 is a real lie algebra of traceless , skew - hermitian matrices . thus , @xmath34 is always a subalgebra of @xmath36 . since @xmath32 is isomorphic to @xmath37 if @xmath38 and @xmath39 if @xmath40 , it suffices to determine @xmath34 . it follows from classical results that a pair of skew - hermitian matrices in @xmath36 almost always generates the full lie algebra @xmath36 ( see lemma 4 in @xcite , for example ) . for the type of quantum systems considered in this paper , explicit criteria ensuring @xmath41 have been established @xcite : [ thm : a ] let @xmath42 and @xmath43 for @xmath44 . the dynamical lie algebra @xmath34 generated by @xmath35 and @xmath31 defined in ( [ eq : hzerop ] ) and ( [ eq : hone ] ) is @xmath36 if @xmath45 , @xmath46 for @xmath47 , and one of the following criteria applies : there exists @xmath48 such that @xmath49 for @xmath50 , or 2 . @xmath51 for @xmath47 but there exists @xmath52 such that @xmath53 for @xmath50 . if @xmath54 then @xmath55 for some @xmath56 is required as well . as has been shown in @xcite , many quantum systems of physical interest indeed satisfy these criteria . however , there are systems of physical interest that do not meet these criteria . for instance , if any of the dipole moments @xmath57 vanish then the system decomposes into independent subsystems and its dynamical lie algebra @xmath34 is a sum of subalgebras of @xmath36 @xcite . but even if all the @xmath57 are non - zero , the dynamical lie algebra of the system may be a proper subalgebra of @xmath36 , for example , if the transition frequencies @xmath58 and the transition dipole moments @xmath57 satisfy @xmath59 as is the case for a system with @xmath0 equally spaced energy levels and uniform dipole moments . in the following we show that the dynamical lie algebra @xmath34 of such a system is a subalgebra of @xmath5 if @xmath2 , and a subalgebra of @xmath3 if @xmath60 , and give criteria ensuring @xmath61 or @xmath62 , respectively . in [ app : d ] , we also briefly discuss why the lie algebra @xmath63 does not arise for the systems considered in this paper . consider a system with hamiltonian @xmath65 , where @xmath66 @xmath67 , @xmath68 , @xmath69 for all @xmath26 , and the transition frequencies @xmath70 and transition dipole moments @xmath57 satisfy the symmetry relation ( [ eq : sym ] ) . we shall prove that the lie algebra @xmath34 is a subalgebra of @xmath5 , which is in general isomorphic to @xmath5 . we show first that @xmath71 . let @xmath72 . using @xmath73 and @xmath74 we can simplify @xmath31 , @xmath75 to compute @xmath76 , we note that @xmath77 . thus , using @xmath78 leads to @xmath79 hence , the energy levels @xmath80 of @xmath76 are @xmath81 and @xmath82 for @xmath83 . consequently , we have @xmath84 let @xmath85 be an isomorphism of the hilbert space of pure states defined by @xmath86 and set @xmath87 as well as @xmath88 and @xmath89 . then the representations of @xmath35 and @xmath31 with respect to the new basis @xmath90 are @xmath91 with @xmath92 and @xmath93 as defined in ( [ eq : basisb ] ) and ( [ eq : yb ] ) , respectively . hence , @xmath35 and @xmath31 are both in @xmath5 and thus the lie algebra @xmath34 they generate must be contained in @xmath5 . since @xmath35 and @xmath31 in ( [ eq : sysb1 ] ) contain a complete set of generators @xmath92 and @xmath93 for @xmath5 ( see [ app : b ] ) , it is natural to expect that they generate the full lie algebra @xmath5 . we shall prove that this is usually , but not inevitably , true . [ ex : b1 ] consider a system of type ( [ eq : sysb2 ] ) for @xmath94 . if @xmath95 , @xmath96 and @xmath97 , @xmath98 and @xmath99 then the basis change ( [ eq : sigmab ] ) leads to @xmath100 with @xmath92 and @xmath93 as defined in ( [ eq : basisb ] ) and ( [ eq : yb ] ) , respectively . therefore , the lie algebra @xmath34 generated by @xmath35 and @xmath31 is a subalgebra of @xmath101 . however , it is easy to verify that @xmath102 . indeed , in this particular case @xmath34 is a three - dimensional subalgebra of @xmath101 spanned by @xmath30 , @xmath31 and @xmath103=\sqrt{6}x_1+\sqrt{5}x_2+\sqrt{3}x_3 $ ] . thus , for certain choices of the parameters @xmath104 and @xmath105 , the lie algebra @xmath34 is a _ proper _ subalgebra of @xmath5 . to find criteria that ensure @xmath61 , consider the generic system @xmath106 with @xmath92 and @xmath93 as defined in ( [ eq : basisb ] ) and ( [ eq : yb ] ) , respectively . as before , @xmath30 and @xmath31 are in @xmath5 and hence the lie algebra @xmath34 they generate must be contained in @xmath5 . [ thm : b1 ] let @xmath107 for @xmath108 and @xmath109 . the dynamical lie algebra @xmath34 generated by the system @xmath110 with @xmath35 and @xmath31 as in ( [ eq : sysb ] ) is @xmath5 if @xmath111 for @xmath112 . using the properties of the generators @xmath92 and @xmath93 leads to : @xmath113,\rmi h_0 ' ] \\ & = \sum_{m=1}^\ell \delta_m\omega_{m-1}^2 y_m \\ v^{(1)}&\equiv [ [ \rmi h_0 ' , v^{(0)}],\rmi h_0']-\omega_{\ell-1}^2 v^{(0 ) } \\ & = \sum_{m=1}^{\ell-1}\delta_m\omega_{m-1}^2(\omega_{m-1}^2-\omega_{\ell-1}^2)y_m\\ v^{(2)}&\equiv [ [ \rmi h_0 ' , v^{(1)}],\rmi h_0']-\omega_{\ell-2}^2 v^{(1 ) } \\ & = \sum_{m=1}^{\ell-2 } \delta_m \omega_{m-1}^2(\omega_{m-1}^2-\omega_{\ell-1}^2 ) ( \omega_{m-1}^2-\omega_{\ell-2}^2 ) y_m\\ & \vdots \\ v^{(\ell-1 ) } & \equiv [ [ \rmi h_0',v^{(\ell-2)}],\rmi h_0']-\omega_1 ^ 2 v^{(\ell-2)}\\ & = \delta_1 \omega_0 ^ 2 \prod_{m=1}^{\ell-1}(\omega_0 ^ 2-\omega_m^2 ) y_1 . \end{aligned}\ ] ] by hypothesis @xmath111 for @xmath114 and @xmath115 , @xmath116 . hence , all the factors in the last expression above are non - zero , i.e. , we have @xmath117 and thus @xmath61 by lemma [ lemma : b ] of [ app : b ] . if @xmath118 for some @xmath114 then a slight modification of the proof above leads to a residual term @xmath119 where @xmath120 with @xmath121 for @xmath122 and @xmath123 otherwise . if the energy levels are either positive and ordered in a non - decreasing sequence , i.e. , @xmath124 , or negative and ordered in a non - increasing sequence , i.e. , @xmath125 , then @xmath126 implies @xmath127 for all @xmath128 . we shall only consider this case in the following . [ thm : b2 ] let @xmath129 for @xmath112 , where @xmath130 , @xmath131 . the dynamical lie algebra @xmath34 generated by the system @xmath110 with @xmath35 and @xmath31 as in ( [ eq : sysb ] ) is @xmath5 if @xmath127 but @xmath132 for all @xmath133 . since @xmath127 for all @xmath128 , we have @xmath134 \\ z & \equiv 2^{-1 } [ x^{(0)},y^{(0 ) } ] = \sum_{m=1}^\ell ( \tilde{\delta}_{m+1}^2-\tilde{\delta}_m^2 ) h_m.\end{aligned}\ ] ] suppose @xmath135 has @xmath136 elements labeled @xmath137 , @xmath138 up to @xmath139 . if @xmath132 for all @xmath133 then @xmath140-v_{m_{\ell ' } } y^{(0 ) } } \\ & = \tilde{\delta}_1 ( v_1-v_{m_{\ell ' } } ) y_1 -\sum_{k=1}^{{\ell'}-1}\tilde{\delta}_{m_k}(v_{m_k}-v_{m_{\ell ' } } ) y_{m_k}\\ x^{(1 ) } & \equiv { [ y^{(0)},z]-v_{m_{\ell ' } } x^{(0 ) } } \\ & = \tilde{\delta}_1(v_1-v_{m_{\ell ' } } ) x_1 -\sum_{k=1}^{{\ell'}-1}\tilde{\delta}_{m_k}(v_{m_k}-v_{m_{\ell ' } } ) x_{m_k}\\ y^{(2 ) } & \equiv { [ z , x^{(1)}]-v_{m_{{\ell'}-1 } } y^{(1 ) } } \\ & = \tilde{\delta}_1(v_1-v_{m_{{\ell'}}})(v_1-v_{m_{{\ell'}-1 } } ) y_1 -\sum_{k=1}^{{\ell'}-2}\tilde{\delta}_{m_k}(v_{m_k}-v_{m_{{\ell ' } } } ) ( v_{m_k}-v_{m_{{\ell'}-1 } } ) y_{m_k}\\ x^{(2 ) } & \equiv { [ y^{(1)},z]-v_{m_{{\ell'}-1 } } x^{(1 ) } } \\ & = \tilde{\delta}_1(v_1-v_{m_{{\ell'}}})(v_1-v_{m_{{\ell'}-1 } } ) x_1 -\sum_{k=1}^{{\ell'}-2}\tilde{\delta}_{m_k}(v_{m_k}-v_{m_{{\ell ' } } } ) ( v_{m_k}-v_{m_{{\ell'}-1 } } ) x_{m_k}\\ & \vdots & \\ y^{(\ell ' ) } & \equiv { \tilde{\delta}_1 \prod_{k=1}^{\ell ' } ( v_1-v_{m_k } ) y_1}\end{aligned}\ ] ] shows that @xmath141 and hence @xmath61 by lemma [ lemma : b ] of [ app : b ] . a similar argument shows that if there exists @xmath142 such that @xmath143 for @xmath144 but @xmath145 , then @xmath146 . using the fact that the generators @xmath93 of @xmath5 are not diagonal with respect to the cartan elements ( see [ app : b ] ) , it generally follows that @xmath34 contains all the generators @xmath147 and @xmath93 and thus @xmath148 as well . an important special case of this type is a system with @xmath2 equally spaced energy levels and uniform transition dipole moments : [ thm : b3 ] the dynamical lie algebra @xmath34 generated by the system @xmath110 with @xmath2 equally spaced energy levels @xmath149 and uniform dipole moments @xmath150 is @xmath5 . let @xmath151 and @xmath152 $ ] . then @xmath153 $ ] , @xmath154 $ ] and @xmath155 $ ] . thus , @xmath156 , @xmath157 . next , set @xmath158 and @xmath159 for @xmath160 , and note that @xmath161 $ ] , @xmath162 $ ] and @xmath163 $ ] . this shows that @xmath34 contains all the generators @xmath147 and @xmath93 of @xmath5 . hence , @xmath61 . we shall now return to the original system ( [ eq : sysb2 ] ) . since we have assumed that the energy levels are ordered in a non - decreasing sequence , we have @xmath164 for all @xmath165 and hence @xmath166 i.e. , the energy levels @xmath167 are negative and form a decreasing ( non - increasing ) sequence . noting that @xmath168 , we thus have @xmath169 and @xmath170 if @xmath171 and @xmath172 for @xmath173 then @xmath61 according to theorem [ thm : b1 ] , independent of the choice of the dipole moments , provided that they are non - zero . if @xmath174 for some @xmath173 then theorem [ thm : b2 ] applies and the dipole moments determine whether we have @xmath61 or a proper subalgebra . in particular , the lie algebra is @xmath5 if the energy levels are equally spaced , @xmath174 for all @xmath165 , and the dipole moments @xmath105 are such that @xmath175 for all @xmath176 , where @xmath177 for @xmath178 and @xmath179 ( with @xmath180 ) . for instance , setting @xmath181 for @xmath112 works since it gives @xmath182 but @xmath183 for @xmath176 . setting @xmath184 also gives @xmath61 due to theorem [ thm : b3 ] . however , recall that there are systems of type ( [ eq : sysb2 ] ) whose lie algebra @xmath34 is a proper subalgebra of @xmath5 , as example [ ex : b1 ] above clearly shows . note that in this example , the energy levels are equally spaced and the dipole moments @xmath105 are such that @xmath185 , @xmath186 and @xmath187 , i.e. , all the @xmath188 are equal and none of the theorems above are applicable . let @xmath4 and consider the system @xmath65 with @xmath190 where @xmath191 and @xmath192 for @xmath193 . note the symmetry of the system . every transition frequency except @xmath194 occurs in pairs , and although @xmath194 may be different from all the other @xmath58 , theorem [ thm : a ] does not apply since @xmath4 and there is no @xmath195 such that @xmath196 . in fact , we shall prove that the lie algebra generated by @xmath35 and @xmath31 is a subalgebra of @xmath3 , which is in general isomorphic to @xmath3 . to show that @xmath198 , we note that @xmath199 implies @xmath200 furthermore , @xmath77 and @xmath201 for @xmath202 leads to @xmath203 hence , @xmath204 and the energy levels @xmath80 of @xmath76 are @xmath205 , @xmath206 and @xmath207 for @xmath208 . thus , we have @xmath209 let @xmath85 be an isomorphism of the hilbert space of pure states defined by @xmath210 and set @xmath87 as well as @xmath211 for @xmath212 , @xmath213 , @xmath214 . then @xmath35 and @xmath31 have the following representations with respect to the new basis @xmath90 @xmath215 where @xmath92 and @xmath93 are as defined in ( [ eq : basisc ] ) and ( [ eq : yc ] ) , respectively , and we note that @xmath216 , @xmath217 . hence , the dynamical lie algebra generated by @xmath35 and @xmath31 must be a subalgebra of @xmath3 . since @xmath35 and @xmath31 in ( [ eq : sysc1 ] ) contain a complete set of generators @xmath92 and @xmath93 for @xmath3 , it is natural to expect that they generate the full lie algebra @xmath3 . we shall prove that this is true in most cases . however , as in case of @xmath5 , a proper subalgebra may also be generated . [ ex : c1 ] consider a system of type ( [ eq : sysc2 ] ) for @xmath94 . if @xmath95 , @xmath218 and @xmath219 , @xmath220 and @xmath221 then the basis change ( [ eq : sigmac ] ) leads to @xmath222 with @xmath92 and @xmath93 as defined in ( [ eq : basisc ] ) and ( [ eq : yc ] ) , respectively . therefore , the lie algebra @xmath34 generated by @xmath35 and @xmath31 is a subalgebra of @xmath223 . however , it is easy to verify that @xmath224 . indeed , @xmath34 is a three - dimensional subalgebra of @xmath223 spanned by @xmath30 , @xmath31 and @xmath225=-(\sqrt{5}x_1 + 2\sqrt{2}x_2 + 3x_3)$ ] . thus , for certain choices of the parameters @xmath104 and @xmath105 , the lie algebra @xmath34 is a _ proper _ subalgebra of @xmath3 . to find conditions that ensure @xmath62 , we consider the generic system @xmath226 with @xmath92 and @xmath93 as in ( [ eq : basisc ] ) and ( [ eq : yc ] ) . clearly , @xmath35 and @xmath31 are in @xmath3 . hence the lie algebra @xmath34 they generate must be contained in @xmath3 . [ thm : c1 ] let @xmath107 for @xmath108 and @xmath227 . the dynamical lie algebra @xmath34 generated by @xmath35 and @xmath31 as in ( [ eq : sysc ] ) is @xmath3 if @xmath228 for @xmath173 . using the properties of the generators @xmath92 and @xmath93 leads to : @xmath229,\rmi h_0 ' ] - \omega_1 ^ 2 ( \rmi h_1 ) \\ & = \sum_{m=2}^{\ell } \delta_m ( \omega_m^2-\omega_1 ^ 2 ) y_m \\ v^{(2 ) } & \equiv [ [ \rmi h_0 ' , v^{(1)}],\rmi h_0']-\omega_2 ^ 2 v^{(1 ) } \\ & = \sum_{m=3}^{\ell}\delta_m(\omega_m^2-\omega_1 ^ 2)(\omega_m^2-\omega_2 ^ 2 ) y_m\\ & \vdots \\ v^{(\ell-1 ) } & \equiv[[\rmi h_0',v^{(\ell-2)}],\rmi h_0']-\omega_{\ell-1}^2 v^{(\ell-2)}\\ & = \delta_\ell \prod_{m=1}^{\ell-1}(\omega_\ell^2-\omega_m^2 ) y_\ell . \end{aligned}\ ] ] by hypothesis @xmath228 for @xmath173 and @xmath230 . hence , all the factors in the last expression above are non - zero , i.e. , we have @xmath231 and thus @xmath62 by lemma [ lemma : c ] of [ app : c ] . if @xmath232 for some @xmath173 then a modification of the proof above leads to a residual term @xmath233 where @xmath234 and @xmath235 for @xmath122 and @xmath123 otherwise . if the energy levels @xmath236 are negative and ordered in an increasing ( non - decreasing ) sequence then @xmath232 implies @xmath237 for all @xmath128 . we shall only consider this case in the following . [ thm : c2 ] let @xmath238 for @xmath239 and @xmath240 , @xmath241 . the dynamical lie algebra @xmath34 generated by the system @xmath110 with @xmath35 and @xmath31 as in ( [ eq : sysc ] ) is @xmath3 if @xmath242 but @xmath243 for all @xmath244 . let @xmath245 $ ] and @xmath246 = \sum_{m=1}^\ell ( \tilde{\delta}_{m-1}^2-\tilde{\delta}_m^2 ) h_m.\ ] ] suppose @xmath247 has @xmath136 elements labeled @xmath137 , @xmath138 up to @xmath139 and let @xmath248 . then @xmath249-v_{m_1 } y^{(0 ) } } & = \sum_{k=2}^{\ell'+1 } \tilde{\delta}_{m_k } ( v_{m_k}-v_{m_1 } ) y_{m_k}\\ x^{(1)}&\equiv { [ y^{(0)},z]-v_{m_1 } x^{(0 ) } } & = \sum_{k=2}^{\ell'+1 } \tilde{\delta}_{m_k } ( v_{m_k}-v_{m_1 } ) x_{m_k}\\ y^{(2)}&\equiv { [ z , x^{(1)}]-v_{m_2 } y^{(1 ) } } & = \sum_{k=3}^{\ell'+1}\tilde{\delta}_{m_k}(v_{m_k}-v_{m_1})(v_{m_k}-v_{m_2})y_{m_k}\\ x^{(2)}&\equiv { [ y^{(1)},z]-v_{m_2 } x^{(1 ) } } & = \sum_{k=3}^{\ell'+1}\tilde{\delta}_{m_k}(v_{m_k}-v_{m_1})(v_{m_k}-v_{m_2})x_{m_k}\\ & \vdots\\ y^{(\ell ' ) } & \equiv { \prod_{k=1}^{\ell ' } \tilde{\delta}_\ell ( v_\ell - v_{m_k } ) y_\ell}\end{aligned}\ ] ] shows that @xmath231 and hence @xmath62 by lemma [ lemma : c ] of [ app : c ] . a similar argument shows that if there exists @xmath142 such that @xmath143 for @xmath250 but @xmath145 , then @xmath251 . using the fact that the generators @xmath93 of @xmath3 are not diagonal with respect to the cartan elements ( see [ app : c ] ) , it generally follows that @xmath34 contains all the generators @xmath147 and @xmath93 and thus @xmath252 as well . an important special case of this type is a system with @xmath4 equally spaced energy levels and uniform transition dipole moments : [ thm : c3 ] the dynamical lie algebra @xmath34 generated by a system @xmath110 with @xmath4 equally spaced energy levels and uniform dipole moments is @xmath3 . we have @xmath253 $ ] , @xmath254 $ ] and @xmath255 $ ] . next , set @xmath256 and @xmath257 for @xmath160 , with @xmath258 and @xmath259 , and note that @xmath260 $ ] , @xmath261 $ ] and @xmath262 $ ] . this shows that @xmath34 contains all the generators @xmath147 , @xmath93 of @xmath3 . hence , @xmath62 . let us now return to the original system ( [ eq : sysc2 ] ) . since we have assumed that the energy levels @xmath104 of the system are ordered in a non - decreasing sequence , i.e. , @xmath263 for all @xmath165 , we have @xmath164 , and hence @xmath264 i.e. , the energy levels @xmath167 are negative and form an increasing sequence . noting that @xmath265 we thus have @xmath266 for @xmath267 and @xmath268 . thus , if @xmath171 and @xmath172 for @xmath173 then @xmath62 according to theorem [ thm : c1 ] , independent of the choice of the dipole moments , provided that they are non - zero . if @xmath174 for some @xmath173 then theorem [ thm : c2 ] applies and the dipole moments determine whether we have @xmath62 or a proper subalgebra . in particular , the lie algebra is @xmath3 if the energy levels are equally spaced , @xmath269 for all @xmath165 , and the dipole moments @xmath45 , @xmath270 , are such that there exists @xmath195 , @xmath271 , so that @xmath143 for all @xmath145 , where @xmath272 ( with @xmath180 ) and @xmath273 . for instance , setting @xmath181 for @xmath112 works since it gives @xmath274 for @xmath108 but @xmath275 . similarly , setting @xmath184 for @xmath112 also gives @xmath62 due to theorem [ thm : c3 ] . however , recall that there are systems of type ( [ eq : sysc2 ] ) whose lie algebra @xmath34 is a proper subalgebra of @xmath3 , as example [ ex : c1 ] above clearly shows . note that in this example , the energy levels are equally spaced and the dipole moments @xmath105 are such that @xmath276 , @xmath277 and @xmath278 , i.e. , all the @xmath188 are equal and none of the theorems above are applicable . our analysis of finite - dimensional , non - decomposable driven quantum systems with nearest neighbour interactions and non - zero dipole moments shows that the dynamical lie algebra @xmath34 generated by the trace - zero part of the internal hamiltonian and the interaction hamiltonian of the system is either @xmath36 , @xmath1 , @xmath279 , or a simple subalgebra of these . although by far the most common case is @xmath36 , which corresponds to density matrix / observable controllability and usually complete controllability @xcite , certain symmetries of the controlled transitions can destroy or reduce the controllability of the system . precisely , we showed that the dynamical lie algebra @xmath34 of a system with symmetrically coupled transitions is a subalgebra of @xmath5 if the system has an odd number of energy levels ( where degenerate levels are to be counted according to multiplicity ) and a subalgebra of @xmath3 if the system has an even number of energy levels . moreover , we established criteria which guarantee in most cases that the dynamical lie algebra is actually isomorphic to either @xmath5 or @xmath3 . in particular , the dynamical lie algebra of a system with equally spaced energy levels and uniform transition dipole moments is @xmath1 if @xmath2 , and @xmath279 if @xmath4 . despite the rather technical nature of the results presented in this paper , we would like to emphasize that the identification of the dynamical lie algebra is a crucial first step towards identification of reachable and non - reachable target states for systems that are not essentially controllable . furthermore , knowledge about the structure of the dynamical lie algebra can be used to develop efficient control schemes for these systems . a standard basis representation for the lie algebra @xmath36 in terms of trace - zero , skew - hermitian @xmath14 matrices is ( see , for example @xcite ) @xmath280 where @xmath47 , @xmath281 and @xmath282 . there are @xmath283 generators @xmath92 and @xmath284 generators of type @xmath285 and @xmath286 each . hence , the total number of generators is @xmath287 and thus the dimension of the lie algebra @xmath36 is @xmath287 . a nice discussion of controllability of @xmath0-level quantum systems in terms of root space decompositions of @xmath36 can be found in @xcite . @xmath1 usually refers to the real lie algebra of trace - zero , anti - symmetric matrices @xcite . however , since we are dealing with subalgebras of @xmath36 generated by @xmath288 skew - hermitian matrices , we require a representation of @xmath1 in terms of trace - zero , skew - hermitian matrices . for @xmath2 , the standard representation of the complex lie algebra @xmath289 @xcite leads to the following skew - hermitian basis for the real lie algebra @xmath5 : + @xmath290 where @xmath112 and @xmath291 . since there are @xmath292 elements @xmath92 , @xmath293 and @xmath294 each , as well as @xmath295 elements @xmath296 , @xmath297 , @xmath298 and @xmath299 each , the total number of basis elements is @xmath300 . thus , the dimension of @xmath5 is @xmath300 . using the general commutation relations @xmath301 = x_{\eps_n } , } & { [ x_{\eps_m } , y_{\eps_m-\eps_n } ] = y_{\eps_n } , } \\ { [ x_{\eps_m } , x_{\eps_n } ] = x_{\eps_m-\eps_n}-x_{\eps_m+\eps_n } , } & { [ x_{\eps_m } , y_{\eps_n } ] = y_{\eps_m-\eps_n}+y_{\eps_m+\eps_n } , } \\ { [ x_{\eps_m } , y_{\eps_m } ] = -2 h_m , } & { [ x_{\eps_m\pm\eps_n } , y_{\eps_m\pm\eps_n } ] = -2(h_m \pm h_n ) , } \\ { [ h_m , x_{\eps_m \pm \eps_n } ] = -y_{\eps_m \pm \eps_n } , } & { [ h_m , y_{\eps_m \pm \eps_n } ] = x_{\eps_m \pm \eps_n } } \end{array}\ ] ] for @xmath302 , shows that the elements @xmath147 and @xmath93 with @xmath303 are not diagonal with respect to the cartan elements @xmath92 of the lie algebra and generate the full lie algebra @xmath5 . furthermore , it generally suffices to prove that the lie algebra @xmath34 generated by @xmath35 and @xmath31 as in ( [ eq : sysb ] ) contains one of these elements to conclude that @xmath61 . we shall demonstrate this explicitly for the case @xmath304 . [ lemma : b ] let @xmath34 be the lie algebra generated by @xmath35 and @xmath31 as defined in ( [ eq : sysb ] ) . if @xmath141 then @xmath305 for @xmath112 and hence @xmath61 . using ( [ eq : rulesb ] ) shows that @xmath306 implies @xmath307=\eps_1 x_1 $ ] and @xmath308=2h_1 $ ] ; thus @xmath309 . furthermore , we have @xmath310 + \eps_1\delta_1 x_1 & = \displaystyle \sum_{m=2}^\ell ( \eps_m-\eps_{m-1 } ) \delta_m x_m \\ & [ z^{(1)},y^{(1 ) } ] & = \displaystyle -\eps_2\delta_2 x_2 - \sum_{m=3}^\ell ( \eps_m-\eps_{m-1 } ) \delta_m x_m \end{array}\ ] ] which shows that @xmath311=-\eps_1\delta_2 x_2 $ ] , i.e. , @xmath312 , and @xmath313=\eps_2 y_2 $ ] , @xmath314 = 2(h_2-h_1)$ ] implies @xmath315 . in general , defining recursively @xmath316 shows that @xmath317=- \eps_k \delta_{k+1 } x_{k+1}$ ] , @xmath318 = \eps_{k+1}y_{k+1}$ ] and @xmath319=2(h_{k+1}-h_k)$ ] . thus , @xmath320 , @xmath321 and @xmath322 are in @xmath34 for @xmath323 . a basis representation for the lie algebra @xmath3 for @xmath4 in terms of trace - zero , skew - hermitian @xmath14 matrices can be derived from the standard basis for @xmath324 @xcite : @xmath325 where @xmath112 and @xmath291 . since there are @xmath292 elements @xmath92 , @xmath326 and @xmath327 each , as well as @xmath295 elements @xmath296 , @xmath297 , @xmath298 and @xmath299 each , the total number of basis elements is @xmath300 and the dimension of @xmath3 is thus @xmath300 . using the general commutation relations @xmath328 = x_{\eps_m+\eps_n } , } & { [ x_{2\eps_n } , y_{\eps_m-\eps_n } ] = y_{\eps_m+\eps_n } , } \\ { [ x_{\eps_m+\eps_n } , x_{\eps_m-\eps_n } ] = 2(x_{2\eps_m}-x_{2\eps_n } ) , } & { [ x_{\eps_m+\eps_n } , y_{\eps_m-\eps_n } ] = 2(y_{2\eps_m}+y_{2\eps_n } ) , } \\ { [ x_{2\eps_m } , y_{2\eps_m } ] = -2 h_m , } & { [ x_{\eps_m\pm\eps_n } , y_{\eps_m\pm\eps_n } ] = -2(h_m \pm h_n ) , } \\ { [ h_m , x_{\eps_m\pm \eps_n } ] = -y_{\eps_m\pm \eps_n } , } & { [ h_m , y_{\eps_m\pm \eps_n } ] = x_{\eps_m\pm \eps_n } } \end{array}\ ] ] for @xmath302 , shows that the elements @xmath147 and @xmath93 with @xmath329 are not diagonal with respect to the cartan elements @xmath92 of the lie algebra and generate the full lie algebra @xmath3 . again , it therefore generally suffices to prove that the lie algebra @xmath34 generated by @xmath35 and @xmath31 as in ( [ eq : sysc ] ) contains one of these elements to conclude that @xmath62 . we demonstrate this explicitly for the case @xmath330 . [ lemma : c ] let @xmath34 be the lie algebra generated by @xmath35 and @xmath31 defined in ( [ eq : sysc ] ) . if @xmath231 then @xmath331 for @xmath270 , hence @xmath62 . using ( [ eq : rulesc ] ) shows that @xmath157 implies @xmath332 = 2 \eps_\ell x_\ell$ ] and @xmath333 = 2 h_\ell$ ] ; thus @xmath334 . furthermore , we have @xmath335 + 2\eps_\ell\delta_\ell x_\ell & = \displaystyle \sum_{m=1}^{\ell-1 } ( \eps_{m+1}-\eps_m ) \delta_m x_m \\ & [ z^{(1)},y^{(1 ) } ] & = \displaystyle \eps_{\ell-1}\delta_{\ell-1 } x_{\ell-1 } - \sum_{m=1}^{\ell-2 } ( \eps_{m+1}-\eps_m ) \delta_m x_m \end{array}\ ] ] which shows that @xmath311=\eps_\ell\delta_{\ell-1 } x_{\ell-1}$ ] , @xmath336=-\eps_{\ell-1 } y_{\ell-1}$ ] , and @xmath337 = 2(h_\ell - h_{\ell-1})$ ] ; thus @xmath338 . in general , defining recursively @xmath339 shows that @xmath340=\eps_{\ell - k } \delta_{\ell - k-1 } x_{\ell - k-1}$ ] , @xmath341=-\eps_{\ell - k-1 } y_{\ell - k-1}$ ] and @xmath342 = 2(h_{\ell - k}-h_{\ell - k-1})$ ] . thus , we have indeed @xmath343 , @xmath344 and @xmath345 in @xmath34 for @xmath346 . using the standard representation for the complex lie algebra @xmath347 @xcite , we can derive the following skew - hermitian basis for @xmath63 : @xmath348 where @xmath112 and @xmath349 . there are @xmath292 elements @xmath92 , as well as @xmath295 elements @xmath296 , @xmath297 , @xmath298 and @xmath299 each , i.e. , the total number of basis elements is @xmath350 . thus , the dimension of @xmath63 is @xmath350 . to see why there is no @xmath351-level system with @xmath65 , where @xmath352 and @xmath353 are as defined in ( [ eq : hzero ] ) and ( [ eq : hone ] ) , respectively , such that @xmath354 , note that @xmath355 forms a minimal , complete set of generators for @xmath63 if @xmath356 . each of the @xmath292 generators @xmath93 has four distinct , non - zero entries , which corresponds to a total of @xmath357 non - zero entries . however , @xmath31 for a @xmath351-level system with only nearest neighbour interactions can have at most @xmath358 non - zero entries on the first super- and sub - diagonal . hence , a @xmath351-level system with dynamical lie algebra @xmath63 must have interactions between non - adjacent energy levels .	the problem of identifying the dynamical lie algebras of finite - level quantum systems subject to external control is considered , with special emphasis on systems that are not completely controllable . in particular , it is shown that the dynamical lie algebra for an @xmath0-level system with symmetrically coupled transitions , such as a system with equally spaced energy levels and uniform transition dipole moments , is a subalgebra of @xmath1 if @xmath2 , and a subalgebra of @xmath3 if @xmath4 . general criteria for obtaining either @xmath5 or @xmath3 are established .
to prove our main theorem , we need to maximize the bell expression @xmath93,\,\boldsymbol{b}\in[\boldsymbol{o}],\,x\in[m],\,\boldsymbol{y}\in[\boldsymbol{m } ] } \beta_{a,\boldsymbol{b},x,\boldsymbol{y } } \ , \operatorname{tr}\left[m^{(a)}_x\,\sigma_{\boldsymbol{b}\|\boldsymbol{y}}\right ] \label{eq : bell_ineqexcltriv}\ ] ] over all sets @xmath94}$ ] of generic dichotomic povm measurements @xmath95 by alice . however , it is known that , to test for bell nonlocality of quantum states , it suffices to examine only von neumann ( rank-1 projective ) measurements @xcite . the same happens for bell violations by assemblages , which we formalize with the following lemma . [ lemma : excludingpovms ] let @xmath32 be an arbitrary well - behaved bell inequality with dichotomic outputs for alice and @xmath44 a generic qubit assemblage . then , the maximal violation of @xmath32 by @xmath44 is attained under rank-1 projective measurements . _ proof_. our proof strategy consists in showing that , for an arbitrary assemblage @xmath44 and generic dichotomic povm measurements @xmath96 , the distribution @xmath97 , given by eq . , obtained from @xmath44 under @xmath96 , is equivalent to a distribution @xmath98 obtained from @xmath44 under a set @xmath99 of von neumann measurements followed by a local mixing of alice s outputs . since @xmath32 is well behaved , this implies that if @xmath97 violates @xmath32 , i.e. if @xmath53 , then @xmath100 , which implies that the maximal violation is always attained under von neumann measurements . since , for all @xmath101 $ ] , the povm measurement operators @xmath102 and @xmath103 are both non - negative , they can be diagonalized as @xmath104 where @xmath105 and @xmath106 are rank-1 orthonormal projectors i.e. @xmath107 , being @xmath108 the kronecker delta acting on @xmath19 , and where @xmath109 for @xmath110 . hence , @xmath111}$ ] defines a set of von neumann measurements . substituting eq . into eq . ( [ eq : pofsigma ] ) , we find that @xmath112 \ , \end{aligned}\ ] ] where we have introduced @xmath113 now , defining the distribution @xmath98 such that @xmath114 $ ] , we can write @xmath115 } q(a\|a',x ) \ \tilde{p}_{\xi}(a',\boldsymbol{b}\|x,\boldsymbol{y } ) \ . \label{eq : povmaslocmix}\ ] ] this , as evident from eq . ( [ eq : deflocmix ] ) , is the expression of a local mixing of alice s outputs applied to the von neumann measurement distribution @xmath116 . we can now continue with the proof of criterion [ crit : conds_for_bell_viol ] . due to lemma [ lemma : excludingpovms ] , we need to maximize the bell expression only over the set of von neumann measurements , i.e. , with measurement operators @xmath47 of the form @xmath117 where each ( unit ) vector @xmath118 represents a direction on the bloch sphere . recall that @xmath119 is the pauli - operator vector with respect to a fixed basis of @xmath19 of one s preference , so that the vectors @xmath120 are the only variables of the optimization . using eqs . and , and the fact that the vectors @xmath121}$ ] are all independent , we get @xmath122 } \left\{\sum_{a\in[2],\,\boldsymbol{b}\in[\boldsymbol{o}],\,\boldsymbol{y}\in[\boldsymbol{m}]}\frac12\beta_{a,\boldsymbol{b},x,\boldsymbol{y } } \ , p(\boldsymbol{b}\|\boldsymbol{y } ) + \max_{\{\hat{\boldsymbol{s}}_x\}}\operatorname{tr}\left [ b_x\ , \hat{\boldsymbol{s}}_x \cdot \boldsymbol{\sigma}\right]\right\}\ . \label{eq : recastexpr}\ ] ] where , for each @xmath101 $ ] , we have introduced the hermitean operator on @xmath19 @xmath123,\,\boldsymbol{y}\in[\boldsymbol{m } ] } \frac12(\beta_{0,\boldsymbol{b},x,\boldsymbol{y}}-\beta_{1,\boldsymbol{b},x,\boldsymbol{y}})\ , \sigma_{\boldsymbol{b}\|\boldsymbol{y } } \ . \label{eq : bobops}\ ] ] note that @xmath124 coincides with the expression inside the brackets of eq . ( [ eq : main_result_r ] ) . then , using that @xmath125 = \boldsymbol r(b_x)\cdot\hat{\boldsymbol{s}}_x$ ] , with the vector function @xmath126 defined in eq . ( [ bldef ] ) , the maximization is finally reduced to @xmath127= \max_{\{\hat{\boldsymbol{s}}_x\}}\ \boldsymbol{r}(b_x)\cdot { \hat{\boldsymbol{s}}}_x \ . \label{eq : reducedprob}\ ] ] clearly , the maximum is @xmath127=\\|{\boldsymbol s}^{\rm opt}_x\\| \ , \label{bestexpr}\ ] ] with @xmath128 for all @xmath101 $ ] , attained by @xmath129 substituting eq . into eq . , one obtains the optimal measurement settings of eq . ( [ eq : opt_meas ] ) . using , in turn , eqs . and , one sees that eq . is equivalent to the left - hand side of eq.([eq : main_result ] ) . in the 2-input , 2-output scenario , by virtue of fine s theorem @xcite , bell nonlocality is equivalent to the violation of the chsh inequality , given by @xmath130 or any of its 8 symmetries ( defined by swapping around the minus sign with the other terms , by applying an overall sign change , or by doing both ) . so , it suffices to show that a violation of eq . ( [ eq:222 ] ) is equivalent to the violation of any of the 8 symmetries of the chsh inequality . in the notation of eq . ( [ eq : bell_ineq ] ) , and omitting the subindices from @xmath74 and @xmath75 , the chsh inequality is expressed as @xmath131 with @xmath132 . its symmetries , in turn , are obtained by replacing @xmath28 or @xmath133 by their negations modulo 2 , by applying an overall sign change to @xmath134 , and by applying any composition of the three . substituting eq . in eqs . ( [ eq : main_result ] ) and ( [ eq : main_result_r ] ) leads to eqs . ( [ eq:222 ] ) and ( [ eq:222_r ] ) , as the reader can straightforwardly verify . this shows that the violation of eq . ( [ eq:222 ] ) , with the measurement direction @xmath135 given by eq . ( [ eq:222_r ] ) , is equivalent to the violation of the chsh inequality . now , note that any of the other symmetries mentioned above either does not explicitly introduce any change in eqs . ( [ eq:222 ] ) and ( [ eq:222_r ] ) or simply corresponds to the relabelings @xmath136 , @xmath137 , for all @xmath138 $ ] , or compositions of the two . none of the latter alters the statements of criterion [ crit : twotwo ] . that is , the violation of eq . ( [ eq:222 ] ) , with @xmath135 given by eq . ( [ eq:222_r ] ) , is equivalent to the violation of any of the symmetries of the chsh inequality , which finishes the proof . for any assemblage @xmath139}$ ] , the correlator @xmath140 , where @xmath141 is a @xmath85-valued unknown observable of bob s subsystem and @xmath142 is a pauli operator on @xmath56 , is given by @xmath143 } ( -1)^{b } \operatorname{tr}\left ( \sigma_{b\|y } \ , a_{\alpha } \right)\\ & = \boldsymbol r \left(\sum_{b\in[2 ] } ( -1)^{b } \ , \sigma_{b\|y } \right)\cdot \hat{\boldsymbol v}_{\alpha } \ , \label{eq : corrbasic2}\end{aligned}\ ] ] where @xmath144 is a unit vector in the bloch sphere in the direction of @xmath145 . using eqs . , ( [ eq:222_r ] ) and the linearity of the vector function @xmath126 , one sees that @xmath146 and , analogously , @xmath147 it is now straightforward to see , from the definition of the euclidian norm , that eq . ( [ eq:222 ] ) is equivalent to eq . ( [ eq:222meas ] ) for @xmath148 the pauli operators in any orthonormal basis of @xmath19 . furthermore , one can also see that the lhs of inequality ( [ eq : australians ] ) is equivalent to the lhs of inequality ( [ eq:222 ] ) evaluated at the projections of @xmath89 and @xmath88 onto the plane orthogonal to @xmath149 , instead of at @xmath89 and @xmath88 themselves .	we study multipartite bell nonlocality in a framework native of multipartite einstein - podolsky - rosen ( epr ) steering scenarios with a single trusted measurement device . we derive a closed - form necessary and sufficient criterion for systems composed of a qubit and @xmath0 untrusted black - box measurement devices to violate under general dichotomic measurements on the qubit a generic bell inequality from a broad family of linear inequalities with arbitrarily many outputs for the @xmath0 untrusted devices and inputs for all @xmath1 parties . the optimal quantum measurements for maximal violation are also obtained . for two users , and two inputs and two outputs per user , our criterion becomes necessary and sufficient for bell nonlocality . furthermore , in that setting , its form generalizes recently obtained steering inequalities , which allows us to provide useful feedback from nonlocality to the detection of steering . our findings constitute a practical tool for the study of the interplay between epr steering and bell nonlocality , with potential applications in multipartite information processing . _ introduction_. composite quantum systems can display exotic forms of non - classical correlations , a phenomenon known under the generic name of _ quantum nonlocality_. quantum nonlocal correlations can appear in three main variants . the first one is _ entanglement _ , which refers to inseparability of quantum states ( described by density matrices ) @xcite . the second one is _ bell nonlocality _ @xcite , the impossibility of explaining measurement statistics ( described by joint probability distributions ) with local hidden - variable ( lhv ) models @xcite . the third variant is called _ einstein - podolsky - rosen ( epr ) steering _ @xcite , after the famous epr paper @xcite . this is an effect by which ensembles of quantum states are remotely prepared by local measurements at distant labs @xcite . the observable data in steering experiments thus consist of the measurement statistics at one lab and quantum states at another . these data are compactly described by a joint mathematical object called _ assemblage _ , composed of a conditional probability distribution and a set of density matrices . hence , epr steering constitutes an intermediate notion between entanglement and bell nonlocality @xcite . apart from their fundamental importance , nonlocal correlations are of practical relevance : they are resources for physical tasks such as quantum key - distribution ( qkd ) @xcite and quantum random - number generation @xcite . entanglement is a resource @xcite in the device - dependent ( dd ) scenario , defined by well - characterized , trusted quantum measurements . bell nonlocality is useful for device - independent ( di ) protocols @xcite , i.e. , where the experimenters possess untrusted black - box measurement apparatuses . in turn , epr steering has been identified @xcite as a resource @xcite for one - sided di situations , where one of the parties has an untrusted black - box device while the other possesses a trusted quantum platform . fully ( both - sided ) di protocols relax the need for device characterization totally , but at the expense of being very demanding experimentally @xcite . one - sided di implementations offer a middle - path alternative , relaxing device characterization only on one side but having , in return , less stringent experimental requirements @xcite for security @xcite than in fully di ones . this is relevant to any asymmetric situation involving users with different levels of quantum control . these developments motivated a great amount of work on the interplay between the different forms of quantum nonlocality . the first problem tackled was that of entangled versus bell nonlocal states ( those capable of exhibiting bell nonlocality ) . all pure entangled states were proven to be bell nonlocal @xcite , but mixed bell local entangled states were found @xcite . later , necessary and sufficient conditions for arbitrary 2-qubit states to be bell nonlocal were derived @xcite . a long list of works then followed these pioneering results ( see , for instance , sec . iii.a of @xcite and refs . therein ) . the second problem was that of entangled versus steerable states ( those capable of exhibiting epr steering ) . unsteerable entangled states , as well as bell local steerable ones , were found @xcite . this led to an impressive amount of work : the sets of entangled , steerable and bell nonlocal states were proven inequivalent under general measurements @xcite ; necessary criteria for a two - qubit state to be steerable were found @xcite ; and constructive methods to test for unsteerability of a state were developed @xcite . finally , in addition to the many known steering inequalities @xcite , a necessary and sufficient criterion for epr steering has been recently obtained for minimal - dimension assemblages @xcite . here we consider a third problem : steerable versus bell nonlocal multipartite assemblages . we derive a closed - form necessary and sufficient criterion for an @xmath1-partite assemblage with a single trusted device , in possession of a user called alice , to violate , under general measurements [ positive operator - valued measure ( povm ) ] by her , a bell inequality . the optimal measurements for the maximal violation are also given . our theorems assume that alice s measurements are dichotomic and that her system is a qubit , but are otherwise general in the number of outcomes for the untrusted devices and of settings for all parties . furthermore , we make only minimal assumptions on the bell inequalities treated , namely , that they are linear and that their violations do not increase under probabilistic local mixings of alice s outputs . thus , many of the most popular bipartite @xcite and multipartite @xcite bell inequalities are within the range of applicability of our criterion . in addition , for @xmath2 users with 2 inputs and 2 outputs per user , our criterion unambiguously characterizes all bell nonlocal assemblages . interestingly , in that setting , our criterion generalizes , in form , recently obtained steering inequalities @xcite . by virtue of this , we provide insight into the detection of epr steering within the framework of bell nonlocality and explain formal links between the two problems . finally , we suggest potential connections of our findings with information - theoretic protocols with asymmetric levels of quantum control among the users involved . _ preliminaries_. we consider @xmath1 space - like separated parties , alice , in possession of a trusted measurement device , and @xmath0 users , @xmath3 , @xmath4 , @xmath5 , in possession of untrusted devices @xcite . this is the @xmath6-sided di scenario . each @xmath7-th untrusted device , for @xmath8 , is treated as a black box with unknown internal functioning , which , given an input @xmath9 $ ] , outputs @xmath10 $ ] , where @xmath11 and the notation @xmath12\coloneqq\{0,\ldots , n-1\}$ ] , for any @xmath13 , is introduced . in addition , we will also use the short - hand notation @xmath14\coloneqq[n]^{n-1}$ ] . alice s subsystem , in turn , is a qubit , on which she can perform any quantum measurement of her choice . the joint system state is specified by an @xmath6-partite conditional probability distribution @xmath15 of the output string @xmath16 given the input string @xmath17 , associated to a ( normalized ) conditional single - partite quantum state @xmath18 on alice s subsystem s hilbert space @xmath19 . these can be conveniently encapsulated in the _ assemblage _ @xmath20,\ , \boldsymbol{y}\in[\boldsymbol{m}]},\ ] ] of ( subnormalized ) conditional quantum states @xmath21 on @xmath19 , with @xmath22 . in other words , @xmath23 provides a concise description of all the observable information in @xmath6-sided di experiments . on the other hand , in the fully di scenario of all @xmath1 users possessing black - box devices , the joint system behavior is described by an @xmath1-partite conditional distribution @xmath24,\ , \boldsymbol{b}\in[\boldsymbol{o } ] , \ , x\in[m],\ , \boldsymbol{y}\in[\boldsymbol{m}]},\ ] ] where @xmath25 is the probability of output values @xmath26 and @xmath27 conditioned on input values @xmath28 and @xmath29 . for ease of notation , we assume throughout that the numbers of inputs and outputs , @xmath30 and @xmath31 , respectively , are the same for all @xmath1 users , but all our results are also valid otherwise . bell inequalities offer a practical tool to test for bell nonlocality in a given distribution @xcite . every linear bell inequality is represented by a pair @xmath32 , with @xmath33,\ , \boldsymbol{b}\in[\boldsymbol{o } ] , \ , x\in[m],\ , \boldsymbol{y}\in[\boldsymbol{m}]}$ ] and @xmath34 , such that @xmath35,\ , \boldsymbol{b}\in[\boldsymbol{o}]\\ x\in[m],\ , \boldsymbol{y}\in[\boldsymbol{m } ] } } \beta_{a,\boldsymbol{b},x,\boldsymbol{y } } \ , p(a,\boldsymbol{b}\|x,\boldsymbol{y } ) \leq \beta_{\rm l}\ ] ] for all bell local @xmath36 . furthermore , in the multipartite scenario , bell inequalities can also be tailored so as to test for different forms of multipartite bell non locality @xcite . our criterion below holds for all linear bell inequalities whose violations do not increase under local probabilistic mixings of alice s outputs , to which we refer , for short , as _ well - behaved bell inequalities_. more precisely , local mixings map @xmath36 into a distribution @xmath37 with elements @xmath38 } q(a\|a',x ) \ p(a',\boldsymbol{b}\|x,\boldsymbol{y } ) \ , \label{eq : deflocmix}\ ] ] where @xmath39 , with @xmath40 } q(a\|a',x)=1 $ ] , characterizes the mixing probability for each input @xmath28 and output @xmath41 . hence , @xmath32 is well behaved if @xmath42 for all @xmath36 for which @xmath43 . local mixings can map local distributions only into local distributions . so , that a bell violation does not increase under such mixings is a basic reasonable property typically satisfied by known inequalities ( including all tight ones and , more generally , all those for which a constant local weight @xcite implies a constant violation ) . examples not satisfying this property can be found among reducible inequalities that have superfluous terms @xcite . finally , we say that @xmath44 _ violates a bell inequality _ @xmath32 if there exists a set @xmath45}$ ] of measurements @xmath46}$ ] , with non - negative measurement operators @xmath47 on @xmath19 fulfilling @xmath48 } m^{(a)}_x=\mathbb1 $ ] for all @xmath49 $ ] , @xmath50 being the identity operator on @xmath19 , such that the distribution @xmath51,\ , \boldsymbol{b}\in[\boldsymbol{o}],\ , x\in[m],\ , \boldsymbol{y}\in[\boldsymbol{m}]}$ ] , defined by @xmath52\ \forall\ , \substack{a\in[o],\ , \boldsymbol{b}\in[\boldsymbol{o}]\\ x\in[m],\ , \boldsymbol{y}\in[\boldsymbol{m } ] } , \end{aligned}\ ] ] violates @xmath32 , i.e. if @xmath53 . our definition of bell violation by an assemblage considers only non - sequential measurements on a single copy of the assemblage . for quantum states , it is known that measurements on multiple copies of the state @xcite , or sequential measurements ( filterings ) @xcite on a single copy , can produce bell violations by entangled states that would otherwise yield local correlations . in fact , it has even been suggested @xcite that every entangled state might be bell nonlocal in this broader sense . something similar is expected to happen with steerable assemblages . however , the conditions for bell violations of assemblages in more general measurement scenarios are outside the scope of the present contribution . _ conditions for bell violations_. before our first theorem , we need to introduce some notation . since @xmath54 , any hermitian operator @xmath55 on @xmath56 can be decomposed in a bloch - sphere - like form @xmath57 $ ] , where @xmath58\in\mathbb{r}$ ] , @xmath59 is the pauli - operator vector , and @xmath60,\operatorname{tr}[o\,y],\operatorname{tr}[o\,z]\right)\in\mathbb{r}^3.\ ] ] if @xmath55 is a state , @xmath61 represents its bloch vector . finally , for any @xmath62 , we denote its euclidean norm by @xmath63 . we can now present our main result , which we prove in the appendix [ app : main ] : [ crit : conds_for_bell_viol ] let @xmath44 be a generic assemblage given by eq . and let @xmath32 be a well - behaved bell inequality with dichotomic measurements for alice . then , @xmath44 violates @xmath32 if , and only if , @xmath64}\left[\sum _ { \substack{a\in[2],\ , \boldsymbol{b}\in[\boldsymbol{o}]\\ \boldsymbol{y}\in[\boldsymbol{m } ] } } \frac12\beta_{a,\boldsymbol{b},x,\boldsymbol{y } } \ , p(\boldsymbol{b}\|\boldsymbol{y } ) + \left\\|{\boldsymbol s}^\mathrm{opt}_x\right\\|\right ] > \beta_l , \label{eq : main_result}\end{aligned}\ ] ] where @xmath65 , \ , \boldsymbol{y}\in\boldsymbol{[m ] } } \frac12\left(\beta_{0,\boldsymbol{b},x,\boldsymbol{y}}-\beta_{1,\boldsymbol{b},x,\boldsymbol{y}}\right)\sigma_{\boldsymbol{b}\|\boldsymbol{y}}\right ) \label{eq : main_result_r}\end{aligned}\ ] ] furthermore , the maximal violation is given by von - neumann measurements along the bloch - sphere directions @xmath66 , i.e. , with @xmath67 eqs . , , and are the solutions to optimizations over general ( povm ) dichotomic measurements . the criterion applies to many of the most - widely used bell inequalities . in the bipartite case , these include , for instance , the chsh @xcite and chained @xcite inequalities , as well as the @xmath68 inequality ( together with its variants for more outputs for bob or more inputs for both ) @xcite . in the multipartite scenario , in turn , criterion [ crit : conds_for_bell_viol ] covers important multipartite @xcite and genuinely multipartite @xcite bell inequalities , as well as @xmath6-sided di genuine @xmath1-partite entanglement witnesses @xcite . for instance , svetlichny s inequality is obtained by taking @xmath69 and @xmath70 otherwise @xcite ; whereas the svetlichny - like chained inequality introduced in ref . @xcite is given by @xmath71_2}$ ] , where @xmath72_2 $ ] stands for @xmath26 modulo @xmath73 . violation of the former implies genuinely multipartite nonlocality ( gmn ) , while violation of the latter implies a strong form of gmn that , for large @xmath30 , is a resource for di quantum secret - sharing protocols against generic nonsignaling ( even post - quantum ) eavesdroppers @xcite . quantum secret sharing ( qss ) is an intrinsically multipartite cryptographic protocol with remarkable security properties @xcite . interestingly , unconditional security of qss has been recently proven in the @xmath6-sided di scenario we consider but in the continuous - variable regime @xcite . _ conditions for 2-input 2-output bipartite bell nonlocality_. as a crucial application of criterion [ crit : conds_for_bell_viol ] , we focus on the case of two parties , each one with dichotomic inputs and outputs . in this scenario , bell nonlocality is equivalent to a chsh violation @xcite , thus , applying criterion [ crit : conds_for_bell_viol ] to the chsh inequality , we automatically get a necessary and sufficient condition for nonlocality . this is formalized by the following corollary , whose proof we leave for the appendix [ app:2input2output ] . for ease of notation , from now on we omit the subindex 1 from the untrusted party s input @xmath74 and output @xmath75 . [ crit : twotwo ] let @xmath44 be a generic assemblage with @xmath76 inputs and @xmath77 outputs per party and @xmath2 . then , @xmath44 is bell local if , and only if , @xmath78 } \left\\|{\boldsymbol t}^\mathrm{opt}_x\right\\| \leq 2 \ , \end{aligned}\ ] ] where @xmath79}(-1)^{b+x\,y}\ , \sigma_{b\|y}\right ) . \label{eq:222_r}\end{aligned}\ ] ] furthermore , if inequality is violated , the maximal violation is given by von neumann measurements along the bloch - sphere directions @xmath80 . _ connection to steering inequalities_. in ref . @xcite a bipartite steering inequality for correlators was derived . there , it was shown that , if @xmath81 and @xmath82 are any two out of the three pauli operators in an arbitrary basis of @xmath56 and @xmath44 is unsteerable , then @xmath83}\langle a_x\,(b_0 + b_1)\rangle^2 } + \sqrt{\sum_{x\in[2]}\langle a_x\,(b_0 - b_1)\rangle^2 } \leq 2 \ .\end{aligned}\ ] ] @xmath84 and @xmath3 are unknown , @xmath85-valued observables of the untrusted part ; and @xmath86}(-1)^a\operatorname{tr}[\sigma_{b\|y}\,a_x]$ ] for all @xmath87 $ ] . later on , in ref . @xcite , the authors found out that a violation of eq . implies not only that @xmath44 is steerable but also that it violates the chsh inequality ( under some measurements for alice not necessarily corresponding to @xmath81 and @xmath82 ) . and @xmath88 onto some fixed chosen plane ( represented in red ) instead of at @xmath89 and @xmath88 themselves . for assemblages for which the plane ( represented in light blue ) of @xmath89 and @xmath88 happens to coincide with the chosen one , are equivalent . for any other assemblage , inequality is more effective than inequality . ] we can explain this implication in light of criterion [ crit : twotwo ] . to this end , we note ( see appendix [ app : steerineq ] ) that eq . can be recast as @xmath90}\langle a_x\,(b_0 + b_1)\rangle^2 } + \sqrt{\sum_{x\in[3]}\langle a_x\,(b_0 - b_1)\rangle^2 } \leq 2 \ , \label{eq:222meas}\end{aligned}\ ] ] where @xmath91 is the third pauli operator complementary to @xmath81 and @xmath82 . clearly , the left - hand side ( lhs ) of eq . is greater or equal than that of eq . . as a consequence , a violation of eq . implies a violation and , therefore , of , consistent with the findings of ref . @xcite . the difference , of course , is that a violation of eq . does not in general imply a violation of eq . . so , while the former gives a necessary and sufficient condition for 2-input 2-output bipartite bell nonlocality , the latter only provides a sufficient one . interestingly , one can show ( see appendix [ app : steerineq ] ) that , in the basis of @xmath56 in which the bloch - sphere directions associated to @xmath81 and @xmath82 are contained in the plane of @xmath89 and @xmath88 , the correlators involving @xmath91 in eq . vanish and the lhss of eq . and . thus coincide ( see fig . [ compaust ] ) . this implies that for every 2-input 2-output bipartite bell nonlocal assemblage there exists a pair of mutually unbiased bases for which the steering in the assemblage is witnessed via a violation of . the implication was also recently formally proven in refs . @xcite by other reasonings . the pair of bases is readily obtained from eq . . _ concluding remarks_. we have unambiguously characterized all the @xmath1-partite assemblages with a single trusted ( qubit ) system that can violate under general dichotomic trusted measurements on the qubit bell inequalities from a very broad family . indeed , most widely used inequalities are within the firepower of our criterion . in addition , the optimal povms for maximal violation were also provided and turn out to always be rank-1 projective von neumman measurements . furthermore , for the important particular case of two users with 2 inputs and 2 outputs per user , our criterion unambiguously characterizes all bell nonlocal correlations in the assemblage . in that setting , we showed that the form of the criterion generalizes that of recently obtained steering inequalities , and sheds light back onto the problem of steering detection within the framework of bell nonlocality . our results hold for the usual scenario of non - sequential measurements on a single copy of the assemblage . hence , our criterion is to qubit assemblages what the famous analytical criterion of ref . @xcite for violation of the chsh inequality @xcite is to 2-qubit states , with the difference that we handle more generic bell inequalities and in the multipartite case . we leave as an open question the conditions for bell violations and bell nonlocality in more general measurement scenarios @xcite . it is important to remark that the optimizations we have solved can also be solved with semi - definite programming ( sdp ) @xcite . however , while sdp can , in general , only give , for each problem instance , a numeric solution , we provided closed - form analytic expressions for the general case . analytic solutions both carry more information and are more practical than numeric ones . this is specially relevant in , e.g. , security proofs , which are naturally formulated symbolically and rarely admit numeric manipulations . in turn , here , it was precisely having closed - form solutions what made the conceptual connections with steering inequalities possible . we have developed a practical toolbox to study of the interplay between epr steering and bell nonlocality , with potential implications in multipartite cryptographic protocols such as quantum secret sharing @xcite . a further interesting prospect would be to explore possible connections with random - number generation schemes that are one - sided di in that they treat photon sources as black boxes @xcite . _ we thank s. p. walborn for his support . we acknowledge financial support from the brazilian agencies cnpq ( national council for scientific and technological development ) , faperj ( research support foundation of the state of rio de janeiro ) , capes ( coordination for the improvement of higher education personnel ) , and inct - iq ( national institute for science and technology of quantum information ) . 99 r. horodecki , p. horodecki , m. horodecki , and k. horodecki , _ quantum entanglement _ , rev . mod . phys . * 81 * , 865 ( 2009 ) . n. brunner , d. cavalcanti , s. pironio , v. scarani and s. wehner , _ bell nonlocality _ , rev . mod . phys . * 86 * , 419 ( 2014 ) . j. s. bell , _ on the einstein podolsky rosen paradox _ , physics * 1 * 195 ( 1964 ) . m. d. reid , p. d. drummond , w. p. bowen , e. g. cavalcanti , p. k. lam , h. a. bachor , u. l. andersen , and g. leuchs , _ _ colloquium : _ the einstein - podolsky - rosen paradox : from concepts to applications _ , rev . mod . phys . * 81 * , 1727 ( 2009 ) . a. einstein , b. podolsky , and n. rosen , _ can quantum - mechanical description of physical reality be considered complete ? _ , phys . rev . * 47 * 777 ( 1935 ) . h. m. wiseman , s. j. jones and a. c. doherty , _ steering , entanglement , nonlocality , and the einstein - podolsky - rosen paradox _ , phys . rev . lett . * 98 * , 140402 ( 2007 ) . j. barrett , l. hardy , and a. kent , _ no signaling and quantum key distribution _ , phys . rev . lett . * 95 * , 010503 ( 2005 ) ; a. acn , n. gisin , and l. masanes , _ from bell s theorem to secure quantum key distribution _ , phys . rev . lett . * 97 * , 120405 ( 2006 ) ; a. acn _ et al_. , _ device - independent security of quantum cryptography against collective attacks _ , phys . rev . lett . * 98 * , 230501 ( 2007 ) ; c. branciard , e. g. cavalcanti , s. p. walborn , v. scarani and h. m. wiseman , _ one - sided device - independent quantum key distribution : security , feasibility , and the connection with steering _ , phys . rev . a * 85 * , 010301(r ) ( 2012 ) . r. colbeck , _ quantum and relativistic protocols for secure multi - party computation _ , phd thesis , arxiv : 0911.3814 , ( 2009 ) ; s. pironio _ et al_. , _ random numbers certified by bell s theorem _ , nature * 464 * , 1021 , ( 2010 ) ; r. gallego _ et al_. , _ full randomness from arbitrarily deterministic events _ , nature communications * 4 * , 2654 ( 2013 ) . r. gallego and l. aolita , _ resource theory of steering _ , phys . rev . x * 5 * , 041008 ( 2015 ) . b. hensen _ et al_. , _ loophole - free bell inequality violation using electron spins separated by 1.3 kilometres _ , nature * 526 * , 682 ( 2015 ) ; l. k. shalm _ et al_. , _ strong loophole - free test of local realism _ , phys . rev . lett . * 115 * , 250402 ( 2015 ) ; m. giustina _ et al_. , _ significant - loophole - free test of bell s theorem with entangled photons _ , phys . rev . lett . * 115 * , 250401 ( 2015 ) . b. wittmann _ et al_. _ loophole - free einstein - podolsky - rosen experiment via quantum steering _ , new j. phys . * 14 * , 053030 ( 2012 ) . v. capasso , d. fortunato , and f. selleri , _ sensitive observables of quantum mechanics _ , int . j. theor . phys . * 7 * , 319 ( 1973 ) . n. gisin , _ bell s inequality holds for all non - product states _ , phys . lett . a * 154 * , 201 ( 1991 ) . d. home and f. selleri , _ bell s theorem and the epr paradox _ , rivista del nuovo cimento * 14(9 ) * , 1 ( 1991 ) . s. popescu and d. rohrlich , _ generic quantum nonlocality _ , phys . lett . a * 166 * , 293 ( 1992 ) . r. f. werner , _ quantum states with einstein - podolsky - rosen correlations admitting a hidden - variable model _ , phys . rev . a * 40 * , 4277 ( 1989 ) . r. horodecki , p. horodecki , and m. horodecki , _ violating bell inequality by mixed spin-1/2 states : necessary and sufficient condition _ , physics letters a * 200 * , 340 - 344 ( 1995 ) . d. j. saunders , s. j. jones , h. m. wiseman , and g. j. pryde , _ experimental epr - steering using bell - local states _ , nat . phys . * 6 * , 845 ( 2010 ) . m. t. quintino , t. vrtesi , d. cavalcanti , r. augusiak , m. demianowicz , a. acn , and n. brunner , _ inequivalence of entanglement , steering , and bell nonlocality for general measurements _ , phys . rev . a * 92 * , 032107 ( 2015 ) . j. bowles , f. hirsch , m. t. quintino , and n. brunner , _ sufficient criterion for guaranteeing that a two - qubit state is unsteerable _ , arxiv:1510.06721 ( 2015 ) . f. hirsch , m. t. quintino , t. vrtesi , m. f. pusey , and n. brunner , _ algorithmic construction of local hidden variable models for entangled quantum states _ , arxiv:1512.00262 ( 2015 ) ; d. cavalcanti , l. guerini , r. rabelo , and p. skrzypczyk , _ general method for constructing local - hidden - state ( and -variable ) models for multiqubit entangled states _ , arxiv:1512.00277 ( 2015 ) . e. g. cavalcanti , s. j. jones , h. m. wiseman and m. d. reid , _ experimental criteria for steering and the einstein - podolsky - rosen paradox _ , phys . rev . a * 80 * , 032112 ( 2009 ) . p. skrzypczyk , m. navascus , and d. cavalcanti , _ quantifying einstein - podolsky - rosen steering _ , phys . rev . lett . * 112 * , 180404 ( 2014 ) . e. g. cavalcanti , c. j. foster , m. fuwa , and h. m. wiseman , _ analog of the clauser - horne - shimony - holt inequality for steering _ , j. opt . soc . am . b * 32 * , a74-a81 ( 2015 ) . m. piani and j. watrous , _ necessary and sufficient quantum information characterization of einstein - podolsky - rosen steering _ , phys . rev . lett . * 114 * , 060404 ( 2015 ) . i. kogias , p. skrzypczyk , d. cavalcanti , a. acn , and g. adesso , _ hierarchy of steering criteria based on moments for all bipartite quantum systems _ , phys . rev . lett . * 115 * , 210401 ( 2015 ) . p. girdhar and e. g. cavalcanti , _ all two qubit states that are steerable via chsh - type correlations are bell - nonlocal _ , arxiv:1601.01703 ( 2016 ) . r. uola , c. budroni , o. ghne , and j .- p . pellonp , _ one - to - one mapping between steering and joint measurability problems _ , phys . rev . lett . * 115 * , 230402 ( 2015 ) . j. f. clauser , m. a. horne , a. shimony , and r. a. holt , _ proposed experiment to test local hidden - variable theories _ , phys . rev . lett . * 23 * , 880 ( 1969 ) . p. m. pearle , _ hidden - variable example based upon data rejection _ , phys . rev . d , * 2 * 1418 ( 1970 ) . s. l. braunstein and c. m. caves , _ wringing out better bell inequalities _ , ann . phys ( n.y ) , * 202 * , 22 ( 1990 ) . d. collins and n. gisin , _ a relevant two qubit bell inequality inequivalent to the chsh inequality _ , j. phys . a : math . gen . * 37 * 1775 ( 2004 ) . n. d. mermin , _ extreme quantum entanglement in a superposition of macroscopically distinct states _ , phys . rev . lett . * 65 * 1838 ( 1990 ) ; m. ardehali , _ bell inequalities with a magnitude of violation that grows exponentially with the number of particles _ , phys . rev . a * 46 * 5375 ( 1992 ) a. v. belinsky and d. n. klyshko , _ a modified @xmath1-particle bell theorem , the corresponding optical experiment and its classical model _ , phys lett . a * 176 * 415 ( 1993 ) . g. svetlichny , _ distinguishing three - body from two - body nonseparability by a bell - type inequality _ , phys . rev . d * 35 * , 3066 ( 1987 ) . d. collins , n. gisin , s. popescu , d. roberts and v. scarani , _ bell - type inequalities to detect true @xmath92-body nonseparability _ , phys . rev . lett . * 88 * 170405 ( 2002 ) ; m. seevinck and g. svetlichny , _ bell - type inequalities for partial separability in @xmath1-particle systems and quantum mechanical violations _ , phys . rev . lett . * 89 * 060401 ( 2002 ) . l. aolita , r. gallego , a. cabello , and a. acn , _ fully nonlocal , monogamous , and random genuinely multipartite quantum correlations _ , phys . rev . lett . * 108 * , 100401 ( 2012 ) . j .- d . bancal , n. gisin , y .- c . liang , and s. pironio , _ device - independent witnesses of genuine multipartite entanglement _ , phys . rev . lett . * 106 * , 250404 ( 2011 ) . k. f. pl and t. vrtesi , _ multisetting bell - type inequalities for detecting genuine multipartite entanglement _ , phys . rev . a * 83 * 062123 ( 2011 ) . d. cavalcanti , p. skrzypczyk , g. h. aguilar , r. v. nery , p. h. souto ribeiro , and s. p. walborn , _ detection of entanglement in asymmetric quantum networks and multipartite quantum steering _ , nat . commun . * 6 * , 7941 ( 2015 ) . a. fine , _ joint distributions , quantum correlations , and commuting observables _ , j. math . phys . * 23 * , 1306 ( 1982 ) ; + a. fine , _ hidden variables , joint probability , and the bell inequalities _ , phys . rev . lett . * 48 * , 291 ( 1982 ) . a. elitzur , s. popescu , and d. rohrlich , phys . lett . a * 162 * , 25 ( 1992 ) . a. garuccio and f. selleri , _ systematic derivation of all the inequalities of einstein locality _ , found phys * 10 * , 209 ( 1980 ) . s. popescu , _ bell s inequalities and density matrices : revealing `` hidden '' nonlocality _ , phys . rev . lett . * 74 * , 2619 ( 1995 ) ; + n. gisin , _ hidden quantum nonlocality revealed by local filters _ , phys . lett . a * 210 * , 151 ( 1996 ) ; a. peres , _ collective tests for quantum nonlocality _ , phys . rev . a 54 , 2685 ( 1996 ) . f. hirsch , m. t. quintino , j. bowles , and n. brunner , _ genuine hidden quantum nonlocality _ , phys . rev . lett . * 111 * , 160402 ( 2013 ) . m. navascus and t. vrtesi , _ activation of nonlocal quantum resources _ , phys . rev . lett . * 106 * , 060403 ( 2011 ) ; + c. palazuelos , _ superactivation of quantum nonlocality _ , phys . rev . lett . * 109 * , 190401 ( 2012 ) . d. cavalcanti , m. l. almeida , v. scarani , and a. acn , _ quantum networks reveal quantum nonlocality _ , nature comms . , 184 ( 2011 ) ; d. cavalcanti , a. acn , n. brunner , and t. vrtesi , _ all quantum states useful for teleportation are nonlocal resources _ , phys . rev . a * 87 * , 042104 ( 2013 ) . a. s. de , u. sen , c. brukner , v. buek , and m. zukowski , _ entanglement swapping of noisy states : a kind of superadditivity in nonclassicality _ , phys . rev . a * 72 * , 042310 ( 2005 ) . l. masanes , y .- c . liang , and a. c. doherty , _ all bipartite entangled states display some hidden nonlocality _ , phys . rev . lett . * 100 * , 090403 ( 2008 ) . m. hillery , v. buek , and a. berthiaume , phys . rev . a * 59 * , 1829 ( 1999 ) . i. kogias , y. xiang , q. he , and g. adesso , _ unconditional security of entanglement - based quantum secret sharing schemes _ , arxiv:1603.03224v1 ( 2016 ) . a. c. s. costa and r. m. angelo , _ quantification of einstein - podolski - rosen steering for two - qubit states _ , arxiv:1510.08030 ( 2015 ) . q. quan , h. zhu , h. fan , and w .- l . yang , _ einstein - podolsky - rosen correlations and bell correlations in the simplest scenario _ , arxiv:1601.00962 ( 2016 ) . d. cavalcanti and p. skrzypczyk , _ quantum steering : a short review with focus on semidefinite programming _ , arxiv:1604.00501v1 ( 2016 ) . z. cao , h. zhou , x. yuan , and x. ma , _ source - independent quantum random number generation _ , phys . rev . x * 6 * , 011020 ( 2016 ) . j. barrett , _ nonsequential positive - operator - valued measurements on entangled mixed states do not always violate a bell inequality _ , phys . rev . a , * 65 * , 042302 ( 2002 ) . * appendix *
it has been known for the last ten years that the cosmological evolution of massive galaxies shows signs of ` down - sizing ' @xcite , i.e. of a progressive decrease of the typical mass of actively star forming galaxies . many pieces of evidence , brought forward also during this meeting ( see e.g. the contributions of bender , kauffmann and danese to these proceedings ) , suggest that the baryonic physics of star formation processes counterbalance the hierarchical ( bottom - up ) growth of dark matter structures , favouring the early formation and growth of the most massive ( and dense ) stellar aggregates . the ubiquity of smbh in galactic nuclei , and their tight relation with their hosts bulge properties @xcite seem to indicate that the formation and growth of galaxies and of their resident black holes have proceeded in parallel , and probably influenced each other in the early stages of galaxy formation . as a matter of fact , the number of theoretical studies dedicated to agn feedback in galaxy formation has flourished in the last five years ( see e.g. loeb s contribution in this proceedings , and references therein ) . furthermore , a recent comprehensive study of 23 000 local agn carried out by the sloan digital sky survey ( sdss , @xcite ) have demonstrated , in a direct quantitative way , that accretion onto smbh and formation of stars are tightly coupled even in the local universe . is it possible to learn more about the formation and growth of structures by just looking at the evolution of the agn population ? the aim of the work presented here is to show to what extent this is indeed possible , and to describe a robust , self - consistent way to unveil the history of cosmic accretion and the build - up of smbh in the nuclei of galaxies , in the form of their evolving mass function . the methodology and the main results will be discussed in the next section , while in section [ sec : times ] i will expand on the consequences of these results for the issue of qso lifetimes . section 4 will summarize my conclusions . under the standard assumption that black holes grow mainly by accretion @xcite , the cosmic evolution of the smbh accretion rate and its associated mass density can be calculated from the luminosity function of agn : @xmath3 , where @xmath4 is the bolometric luminosity produced by a smbh accreting at a rate of @xmath5 with a radiative efficiency @xmath6 . in practice , the accreting black hole population is always selected through observations in specific wavebands . crucial is therefore the knowledge of two factors : the completeness of any specific agn survey , and the bolometric correction needed in order to estimate @xmath7 from the observed luminosity in any specific band . on both these issues , huge progress has been made in the last few years ( see e.g. @xcite ) . in order to progress from the study of bh mass densities to that of bh mass functions , we need to break the degeneracy between mass and accretion rate of any given observed agn . while in most semi - analytic works this is done by assuming a constant eddington ratio for all sources , here i will propose an alternative , physically motivated , method . in a recent paper @xcite it has been shown that the hard ( 2 - 10 kev ) x - ray luminosity of an accreting black holes is related to its mass and its core radio ( at 5ghz ) luminosity by the following relation ( the `` fundamental plane '' of black hole activity ) : @xmath8 ; which can be inverted to relate bh masses to observed nuclear radio and x - ray luminosities : @xmath9 . one of the consequences of this relation is that , in an ideal case , the _ conditional radio / x - ray _ luminosity function of active black holes , i.e. the number of sources per unit co - moving volume per unit logarithm of radio and x - ray luminosity , @xmath10 , could be used to reconstruct the mass function of the underlying black hole population . in fact , the current lack of the exact knowledge of @xmath11 can be ( at least partially ) superseded , given the two separate radio , @xmath12 , and x - ray , @xmath13 , luminosity functions at redshift @xmath14 , and an _ independent _ estimate of the black hole mass function , @xmath15 at the same redshift . by taking into account the fundamental plane relationship , we have that , at any @xmath14 , the conditional luminosity function @xmath16 has to satisfy the following integral constraints : @xmath17 @xmath18 given observational estimates of @xmath19 , @xmath20 and @xmath21 , we start with an initial guess for @xmath16 , and proceed via successive iterations , minimizing the differences between the projections of the conditional luminosity function onto the x - ray and radio luminosity axes and the observed luminosity functions , until a conditional lf is obtained simultaneously satisfying eqs . ( [ eq : int_ij ] ) and ( [ eq : int_m ] ) . once such an estimate of @xmath16 is found , it is possible to derive the local distribution of the x - ray to eddington ratio , and from this , given an appropriate bolometric correction ( i.e. a specific functional form @xmath22 , where @xmath23 ) , the desired accretion rate function . the redshift evolution of the smbh population can then be computed integrating _ backwards _ the continuity equation that describes smbh evolution driven by accretion only @xcite : @xmath24}{\partial m}=0,\ ] ] where the mean accretion rate as a function of black hole mass and time , @xmath25 is calculated directly from the accretion rate distribution function at time @xmath26 . starting from @xmath0 , the local bhmf , as determined independently from the galaxy velocity dispersion distribution and the @xmath27 relation , can be used as a boundary condition to integrate eq . ( [ eq : continuity ] ) up to the redshift where hard x - rays and radio luminosity functions of agn can be reliably estimated . the only parameters needed are the accretion efficiency @xmath28 , and the functional form @xmath29 . [ cols="^,^ " , ] here i present the results obtained using the most recent determinations of the hard x - rays and radio lf @xcite , fixing @xmath30 , and assuming for the function @xmath31 a simple broken power - law expression , with @xmath32 at low accretion rates ( radiatively inefficient regime ) and @xmath33 at high luminosities ( radiatively efficient regime , see @xcite for details ) , with the transition being placed at @xmath34 @xcite . the evolution of the black hole mass function is shown in the left panel of figure [ fig : mf ] . as opposed to the standard picture of hierarchical mass build up of dark matter halos in cdm cosmologies , supermassive black holes growing by accretion between @xmath35 and now have a mass function which is more and more dominated by larger mass objects at higher redshift . thus , most of the more massive black holes ( @xmath2 ) were already in place at @xmath1 , and as a result , the `` typical '' smbh mass decreases with decreasing redshift . smbh appear to be growing in anti - hierarchical fashion @xcite . the right panel of figure [ fig : mf ] shows instead the redshift evolution of the derived accretion rate function ( expressed here as x - ray to eddington ratio , @xmath36 ) . while the number of sources accreting at low rates increases monotonically with decreasing redshift , the situation is different for rapidly accreting objects , that increase in number with increasing redshift . the cut - off redshift , above which the number of sources declines again , is a function of the typical x - ray to eddington ratio , being lower for lower accretion rate sources . once again , it is clear that the `` typical '' x - ray to eddington ratio , which can be approximately identified with the knee of the accretion rate functions of fig.1 , decreases with decreasing redshift , crossing the critical rate that separates radiatively inefficient from efficient regimes ( here fixed at 10@xmath37 , vertical dashed line ) around @xmath38 . in all models that derive the properties of the smbh population from the observed qso evolution , a key element is represented by the typical quasar lifetime or by the almost equivalent activity duty cycle . however , the significance of these parameters is limited to the standard case in which , on the basis of an observed luminosity function in a specific waveband , one tries to derive the distribution of either bh masses or accretion rates . usually , a constant eddington ratio is assumed in this case , which implies that qso are considered as on - off switches . then , the duty cycle is simply the fraction of black holes active at any time , and the lifetime is the integral of the duty cycle over the age of the universe ( see e.g. @xcite and references therein ) . the picture discussed here is different , in that a broad distribution of eddington rates is not only allowed , but actually calculated for the smbh population at every redshift . when this is the case , a more meaningful definition of activity lifetime is needed . let us first define the mean eddington rate for object of mass @xmath39 at redshift @xmath0 @xmath40 and then introducing the _ mean accretion weighted lifetime _ of a smbh with a given mass _ today _ : @xmath41 . the ratio of @xmath42 to the salpeter time , @xmath43 yrs , gives the mean number of @xmath44-folds of mass growth for objects with mass @xmath39 up to redshift @xmath14 . the ratio of @xmath42 to the hubble time @xmath45 , instead , is a measure of the activity duty cycle of smbh . it is also interesting here to calculate `` partial '' lifetimes in a given redshift interval @xmath46 : @xmath47 , calculated for three different redshift intervals : @xmath48 ( solid line ) , @xmath49 ( dashed line ) and @xmath50 ( dot - dashed line ) . the horizontal dotted line is the salpeter time for accretion efficiency of 10% . ] in figure [ fig : tau ] , i show @xmath51 for three redshift intervals : @xmath52 ; @xmath53 and @xmath54 . the accretion weighted lifetime for bh of any given mass between @xmath55 is of course just the sum of the three . the anti - hierarchical nature of mass build - up in actively accreting agn and qsos is again clearly illustrated by this plot . in fact , the major growth episode of a smbh must coincide with the period when @xmath56 . this happens at @xmath57 for @xmath58 , between redshift 1 and 2 for @xmath59 , and at @xmath50 for @xmath60 . supermassive black holes with masses larger than @xmath61 today , must have experienced their major episodes of growth at redshift higher than 3 . it also interesting to note that the objects that dominate the smbh mass function today , i.e. those in the range of masses around @xmath62 , where @xmath63 peaks , mainly grew around @xmath64 , which is when most of the x - ray background light we see today was emitted @xcite . i have presented a new method to study the growth of accreting supermassive black holes , based on the simultaneous evolution of the agn radio and hard ( 2 - 10 kev ) x - ray luminosity functions . the method is based on the locally observed trivariate correlation between black hole mass , x - ray and radio luminosity ( the so - called fundamental plane of black hole activity , @xcite ) . thanks to this correlation , it is possible for the first time to break the degeneracy between luminosity , mass and accretion rate : qso and agn not only grow in mass during their evolution , but also accrete a different rates depending on their mass and age . qualitatively ( i.e. independently on the values of the model parameters ) , the evolution of the black hole mass function between @xmath0 and @xmath1 shows clear signs of an _ anti - hierarchical _ behaviour : while the majority of the most massive objects ( @xmath2 ) were already in place at @xmath35 , lower mass ones mainly grew at progressively lower redshift , so that the average black hole mass increases with increasing redshift . at the same time , the typical accretion rate of smbh decreases with decreasing redshift . consequently , the lifetime of an actively growing bh , measured through the partial mean accretion weighted lifetime ( see section [ sec : times ] ) , is a strong function of both redshift and black hole mass . broadly speaking , what is presented here is the smbh analog of the down - sizing of star forming galaxies . future more direct and detailed comparisons between smbh mass functions , such as those shown here , and galaxy luminosity functions , both at low and high redshifts , will hold the key of our understanding of the complex interplay between formation and growth of black holes and galaxies . j. kormendy , d. richstone : ara&a * 33 * , 581 ( 1995 ) ; j. magorrian , et al . : aj * 115 * , 2285 ( 1998 ) ; l. ferrarese , d. merritt : apjl * 539 * , l9 ( 2000 ) ; k. gebhardt , et al . : apjl * 539 * , l13 ( 2000 ) ; s. tremaine , et al . : apj * 574 * , 740 ( 2002 ) ; a. marconi , l. k. hunt : apjl * 589 * , l21 ( 2003 )	i present a new method to unveil the history of cosmic accretion and the build - up of supermassive black holes ( smbh ) in the nuclei of galaxies , based on observations of the evolving radio and ( hard ) x - ray luminosity functions of agn . the fundamental plane of black hole activity discovered by merloni , heinz & di matteo ( 2003 ) is used as a mass and accretion rate estimator . i adopt the local bh mass function as a boundary condition to integrate backwards in time the continuity equation for the smbh evolution , neglecting the role of mergers . under the most general assumption that accretion proceeds in a radiatively efficient way above a certain rate , and in a radiatively inefficient way below , the redshift evolution of the mass and accretion rate functions are calculated self - consistently . the only tunable parameters are the accretion efficiency and the critical ratio of the x - ray to eddington luminosity at which the transition between accretion modes takes place . the evolution of the bh mass function between @xmath0 and @xmath1 shows clear signs of an anti - hierarchical behaviour : while the majority of the most massive objects ( @xmath2 ) were already in place at @xmath1 , lower mass ones mainly grew at progressively lower redshift . as an example , i will discuss the consequences of these results for the lifetimes of accreting black holes .
periodically - kicked quantum systems , such as the @xmath0-kicked particle ( @xmath0-kp ) , have long played a central role in studies of quantum chaos and the correspondence between quantum behavior and the underlying classical dynamics @xcite . advances in the manipulation of cold atoms have permitted the experimental realization of these systems in pulsed optical lattices @xcite . experiments with sodium and cesium atoms have demonstrated the phenomenon of `` dynamical localization '' ( dl ) the quantum suppression of classical chaotic diffusion and established the suitability of these systems as an arena for the study of effects arising from quantum chaos . when treating conservative quantum systems it is frequently useful to study the system s energy - eigenstates , and for periodically driven systems the appropriate generalization of these states is given by the floquet states . this approach has provided extensive insight into the properties of the standard qkp , and has shown , for example , that dl arises directly from the exponential localization of the system s floquet states @xcite . observed momentum oscillations associated with chaos - assisted tunneling , in experiments using periodically - driven cold atoms @xcite and becs @xcite have also been analysed with floquet theory ; it was found that the oscillation period is determined by the splittings of the floquet phases of a pair of symmetry - related eigenstates . the statistics of qkp floquet quasi - energy spectrum have been studied extensively see e.g. @xcite and compared with the predictions of random matrix theory . notably , though , the @xmath0-qkp has poissonian short - range statistics ( which are typically associated with integrable dynamics ) even for very large values of @xmath1 , where the dynamics is fully chaotic . this has been shown to be a further consequence of dl @xcite . however , a series of recent theoretical @xcite and experimental @xcite studies of cold atom systems subjected to repeating cycles of unequally spaced kicks revealed dynamics rather different from that found in the corresponding standard qkp systems . two types of unequally @xmath0-kicked systems were investigated . the first @xcite consists of a @xmath0-kp with small perturbations made to the kick - period . we term it the perturbed - period kp . in the second system , @xcite the system is periodically subjected to _ pairs _ of closely - spaced kicks . this is referred to as the double @xmath0-kp or 2@xmath0-kp . in a theoretical study , the perturbed - period kp was found to yield a quantum momentum current even in the chaotic regime @xcite . this was unexpected in a chaotic hamiltonian system , since to date only mixed phase - space ratchet mechanisms had been investigated @xcite . a simple definition of a ratchet is a spatially periodic device which produces a current without net bias ( ie the time and space - averaged forces are zero ) . most of the extensive ratchet literature deals with dissipative or brownian ratchets @xcite and comparatively little theoretical work has been undertaken on hamiltonian ratchets , which are dissipation and noise - free . in @xcite , a momentum distribution with a non - zero average ( constant in time ) was obtained experimentally from an atomic cloud with initial zero average momentum . we are unaware of any other experimental studies of hamiltonian quantum ratchets : all implementations to date have been of dissipative / brownian ratchets . hence the results from @xcite and also reproduced here , represent the only implementation of a type of hamiltonian quantum ratchet , whether chaotic or mixed phase - space . in ref . @xcite it was also proposed that the chaotic diffusive properties of the perturbed - period kp could be exploited to filter cold atoms , according to their momenta , by controlling the dynamical localization . for the second system , the 2@xmath0-kp , a theoretical and experimental study @xcite revealed that the diffusion is dominated by long - ranged correlations which control escape from well - defined momentum trapping regions . this combination of strong chaotic diffusion and long - ranged correlations is novel : strong chaos is generally associated with rapidly decaying correlations . it is clear that floquet theory is central to the analysis of chaotic , time - periodic quantum systems . the need to understand further the chaotic hamiltonian ratchet as well as the 2@xmath0-kp motivated this study of the floquet states of these systems . the paper is organized as follows . in the next section we review the well - known @xmath0-kp , then introduce the perturbed @xmath0-kp and the double @xmath0-kp systems . in section iii we give a brief review of the floquet approach . in section iv we compare the results with recently obtained experiments on these systems . finally in section v we give our conclusions . the hamiltonian for the usual @xmath0-kp can be written as @xmath2 where @xmath1 is the kick strength and @xmath3 is the time interval between successive kicks . consider its effect on an ensemble of particles with a gaussian momentum distribution centered on @xmath4 , @xmath5 $ ] . the classical dynamics depends only on the parameter @xmath1 , and for values larger than @xmath6 , the chaotic diffusion is not bounded by classical barriers . in this regime the ensemble will diffuse in momentum space , its average energy growing linearly with time as @xmath7 , where , to lowest order , the diffusion rate is given by @xmath8 . the distribution will thus remain gaussian , although its width will increase with time as @xmath9 . in contrast , the quantum system only follows this behavior up to a timescale @xmath10 @xcite , after which diffusion in momentum space is suppressed dynamical localization ( dl ) . such a system will asymptotically evolve towards a characteristic exponential momentum distribution , @xmath11 $ ] , with constant width @xmath12 , which thus acts as an experimental fingerprint for dl @xcite . as dl is a wave - coherent effect , the quantum system must preserve coherence over at least the timescale @xmath13 for this effect to be observable . indeed , it has been verified experimentally @xcite that the dl profile does not survive the presence of noise or dissipation , and that with decoherence a more gaussian profile for @xmath14 will be produced . in both classical and quantum cases the behavior of the standard @xmath0-kp is essentially independent of @xmath4 since , even for modest values of @xmath1 , the effects of small fluctuations in the structure of phase - space are on negligible scales relative to @xmath15 . even if there are small stable islands , they are of size @xmath16 so have little effect on the general form of @xmath14 , since typically @xmath17 . the classical dynamics of the @xmath0-kp is obtained by iterating the well - known `` standard map '' . for the perturbed - period and double @xmath0- kicked systems , on the other hand , the dynamics is given by a 2-kick map : @xmath18 clearly , setting @xmath19 and @xmath20 , we recover the standard map . for the perturbed - period kp , the lengths of the two kicking periods are @xmath21 and @xmath22 , where @xmath23 . the perturbation thus consists of slightly altering the kicking period about its mean . for the double @xmath0-kp we take @xmath24 , @xmath25 , although we shall also show the effect of interchanging the two kick periods . it should be noted that these systems are _ time - periodic _ , with period @xmath26 , and are thus quite distinct from the recent interesting study of two independent kicking sequences , which can be non - periodic and hence non - localizing @xcite . as in the standard map , we consider a sinusoidal potential @xmath27 . however , to obtain a ratchet current in the case of the perturbed - period kp , we need to break the spatio - temporal symmetries , and so we add an additional `` rocking '' linear potential of strength @xmath28 . in this case the form of the potential is @xmath29 $ ] , where @xmath30 is the kick number . in experimental implementations of this system , the rocking linear term was obtained by means of an accelerated lattice @xcite . we first consider in general terms how the introduction of the second timescale @xmath31 modifies the classical behavior of the standard map . if we neglect all correlations , the standard map has a constant momentum diffusion rate , @xmath32 this is what one would expect if the momenta at consecutive kicks are uncorrelated and so evolve as a random walk . however , unless @xmath1 is exceedingly large , the time - evolution of the standard map will contain some short - range ( 2-kick and 3-kick ) correlations . including these corrections yields a modified diffusion constant @xmath33 $ ] . of particular interest is the @xmath34 term , representing correlations @xmath35 between nearest - but - one kicks ( the 2-kick correlation ) . for the modified systems it is also possible to obtain analytically the important correlations @xcite . for instance , for the perturbed - period system the corrected diffusion is @xmath36 $ ] . we see that we have a modified 2-kick correlation which _ oscillates _ with the initial momentum @xmath4 . this effect is clearly most significant for values of @xmath1 such that @xmath37 . the key point is that perturbing the kick spacings @xmath3 by a small amount can result in large scale ( relative to @xmath38 ) variations in the classical momentum diffusion , and that these are present even in fully chaotic regimes ( we take this to mean the absence of visible stable structures on the poincar surface of section ) . for the analysis of experiments , one must now consider a _ local _ diffusion rate @xmath39 , which depends on the initial relative momentum between the atoms and the optical lattice . in @xcite , the perturbed - period system was implemented experimentally with a cloud of cesium atoms for the case @xmath40 . it was verified that the energy absorbed by the cloud after dynamical localization @xmath41 as expected . however , @xmath40 corresponds to a symmetric potential . the case @xmath42 in the perturbed - period system is particularly interesting since then the momentum diffusion is asymmetric about @xmath43 . this implies that atoms with positive momenta will absorb kinetic energy at different rates from those with momenta of the same magnitude but moving in the opposite direction . this asymmetric momentum diffusion represents a type of fully chaotic _ momentum _ ratchet : in other words , roughly equal numbers of particles will diffuse to the left or to the right , but those diffusing to the right , on average , move faster . in a further experimental study of the perturbed - period kp @xcite , the @xmath42 potential was implemented by means of an accelerated lattice . it was found that an atomic cloud prepared initially with a gaussian momentum distribution centered on @xmath44 evolved into a distribution with non - zero , but constant , @xmath45 which persisted even beyond the break - time @xmath13 , as expected from the theory @xcite . this type of chaotic directed motion was first identified in a slightly different system : a kicked asymmetric double - well potential @xcite . however , the latter potential gives rise to a rather more complicated diffusive behavior , and also has proved much harder to implement experimentally . for these reasons here we do not consider the case of the asymmetric double - well ratchet , but note that our floquet analysis of the perturbed - period kp can be carried - over to the system investigated in @xcite . the second system we consider explicitly in this work , the 2@xmath0-kp , has diffusive behavior which is qualitatively different to both the standard map and perturbed - period kp . while for these other kicked systems we can analyze the diffusion as an uncorrelated term , @xmath46 , corrected by short - ranged correlations ( typically only 2 or 3-kick correlations for @xmath47 ) , for the double @xmath0-kp , we find that the diffusion at long times is dominated by families @xcite of long - ranged `` global '' correlations ( `` global '' in the sense that they correlate all kicks up to the time under consideration ) . at short times , the diffusion is dominated by a 1-kick correlation not present in other kick systems . at longer times , the global diffusion terms , though weak , accumulate and eventually become dominant . the method of correlations provides a generic and accurate way of interpreting experimental data for this system @xcite . there is also a simple physical picture . for particles subjected to kicks of form @xmath48 , consecutive kicks will be out of phase and will hence cancel if @xmath49 where @xmath50 . in other words , an impulse @xmath51 will be immediately followed by another which cancels it , since @xmath52 . this cancellation means that particles become trapped at these momenta . in contrast , particles for which @xmath53 , will experience enhanced diffusion . it was shown in @xcite that the new types of global families of correlations control the escape from , and through , these `` trapping regions '' . an unexpected feature of the classical calculations ( and also seen in experiment ) was the observation that particles initially prepared in the trapping regions will eventually gain more energy than those initially prepared in regions of enhanced diffusion , after a timescale @xmath54@xcite . if a hamiltonian has a @xmath3-periodic time dependence , @xmath55 , then the floquet theorem implies that solutions to the time - dependent schrdinger equation can be written in the form @xmath56 \phi(t)\ ] ] where @xmath57 is a @xmath3-periodic function called a floquet state , and @xmath31 is a quantity with dimensions of energy , termed a quasienergy . this type of relation is familiar in the context of solid - state physics , where a hamiltonian s invariance under discrete shifts of _ spatial _ position ( typically arising from a lattice structure ) allow solutions to be written analogously in terms of bloch states and quasimomenta . the floquet states provide a complete basis , and thus the time - evolution of a general quantum state under periodic driving can be expressed as @xmath58 \phi_n(t ) , \label{floquet}\ ] ] where @xmath59 are _ time - independent _ expansion coefficients . it is clear from this expression that the floquet states and quasienergies play a similar role for periodically - driven systems to that of energy eigenstates and eigenvalues in the time - independent case . the time - evolution operator @xmath60 may be used to evolve a quantum state from time @xmath61 to a time @xmath62 . for a time - periodic system , the single - period propagator @xmath63 allows a quantum system to be evolved `` stroboscopically '' at intervals separated by the period @xmath3 with great efficiency , by defining the quantum map @xmath64^n \psi(0)$ ] . in terms of floquet states , it is straightforward to show that the time - evolution operator is given by @xmath65 \label{define } \\ { } & = & \sum_n e^{-i \epsilon_n t / \hbar } @xmath66 is the time - ordering operator , and thus it can be seen that the quasienergies and floquet states can be conveniently obtained by simply diagonalising the one - period propagator . the eigenvectors of this operator are the floquet states , while its eigenvalues are related to the quasienergies via @xmath67 $ ] . obtaining an explicit form for this propagator is normally a complicated procedure , as in general the driving field does not commute with the static hamiltonian . for the case of @xmath0-kicking , however , the problem is simplified considerably which allows an analytic form for the propagator to be written . for the qkp ( eq.[ham_kp ] ) the propagator is given by @xmath68 \exp\left [ -i ( k / \hbar ) \sin x \right ] .\ ] ] using a basis of plane - wave states , the matrix - elements @xmath69 of this operator can easily be shown to be @xmath70 j_{m - n}(k / \hbar ) \delta ( q - q ' ) , \label{prop}\ ] ] where @xmath71 is the quasimomentum ( @xmath72 ) and @xmath73 is the @xmath74-th bessel function of the first kind . for practical purposes it is useful to note that @xmath75 decreases extremely rapidly with increasing @xmath76 , thus giving @xmath77 an effectively banded - structure . the single - period propagators for the unequally - kicked systems can now be expressed as the product of two matrices of this form , @xmath78 , where @xmath79 and @xmath80 are the two kick - periods . it should be noted that although the single - kick propagators do not conserve quasimomentum , their product does . as a consequence the evolution of an ensemble of non - interacting particles which can be modelled by the evolution of a superposition of states with different quasimomenta , represents a computationally efficient procedure : we can consider each quasi - momentum component independently . in our study of floquet states in effect this means we can diagonalise the smaller matrix given by [ prop ] , which is block - diagonal in @xmath71 . having obtained the floquet states by diagonalising ( [ prop ] ) , it is useful to analyse their structure , particularly their spread in momentum space . we do this by evaluating a localization length , @xmath81 of each floquet state at @xmath82 . we note that in this case we can not assume that the floquet states have the usual exponential momentum distribution @xmath83 $ ] , of the usual @xmath0-kp . hence we take @xmath81 to be simply the root mean square deviation from the mean , @xmath84 . in this expression @xmath85 is the mean momentum of the @xmath74-th floquet state at @xmath82 , @xmath86 , and @xmath87 . if the floquet states do not have a strong time - dependence , such as for the standard , or even the perturbed - period kp , this is adequate to quantify the degree of spreading in momentum space . we shall see , however , that for the case of the 2@xmath0 kicked system it is not sufficient to measure the localization at a single time , due to the extremely strong time - dependence of the floquet states . fig.[fig1 ] shows a plot of the experimental ratchet current obtained in @xcite . a series of momentum distributions @xmath88 as a function of @xmath4 for a cloud of cold cesium atoms in an optical lattice pulsed with unequal periods . the momentum distribution is essentially unchanged after about 60 kicks ; the plotted values correspond to about @xmath89 kicks , hence well after dynamical localization . full details are given in @xcite , but by employing an accelerated lattice , the experiment simulated an effective rocking potential with @xmath90 . the first moment of each localized distribution @xmath91 was then calculated and plotted as a function of @xmath4 . in particular , a distribution centered at @xmath44 initially , and with zero initial momentum current @xmath92 at @xmath82 , yielded a finite and constant momentum current @xmath93 at long times . for non - zero initial momenta @xmath94 , an oscillation @xmath95 , was observed and is seen in fig.[fig1 ] . this may be qualitatively understood from the form of the classical two - kick momentum - diffusion correction introduced in sec.ii , @xmath96 . if we consider a very small momentum displacement @xmath97 , differential absorption of energy for particles moving to the left @xmath98 or right @xmath99 is proportional to the gradient @xmath100 . an accurate analytical form for the classical current was derived in @xcite . of course , it follows that a finite and persistent constant momentum current is also obtained classically . it was found in @xcite that asymmetric diffusion persists only on a timescale @xmath101 and for this ( unbounded ) chaotic system , the acquired momentum asymmetry is never lost . for a bounded ( `` compact phase - space '' ) system , such asymmetries would vanish on a long time - scale , since the distribution of a fully chaotic system would eventually become uniform . for this reason , until recently , it was argued that a fully chaotic system could not generate directed motion . so , although as shown in @xcite , the fully chaotic classical system can keep a constant current for long times , practical implementation is ultimately less interesting since the average kinetic energy of the ensemble grows linearly with time and without limit . hence , this type of chaotic ratchet is of most interest as a _ quantum _ rather than a _ classical _ ratchet since in the quantum case dl halts the diffusion and `` freezes - in '' the asymmetry , without the need for classical barriers like tori . , for the perturbed - period kp , obtained with cold cesium atoms in a pulsed optical lattice for @xmath47 , @xmath102 . the solid line is a best - fit to the data , showing that the current oscillates sinusoidally as @xmath103 for @xmath42.,width=288 ] for the perturbed - period kp obtained with cold cesium atoms in a pulsed optical lattice for @xmath47 , @xmath102 . the distributions have localized , and hence remain essentially constant with time . the results show clearly that the origin of the net non - zero value of @xmath45 obtained at long times is in the asymmetry of the dl profiles . as expected , the asymmetry is reversed by changing the sign of @xmath28 , the amplitude of the rocking potential.,height=288 ] in fig.[fig2 ] we reproduce two experimental momentum distributions for @xmath47 obtained with cesium atoms in ref.@xcite , for @xmath104 . we clearly see that the origin of the non - zero momentum current is the asymmetric momentum distribution . as expected , fig.[fig2 ] shows that changing the sign of @xmath28 reverses the asymmetry . at this stage it may be unclear to the reader what the significance of altering the sign of @xmath28 in the experiment might be , since after all , the rocking potential involves alternating impulses @xmath105 . in fact the distinction ( as may be ascertained from the form of the classical diffusion ) is between the case where an impulse @xmath106 precedes free evolution for a time interval @xmath107 ( obviously followed by an impulse @xmath108 and interval @xmath109 ) and the separate experimental case where an impulse @xmath108 precedes free evolution for a time - interval @xmath107 and so forth ( which corresponds to a reversed current ) . note that the experimental range of @xmath110 does correspond to a classical surface of section with some islands . however , we note that classical quantities such as the average energy are very accurately given by diffusion rates ( with 2 and 3-kick corrections ) . the essential mechanism is asymmetric chaotic diffusion : similar behavior was found at larger @xmath1 in @xcite in regimes where there are no visible classical islands ( but for which experiments are not available ) ; hence , in the analysis of this type of ratchet , the presence ( or otherwise ) of small stable islands is immaterial . what is important , though , is that since the asymmetric diffusion term is @xmath111 , we need @xmath112 . thus the much - studied ( for the standard map ) parameter value @xmath113 does _ not _ produce asymmetry , since @xmath114 . values of @xmath115 , @xmath116 , on the other hand , turned out to be experimentally convenient and produced the strongest asymmetries . of floquet states as a function of average momentum @xmath117 ( i.e. @xmath45 ) for @xmath118 , @xmath119=1 . results are shown ( a ) for the standard qkp case , i.e. @xmath120 ( a non - zero @xmath28 was used to break spatial symmetry ) , @xmath121 , ( b ) for @xmath122 , @xmath120 and ( c ) @xmath122 , @xmath42 . the graph shows that for the standard kicked rotor the @xmath81 are distributed within a narrow range in comparison with other two below . for the rocking case , @xmath81 oscillates with @xmath117 as expected from the 2-kick correction @xmath123 ; the oscillations of the two lower graphs are shifted relative to each other by a phase @xmath124 . the density of eigenstates corresponding to average momentum range is roughly the same in all three cases.,width=288,height=288 ] we now examine the form of the underlying floquet states . in fig.[fig3 ] we compare the localization lengths for the standard qkp , with those of the perturbed - period kp for @xmath118 , @xmath122 . the difference is quite striking ; while the standard qkp eigenstates are quite uniform across all regions of phase - space , the perturbed - period localization lengths oscillate sinusoidally with @xmath117 , with a period of @xmath125 . introducing the additional rocking potential with the accelerated lattice ( @xmath42 ) clearly leads to a @xmath124 shift in the oscillations . inspecting fig.[fig3](c ) for @xmath126 , we see that for positive momenta the localization lengths are increasing , while for negative momenta , the localization lengths decrease . note the nearly regular row of states for the standard qkp case with @xmath127 . these correspond to states localized on a series of stable islands separated by @xmath128 , due to the momentum periodicity of phase - space in that case . we have chosen a parameter range for which @xmath129 : that is , the localization length of each state is much smaller than the oscillation in @xmath117 . hence individual floquet states really do sample `` local '' diffusion rates . we found that if we move towards a regime where @xmath130 , the conclusions remain valid , but the amplitude of the oscillations is considerably damped . similarly , if the sign of @xmath131 changes , so does the sign of the sinusoidal oscillation . we now consider the shape of the floquet states in detail . in fig.[fig4 ] we show the momentum distributions @xmath132 for floquet states of the standard qkp . the distributions ( with @xmath14 on a logarithmic scale ) all show the well - known triangular form @xcite the hallmark of dynamical localization . it may be clearly seen that the localization lengths vary little from state to state . in fig.[fig5 ] , by contrast , the localization lengths of the floquet states of the perturbed period @xmath0-kp display a strong dependence on the mean momentum of the states . in addition , the figure shows that states localized close to @xmath133 are markedly asymmetric . the states are considerably extended towards positive momentum , but are strongly localized towards negative @xmath134 . this behavior neatly accounts for the form of the experimental momentum distribution shown in fig.1 , which for @xmath42 were also more extended towards positive @xmath134 . the states localized near @xmath135 and @xmath136 correspond to , respectively , minima and maxima of the classical diffusion . they are roughly symmetrical ( typically ) but vary by up to a factor of @xmath137 in @xmath81 . in ref.@xcite it was proposed that the observed variation in the energy absorption rates between atoms prepared with an initial drift momentum @xmath138 ( which absorb very little energy ) and those with @xmath139 might be exploited to filter traffic of atoms through an optical lattice . the form of the underlying floquet states explains this differential rate of energy absorption . subsequently , it was found experimentally that the double @xmath0-kp in fact shows much more pronounced differential absorption rates , without requiring the application of a rocking field @xmath28 . we next report a study of the floquet states of this system . , @xmath140 . as expected , all the states are exponentially localized , giving the characteristic triangular shape of @xmath14 when plotted on a logarithmic scale . they all have approximately similar localization lengths.,width=288,height=144 ] , @xmath122 , @xmath42 and @xmath119=1 . here we plot @xmath132 as a function of @xmath134 . ( a ) states with @xmath141 . this corresponds to a minimum of the 2-kick correction @xmath142 . the states are narrow , but in general , roughly symmetric . ( b ) states with @xmath143 . the typical state here is asymmetric ( c ) states with @xmath144 . this corresponds to a maximum of the 2-kick correction . states here are generally symmetrical , but broad and flat - topped.,title="fig:",width=288 ] + -kp realization with cesium atoms ( see @xcite for details ) . each data point ( star ) shows the energy absorbed ( after 100 kicks , @xmath145 , @xmath140 ) by a cloud of atoms with average momentum @xmath146 ( relative to the optical lattice ) at initial time , @xmath82 . with increasing @xmath31 , we see the minima ( maxima ) in the energy flip into maxima ( minima ) as a long - ranged family of classical correlations gradually overtakes the 1-kick classical correlation . the dashed lines represent a classical simulation using 100,000 particles , all with momenta @xmath4 at @xmath82 , and @xmath1 within the range @xmath147 . * ( a ) * @xmath148 . regime where a one - kick correlation is the dominant correction to the classical diffusion . here , atoms prepared near the trapping regions @xmath149 remain trapped . results follow closely the formula @xmath150 . * ( b)@xmath151 . regime showing the inverted peaks of the poisson correlation terms analyzed in @xcite , which determine the momentum trapping very close to the resonant condition @xmath152 . ( c ) * @xmath153 . regime dominated by correlation family @xmath154 , but sharp inverted peaks due to the poisson correlations are still visible.,title="fig:",width=288 ] + a study of the experimental and classical behavior of the double @xmath0-kp was carried out in @xcite . the classical dynamics is very different to that of the perturbed period kp . at very short times , the chaotic diffusion comprises an uncorrelated diffusion term @xmath46 and one dominant 1-kick correction . it was found in @xcite that one can approximate the growth in the mean energy with time @xmath155 , by the simple expression @xmath156 $ ] . in fig.[fig6](a ) we show experimental results for cesium atoms which localized in this regime . the experiment measured the energy of a series of clouds of @xmath157 atoms moving through the pulsed optical lattice with varying average drift momenta @xmath4 . for fig.[fig6](a ) , the simple expression given above gives an excellent fit to the experiment , if we take @xmath158 , where @xmath13 is the break time . this regime corresponds to @xmath159 . however , a more detailed study of the classical correlations showed that for later times , a new type of correction appeared . families of long - ranged , or `` global '' , correlations which coupled _ all _ kicks appeared . these corrections are individually very weak , but accumulate to eventually dominate the diffusive process . one family ( termed the `` poisson family '' in @xcite ) was shown to lead to well - localized , inverted peaks in the energy absorption at values of @xmath160 , where @xmath161 . these values of @xmath4 correspond to trapping regions in phase - space ( at low values of @xmath1 , structures corresponding to islands and broken phase - space barriers are evident ) . however there is no need to investigate detailed transport through this complex mixed phase - space structure , as the correlations give us a generic and quantitative handle on the energy diffusion with time . in this intermediate regime , dominated by the poisson correlations , atoms prepared outside the trapping regions rapidly diffuse across the regions between them . particles prepared in the trapping regions remain there . this regime occurs for @xmath162 and corresponds approximately to the experimental results shown in fig.[fig6 ] . finally , at the longest timescales , there is the @xmath154 correction investigated in @xcite , which is a long - ranged global - correlation family . @xmath154 results in an oscillation of the form @xmath163 and becomes dominant at the longest timescales . the oscillation is of the same period as the 1-kick correlation , but is of opposite sign . this means that at the longest timescales , the minima in energy absorption shown in fig.[fig6](a ) become maxima in energy absorption ; and vice - versa : the maxima become minima . fig.6(c ) shows experiments tending towards this regime . the inverted peaks of the poisson family are still in evidence , but a @xmath164 oscillation is clearly superposed . this is a somewhat counter - intuitive result since it implies that atoms initially prepared in the momentum trapping regions are the ones which at long times , for @xmath165 , will absorb the most energy ( there are no further reversals of this behavior at even longer times ) . -kp ( @xmath166 , @xmath167 ) corresponding to the three classical diffusion regimes investigated in the experiments in @xcite.,width=288 ] the classical analysis thus reveals that there are three distinct classical diffusive regimes occurring at three timescales . we can expect that the corresponding _ quantum _ behavior will depend on which regime is dominant when dynamical localization arrests the quantum momentum diffusion . to investigate this , we now investigate how the form of the floquet states varies in these different regimes . in fig.7(a ) we show the localization lengths of the floquet states for a weak kicking - strength . it can clearly be seen that the localization varies periodically as a function of momentum , staying within the range @xmath168 for the majority of points , with the exception of a series of sharp cusp - like features at which the localization dramatically falls . these location of these cusps exactly corresponds to the the `` trapping momenta '' , @xmath169 , predicted from classical arguments . the floquet states centered in the trapping regions have widths of @xmath170 , much narrower than states localized on stable islands , which are also visible in this figure as regular strings of points at @xmath171 . at the experimental values of @xmath47 and @xmath172 a similar behavior is produced , with the broadest floquet states having localization lengths of @xmath173 , while the narrowest have widths of @xmath174 , over one thousand times narrower . the floquet states at the tips of the cusps have such low localization lengths that they are effectively pure plane - wave states ( this can be further corroborated by evaluating the inverse participation ratio for these states , which indeed takes a value of almost unity ) . it is thus unsurprising that the presence of these states corresponds to the classical trapping effect , as a quantum system prepared in such a state will have a vanishingly small overlap with any other state and so will remain frozen ( or trapped ) in its initial state . it is important to note , however , that this quantum trapping effect depends critically on the _ order _ of the two kick- periods that is , whether the system is driven with a short - long kick - sequence or the inverse long - short ordering . this may appear surprising at first , since the floquet states are periodic , with the same period @xmath175 as the driving , and this period is not altered by interchanging the order of the kicks . although it is frequently neglected , however , it is important to recall that the floquet states do have an explicit time - dependence within each period , and this _ is _ able to produce substantially different behavior @xcite when the phase of the driving field is altered . to illustrate this , we show in fig.[floq ] the time - evolution of one of the localized floquet states , which experiences @xmath0-kicks at times @xmath176 and @xmath177 . as can be seen , the state has only a trivial time - evolution during the first time - interval ( @xmath178 ) , since it is almost a plane wave and is thus approximately an eigenstate of the free hamiltonian . the first kick at @xmath79 causes the wave - packet to spread considerably in momentum space , before the second kick restores this broadened state to its original narrow form . thus in this brief window of time between the two kicks , even the most localized floquet states have a considerable spread in momentum . as a consequence , if the phase of the kicking field is shifted so that the system experiences the short - long kick - sequence , none of the floquet states are sharply localized in momentum at @xmath82 . we show in fig.[evolve ] the time - evolution of the system s kinetic energy when it is prepared in a momentum eigenstate in a trapping region . for the long - short kick - sequence this state projects onto essentially a single floquet state at @xmath82 , and so its time evolution is trivial and its energy remains constant . for the case of the short - long kick - sequence , however , the initial state projects onto a number of floquet states ( eq.[floquet ] ) , giving rise to a complicated quasi - periodic behavior arising from beating between the different quasienergies . , for physical parameters : @xmath179 , @xmath180 and @xmath167 . initially the floquet state is sharply peaked at @xmath181 , in the center of a trapping region . the first kick at @xmath182 causes the state to spread across a much broader range of momentum , until the second kick at @xmath183 restores the localized state.,width=288 ] -kp , for the same physical parameters as in fig.7a . the solid line shows the evolution of the system under the long - short kicking sequence , and shows little features . the dotted line shows the result of the short - long sequence , and exhibits a complicated quasi - periodic behavior.,width=288 ] in fig.7(b ) , we see the effect of increasing the kick - strength . in this regime we find there is an almost constant localization length for momenta in between the trapping regions , which are again signalled by sharp cusp - like structures . this indicates that the floquet states are confined between the classical broken phase barriers in the trapping regions . early studies indicate that the level statistics of the corresponding quasienergies are not pure poisson in this regime , as would be the case for the standard qkp . in fig 7(c ) , we see an inversion of the broad momentum modulation in fig 7(a ) , similar to the reversal seen in the experiment . in this regime , the eigenstates localized in the trapping regions near @xmath184 are typically _ broader _ than those localized in between . we have presented a study of the floquet states of @xmath0-kicked particles pulsed with unequal periods , and used them to analyze experimental data on these systems . we conclude that the chaotic ratchet effect proposed in @xcite and observed experimentally in @xcite is associated with asymmetric floquet states localized around @xmath133 . we conclude also that the behavior of the localization lengths of the floquet states for the double @xmath0-kicked rotor broadly accompany the change over between the three distinct classical diffusion regimes investigated experimentally in @xcite .	recently , cesium atoms in optical lattices subjected to cycles of unequally - spaced pulses have been found to show interesting behavior : they represent the first experimental demonstration of a hamiltonian ratchet mechanism , and they show strong variability of the dynamical localization lengths as a function of initial momentum . the behavior differs qualitatively from corresponding atomic systems pulsed with equal periods , which are a textbook implementation of a well - studied quantum chaos paradigm , the quantum @xmath0-kicked particle ( @xmath0-qkp ) . we investigate here the properties of the corresponding eigenstates ( floquet states ) in the parameter regime of the new experiments and compare them with those of the eigenstates of the @xmath0-qkp at similar kicking strengths . we show that , with the properties of the floquet states , we can shed light on the form of the observed ratchet current as well as variations in the dynamical localization length .
the exploration of symmetries and their breaking have always been an active and interesting area of research in nuclear physics . one of the well known examples , that can be cited here , is the nuclear @xmath3 decay which violates parity that led to the discovery of the weak interaction . our present concern , however , is the strong interaction where , in particular , we focus attention to the charge symmetry violation ( csv ) in nucleon - nucleon ( @xmath4 ) interaction . charge symmetry implies invariance of the @xmath4 interaction under rotation in isospin space , which in nature , is violated . the csv , at the fundamental level is caused by the finite mass difference between up @xmath5 and down @xmath6 quarks @xcite . as a consequence , at the hadronic level , charge symmetry ( cs ) is violated due to non - degenerate mass of hadrons of the same isospin multiplet . the general goal of the research in this area is to find small but observable effects of csv which might provide significant insight into the strong interaction dynamics . there are several experimental data which indicate csv in @xmath4 interaction . for instance , the difference between @xmath7 and @xmath8 scattering lengths at @xmath9s@xmath10 state is non - zero @xcite . other convincing evidence of csv comes from the binding energy difference of mirror nuclei which is known as okamoto - nolen - schifer ( ons ) anomaly @xcite . the modern manifestation of csv includes difference of neutron - proton form factors , hadronic correction to @xmath11 @xcite , the observation of the decay of @xmath12 etc @xcite . in nuclear physics , one constructs csv potential to see its consequences on various observables . the construction of csv potential involves evaluation of the @xmath4 scattering diagrams with intermediate states that include mixing of various isospin states like @xmath1-@xmath2 or @xmath13-@xmath14 mesons . the former is found to be most dominant @xcite which we consider here . most of the calculations performed initially to construct csv potential considered the on - shell @xcite or constant @xmath1-@xmath2 mixing amplitude @xcite , which are claimed to be successful in explaining various csv observables @xcite . this success has been called into question @xcite on the ground of the use of on - shell mixing amplitude for the construction of csv potential . first in @xcite and then in @xcite , it is shown that the @xmath1-@xmath2 mixing has strong momentum dependence which even changes its sign as one moves away from the @xmath1 ( or @xmath2 ) pole to the space - like region which is relevant for the construction of the csv potential . therefore inclusion of off - shell corrections are necessary for the calculation of csv potential . we here deal with such mixing amplitude induced by the @xmath15-@xmath15 loop incorporating off - shell corrections . in vacuum , the charge symmetry is broken explicitly due to the non - degenerate nucleon masses . in matter , there can be another source of symmetry breaking if the ground state contains unequal number of neutrons ( @xmath16 ) and protons ( @xmath17 ) giving rise to ground state induced mixing of various charged states like @xmath1-@xmath2 meson even in the limit @xmath18 . this additional source of symmetry breaking for the construction of csv potential has , to the best of our knowledge , not been considered before . the possibility of such matter induced mixing was first studied in @xcite and was subsequently studied in @xcite . for the case of @xmath13-@xmath14 meson also such asymmetry driven mixing is studied in @xcite . but none of these deal with the construction of two - body potential and the calculations are mostly confined to the time - like region where the main motivation is to investigate the role of such matter induced mixing on the dilepton spectrum observed in heavy ion collisions , pion form factor , meson dispersion relations etc . @xcite . in ref.@xcite , attempt has been made to calculate the density dependent csv potential where only the effect of the scalar mean field on the nucleon mass is considered excluding the possibility of matter driven mixing . all existing matter induced mixing calculations , however , suggest that , at least in the @xmath1-@xmath2 sector , the inclusion of such a matter induced mixing amplitude into the two body @xmath4 interaction potential can significantly change the results both qualitatively and quantitatively . it is also to be noted that such mixing amplitudes , in asymmetric nuclear matter ( anm ) , have non - zero contribution even if the quark or nucleon masses are taken to be equal @xcite . we consider both of these mechanisms to construct the csv potential . physically , in dense system , intermediate mesons might be absorbed and re - emitted from the fermi spheres . in symmetric nuclear matter ( snm ) the emission and absorption involving different isospin states like @xmath1 and @xmath2 cancel when the contributions of both the proton and neutron fermi spheres are added provided the nucleon masses are taken to be equal . in anm , on the other hand , the unbalanced contributions coming from the scattering of neutron and proton fermi spheres , lead to the mixing which depends both on the density @xmath19 and the asymmetry parameter @xmath20 $ ] . inclusion of this process is depicted by the second diagram in fig.[fig00 ] represented by @xmath21 which is non - zero even in symmetric nuclear matter if explicit mass differences of nucleons are retained . in the first diagram , @xmath22 involves nn loop denoted by the circle . the other important element which we include here is the contribution coming from the external legs . this is another source of explicit symmetry violation which significantly modify the csv potential in vacuum as has been shown only recently by the present authors @xcite . this paper is organized as follows . in sec.ii we present the formalism where the three momentum dependent @xmath23-@xmath2 mixing amplitude is calculated to construct the csv potential in matter . the numerical results are discussed in sec.iii . finally , we summarize in sec.iv . we start with the following effective lagrangians to describe @xmath24 and @xmath25 interactions : _ nn & = & g_\|_^_ , [ lag : omega ] + _ nn & = & g _ ^ _ , [ lag : rho ] where @xmath26 is the ratio of vector to tensor coupling , @xmath27 is the average nucleon mass and @xmath28 is the isospin operator . @xmath29 and @xmath30 represent the nucleon and meson fields , respectively , and @xmath31 s stand for the meson - nucleon coupling constants . the tensor coupling of @xmath2 is not included in the present calculation as it is negligible compared to the vector coupling . the matrix element , which is required for the construction of csv @xmath4 potential is obtained from the relevant feynman diagram @xcite : ^nn_(q ) & = & [ \|u_n(p_3)_^(q ) u_n(p_1 ) ] _ ^(q ) ^_(q^2 ) + & & _ ^(q ) [ \|u_n(p_4)_^(-q ) u_n(p_2 ) ] . [ ma0 ] in the limit @xmath32 , eq.([ma0 ] ) gives the momentum space csv @xmath4 potential , @xmath33 . here @xmath34 , @xmath35 $ ] denote the vertex factors , @xmath36 is the dirac spinor and @xmath37 , @xmath38 is the meson propagator . @xmath39 and @xmath40 are the four momenta of nucleon and meson , respectively . in the present calculation , @xmath1-@xmath2 mixing amplitude ( _ i.e. _ polarization tensor ) @xmath41 is generated by the difference between proton and neutron loop contributions : ^_(q^2 ) = ^(p)_(q^2 ) -^(n)_(q^2 ) . [ se ] explicitly , the polarization tensor is given by & & i^(n)_(q^2 ) = , + & & [ pol : total0 ] where @xmath42 denotes the four momentum of the nucleon in the loops ( _ see _ fig.[fig00 ] ) and @xmath43 is the in - medium nucleon propagator consisting of free @xmath44 and density dependent @xmath45 parts @xcite , [ nucl : prop ] g^f_n(k)&= & , [ nucl : prop_vac ] + g^d_n(k)&=&(k/+m_n)(k_0-e_n)(k_n-\|k\| ) [ nucl : prop_med ] the subscript @xmath15 stands for nucleon index ( _ i.e. _ @xmath46 or @xmath16 ) , @xmath47 denotes the fermi momentum of nucleon , nucleon energy @xmath48 and nucleon mass is denoted by @xmath49 . @xmath50 is the fermi distribution function at zero temperature . the origin of @xmath51 , in addition to the free propagator resides in the fact that here one deals with vacuum containing real particles which when acted upon the annihilation operator does not vanish ( _ see _ * appendix a * for details ) . the appearance of the delta function in eq.([nucl : prop_med ] ) indicates the nucleons are on - shell while @xmath50 ensures that propagating nucleons have momentum less than @xmath47 . likewise , the polarization tensor of eq.([pol : total0 ] ) also contains a vacuum @xmath52 $ ] and a density dependent @xmath53 $ ] parts as shown in fig.[fig00 ] . it is to be noted that the density dependent part given by the combination of @xmath54 corresponds to scattering that we have discussed already , whereas the term proportional to @xmath55 vanishes for low energy excitation @xcite . the vacuum part , _ viz . _ @xmath52 $ ] on the other hand involves @xmath56 which gives rise to usual csv part of the potential due to the splitting of the neutron and proton mass . it might be worthwhile to mention here that eq.([nucl : prop_med ] ) can induce charge symmetry breaking in asymmetric nuclear matter due to the appearance of the fermi distribution function in the propagator itself which can distinguish between neutron and proton , even if their mass are taken to be degenerate . evidently , this is an exclusive medium driven effect where , as mentioned already in the introduction , the charge symmetry is broken by the ground state . the total charge symmetry breaking would involve both the contributions where , it is clear that even for @xmath57 , the medium dependent term can contribute if the non - degenerate nucleon masses are considered . note that the polarization tensor @xmath41 can be expressed as the sum of longitudinal component @xmath58 $ ] and transverse component @xmath59 $ ] which will be useful to simplify the matrix element given in eq.([ma0 ] ) . ^_(q^2 ) = ^l_(q^2 ) a^ + ^t_(q^2 ) b^ , where @xmath60 and @xmath61 are the longitudinal and transverse projection operators @xcite . we define @xmath62 and @xmath63 . we , in the present calculation , use the average of longitudinal and transverse components of the polarization tensor instead of @xmath64 and @xmath65 . the average mixing amplitude is denoted by & = & \|_vac(q^2 ) + \|_med(q^2 ) . [ pol : av0 ] in the last line of eq.([pol : av0 ] ) , @xmath66 and @xmath67 denote the average mixing amplitudes of vacuum and density dependent parts , respectively . to obtain @xmath66 and @xmath67 one would calculate the total polarization tensor given in eq.([pol : total0 ] ) . after evaluating the trace of eq.([pol : total0 ] ) , we find the following vacuum and density dependent parts of the polarization tensor . ^(n)_vac(q^2 ) & = & q^ , [ pol : vac0 ] and ^\{vv(n)}_med(q^2 ) & = & 16 ( k_n-\|k\| ) + & & , [ pol : dens_vv0 ] + ^\{tv(n)}_med(q^2 ) & = & 4 c _ ( k_n-\|k\| ) + & & , [ pol : dens_tv0 ] where @xmath68 and @xmath69 . it is to be mentioned that both @xmath70 and @xmath71 obey the current conservation as @xmath72 and @xmath73 . the superscripts @xmath74 and @xmath75 in eqs.([pol : vac0 ] ) -([pol : dens_tv0 ] ) indicate the vector - vector and tensor - vector interactions , respectively . the dimensional counting shows that vacuum part of the polarization tensor [ both @xmath76 and @xmath77 is ultraviolet divergent and dimensional regularization @xcite is used to isolate the divergent parts . since the mixing amplitude is generated by the difference between the proton and neutron loop contributions , the divergent parts cancel out yielding the vacuum amplitude finite . ^_vac(q^2 ) & = & ^(p)_vac(q^2)-^(n)_vac(q^2 ) + & = & q^2q^^1_0 dx + & & ( ) . [ pol : vac1 ] eq.([pol : vac1 ] ) shows the four - momentum dependent vacuum polarization tensor . from the above equation one can calculate longitudinal @xmath78 and transverse @xmath79 components of the vacuum mixing amplitude and in the limit @xmath32 , @xmath80 . therefore , the average vacuum mixing amplitude is & = & - ( 2 + 3c_)()q^2 + & & - ^2 . [ pol : vac3 ] eq.([pol : vac3 ] ) represents the three momentum dependent vacuum mixing amplitude . this mixing amplitude vanishes for @xmath81 and then no csv potential in vacuum will exist . to calculate density dependent mixing amplitude from eq.([pol : dens_vv0 ] ) and ( [ pol : dens_tv0 ] ) we consider @xmath82 . in the limit @xmath32 , one finds following expressions : ^00(n)_med(q^2 ) & = & - [ pol : dens_00 ] + & & + ^11(n)_med(q^2 ) & = & [ pol : dens_11 ] note that the terms within the first curly braces of both eq.([pol : dens_00 ] ) and eq.([pol : dens_11 ] ) arise from the vector - vector interaction while the terms within the second curly braces arise from tensor - vector interaction of the density dependent polarization tensor . the @xmath83 component of the density dependent polarization tensor vanishes _ i.e. _ @xmath84 . now where ^l_med(q^2 ) & = & - , + ^t_med(q^2 ) & = & - . with the suitable expansion of eqs.([pol : dens_00 ] ) and ( [ pol : dens_11 ] ) in terms of @xmath85 and keeping @xmath86 terms we get where ^ & = & . [ dprm ] + ^ & = & , + & & [ aprm ] clearly , @xmath87 is also three momentum dependent and if @xmath18 it vanishes in snm but is non - vanishing in anm . in the present calculation nucleon masses are considered non - degenerate and the asymmetry parameter @xmath88 . to construct csv potential we take non - relativistic ( nr ) limit of the dirac spinors in which case we obtain u_n ( ) ( 1 - - ) ( c 1 + + ) , [ sp ] where @xmath89 is the spin of nucleon . @xmath90 denotes the average three momentum of the interacting nucleon pair and @xmath91 stands for the three momentum of the meson . the explicit expression of full csv potential in momentum space can be obtained from eq.(13 ) of ref.@xcite by replacing the mixing amplitude @xmath92 with @xmath93 . v^nn_csv(q)&=&- + & & . [ pot : mom_full ] eq.([pot : mom_full ] ) presents the full csv @xmath4 potential in momentum space in matter . here @xmath94 and @xmath95 is the total spin of the interacting nucleon pair . we define @xmath96 , @xmath97 and @xmath98 . it is be mentioned that the spin dependent parts of the potential appear because of the contribution of the external nucleon legs shown in fig.[fig00 ] . on the other hand , @xmath99 and @xmath100 arise due to expansion of the relativistic energy @xmath101 of the spinors . in matter , @xmath33 , consists of two parts , one contains the vacuum mixing amplitude and other contains the density dependent mixing amplitude . the former is denoted by @xmath102 and later by @xmath103 . thus , @xmath104 . from eq.([pot : mom_full ] ) we extract the following term which , in coordinate space gives rise to the @xmath105-function potential . v^nn_csv & = & ( + ^ ) + & & . [ pot : contact ] to avoid the appearance of @xmath105-function potential in coordinate space , one should introduce form factors [ @xmath106 , for which the meson - nucleon coupling constants become momentum dependent _ i.e. _ @xmath107 . here we use the following form factor . f_i(q^2 ) = ( ) , [ form ] where @xmath108 is the cut - off parameter governing the range of the suppression and @xmath109 denotes the mass of exchanged meson . the full csv potential presented in eq.([pot : mom_full ] ) contains both class @xmath0 and class @xmath110 potentials , and both break the charge symmetry in @xmath4 interactions . the terms within the first and the second curly braces represent class @xmath111 and class @xmath110 potentials , respectively . class @xmath111 potential differentiates between @xmath8 and @xmath7 systems while class @xmath112 @xmath4 potential exists in the @xmath113 system only . we , in this paper , restrict ourselves to class @xmath111 potential only . the coordinate space potential can be easily obtained by fourier transformation of @xmath33 _ i.e. _ v^nn_csv(r)= v^nn_csv(q ) e^-iqr we drop the term @xmath99 from eq.([pot : mom_full ] ) while taking the fourier transform as it is not important in the present context . however , this term becomes important to fit @xmath114 and @xmath115 phase shifts simultaneously . now the csv potential in coordinate space without @xmath116 reduces to v^nn_vac(r ) & = & - t^+_3 , [ pot : vac_cord ] and v^nn_med(r)&= & - t^+_3 , [ pot : dens_cord ] where @xmath117 . the explicit expressions of @xmath118 and @xmath119 are given in ref.@xcite . in eqs.([pot : vac_cord ] ) and ( [ pot : dens_cord ] ) , the @xmath120 independent terms represent central parts without contributions of external nucleon legs . the potentials presented in eqs.([pot : vac_cord ] ) and ( [ pot : dens_cord ] ) do not include the form factors so that these potentials diverge near the core . the problem of divergence near the core can be removed by incorporating form factors as discussed before . with the inclusion of form factors , @xmath121 and @xmath122 take the following form : v^nn_vac(r ) & = & - t^+_3 , [ pot : vac_ff ] and v^nn_med(r ) & = & -t^+_3 . [ pot : dens_ff ] in eqs.([pot : vac_ff ] ) and ( [ pot : dens_ff ] ) , @xmath123 and & = & ( ) , [ lambda ] + a_i = ( ) , & & b_i = ( ) , [ ab ] + where i(j ) = , &&(i j ) . note that eqs.([pot : vac_ff ] ) and ( [ pot : dens_ff ] ) contain the contribution of @xmath116 and the problem of divergence near the core is removed . using eqs.([pot : vac_cord ] ) and ( [ pot : dens_cord ] ) we show the difference of csv potentials between @xmath8 and @xmath7 systems _ i.e. _ @xmath124 in fig.[fig01 ] for @xmath9s@xmath10 state at nuclear matter density @xmath125 @xmath126 with asymmetry parameter @xmath127 . the dashed and dotted curves show @xmath128 and @xmath129 , respectively . the total contribution ( _ i.e. _ @xmath130 ) is shown by the solid curve . it is observed that the density dependent csv potential can not be neglected while estimating csv observables such as binding energy difference of mirror nuclei . + fig.[fig02 ] displays @xmath131 for @xmath9s@xmath10 state including the fourier transform of @xmath132 and form factors . @xmath128 and @xmath129 are represented by dashed and dotted curves . solid curve shows the sum of these two contributions . note that incorporation of form factors remove the problem of divergence near the core . it is observed that the csv @xmath4 potential changes its sign due to inclusion of the fourier transform of @xmath132 . in this work we have constructed the csv @xmath4 potential in dense matter using asymmetry driven momentum dependent @xmath23-@xmath2 mixing amplitude within the framework of one - boson exchange model . furthermore , the correction to the central part of the csv potential due to external nucleon legs are also considered . the closed - form analytic expressions both for vacuum and density dependent csv @xmath4 potentials in coordinate space are presented . we have shown that the vacuum mixing amplitude and the density dependent mixing amplitude are of similar order of magnitude and both contribute with the same sign to the csv potential . the contribution of density dependent csv potential is not negligible in comparison to the vacuum csv potential . note that the time - ordered product in eq.([app02 ] ) involves negative sign for fermions . the fermion field contains both the positive- and negative - energy solutions . the modal expansion for the fermion fields are , here @xmath135 and @xmath136 are the creation and annihilation operators for particles and likewise @xmath137 and @xmath138 are the creation and annihilation operators for antiparticles . the only nonvanishing anticommutation relations are in eq.([app13 ] ) @xmath144 has been replaced by @xmath101 . the first term of eq.([app13 ] ) represents particle propagation above the fermi sea and the second term indicates the propagation of holes inside the fermi sea . the last term shows the propagation of holes in the infinite dirac sea .	we construct density dependent class @xmath0 charge symmetry violating ( csv ) potential due to mixing of @xmath1-@xmath2 meson with off - shell corrections . here in addition to the usual vacuum contribution , the matter induced mixing of @xmath1-@xmath2 is also included . it is observed that the contribution of density dependent csv potential is comparable to that of the vacuum contribution .
we begin this section by fixing the notation . let @xmath3 be monic polynomials of degree @xmath4 in @xmath5 and orthogonal , with respect to a weight , @xmath6 $ ] , @xmath7 where @xmath8 is the square of the weighted @xmath9 norm of @xmath10 also , @xmath11 for convenience we set @xmath12 . the recurrence relation follows from the orthogonality condition : @xmath13 where @xmath14 , the @xmath15 are real and @xmath16 are strictly positive . in this paper we describe a formalism which will facilitate the determination of the recurrence coefficients for polynomials with singular weights . two points of view lead to this problem : on one hand the x - ray problem @xcite of condensed matter theory , on the other hand related problems in random matrix theory which involve the asymptotics of the fredholm determinant of finite convolution operators with discontinuous symbols @xcite . this paper is the first in a series that systematically study orthogonal polynomial where the otherwise smooth weights have been singularly deformed . the ultimate aim is the computation for large @xmath4 of the determinant @xmath17 of the @xmath18 moments or hankel matrix @xmath19 with moments @xmath20 where @xmath21 , thereby doing what has been done for the determinants of @xmath18 toeplitz matrices with singular generating functions @xcite . the deformed weight with one jump is @xmath22 where @xmath23 is the position of the jump , @xmath24 is the heaviside step function and the real @xmath25 parametrises the height of the jump . more generally , we take @xmath26 to be the canonical jump function @xmath27 and @xmath28 . the actions of the ladder operators on @xmath29 and @xmath30 are @xmath31 @xmath32 where @xmath33 . if @xmath34 and @xmath35 are non - vanishing one must add @xmath36 to and respectively . now @xmath0 and @xmath1 , the coefficient functions in the ladder operators , satisfy identities analogous to those found for smooth weights @xcite : @xmath37 the derivation of - will be published in a forthcoming paper where the weight has several jumps and @xmath38 $ ] is the jacobi weight . multiplying the recurrence relation evaluated at @xmath39 by @xmath40 and noting as well as we arrive at the universal equality @xmath41 similarly , squaring @xmath42 we find a second universal equation @xmath43 note that in the expressions for @xmath0 and @xmath1 only @xmath44 , the `` potential '' associated with the smooth reference weight , appears . the discontinuities give rise to @xmath45 and @xmath46 it is clear from and that if @xmath47 is rational , then @xmath0 and @xmath1 are also rational . this is particularly useful for our purpose which is the determination of the recurrence coefficients , for in this situation by comparing residues on both sides of and we should find the required difference equations @xcite . in the following section the above approach is exemplified by the hermite weight , @xmath48 and @xmath49 given by . it turns out that in this situation @xmath50 and @xmath51 are related to @xmath52 and @xmath53 in a very simple way . now , @xmath48 , so that @xmath54 , and @xmath55 as in . also , @xmath56 which are independent of the particular choice of @xmath2 and @xmath57 particular to @xmath58 note that @xmath59 is the value of @xmath60 at @xmath61 instead of proceeding with the full machinery of and we take advantage of the fact that @xmath54 . from orthogonality and the recurrence relation , we have @xmath62 by integration by parts . the string equation , @xmath63 is an immediate consequence of the orthogonality condition . again , an integration by parts and noting that @xmath64 produces @xmath65 it should be pointed out here that in general neither the string equation nor will provide the complete set of difference equations for the recurrence coefficients which can be seen if @xmath2 were the jacobi weight . in such a situation the compatibility conditions and must be used . now and become @xmath66 and @xmath67 equations and , supplemented by the initial conditions @xmath68^{-1 } \quad \textrm{and } \quad r_0({{\tilde x } } ) = 0 \ , \ ] ] can be iterated to determine the recurrence coefficients numerically . also , explicit solutions to and can be produced for small @xmath4 . if and are combined with the evolution equations to be derived in this section , the painlev iv mentioned in the abstract is found . we begin with the @xmath9 norm @xmath70 , , which entails @xmath71 and thus @xmath72 since @xmath73 . with , @xmath74 which is the first toda equation . taking the derivative with respect to @xmath69 of at @xmath75 and using the definition of the monic polynomials then gives @xmath76 since @xmath77 is an immediate consequence of the recurrence relation . therefore @xmath78 the second toda equation . eliminating @xmath79 from and the second toda equation , gives @xmath53 in terms of @xmath50 and @xmath80 : @xmath81 using the first toda equation to express @xmath82 in terms of @xmath50 and @xmath83 and substituting into produces a particular painlev iv @xcite , @xmath84 which can be brought into the canonical form with the replacements @xmath85 and @xmath86 . is supplemented by the boundary conditions @xmath87 . in a recent paper @xcite , a painlev iv was derived for the discontinuous hermite weight using an entirely different method . based on and the derivative of the logarithm of the hankel determinant @xmath88 can be computed as @xmath89 where has been used in the first line , which can be summed by the christoffel - darboux formula , @xmath90 in the limit @xmath91 we find , in general , @xmath92 using the the ladder operators and . with this entails @xmath93 the apparent pole at @xmath94 can be shown to have vanishing residue by considering @xmath95 : @xmath96 where the last equality is due to . a further regular term can be found as a contribution from the taylor series of @xmath97 about @xmath94 namely , @xmath98 where @xmath99 . using the fact that @xmath64 and , we see that the first two terms of combined into @xmath100 cancel the third . we are therefore left with the regular term : @xmath101 using it follows from @xmath102 that reproduces the second derivative , @xmath103 . let @xmath104 be the free energy . expressing in terms of @xmath105 and finally in terms of @xmath106 , where @xmath107 , the free energy reads @xmath108 where @xmath109 is the free energy corresponding to @xmath110 . note that @xmath111 , which gives rise to the sum rule @xmath112 with a minor change of variables becomes the toda molecule equation . first we note that @xmath113 defining @xmath114 it then follows @xmath115 we may express @xmath105 in terms of the derivatives of the free energy , by noting that @xmath116 and @xmath117 . one finds with @xmath118 where @xmath119 . for @xmath120 we find the asymptotic expansion @xmath121 \nonumber \\ \fl r_n(0 ) & = & -b ( -1)^n \cos ( 2 b \ln n + b ) - \frac{b^2}{2n } \big ( 1+\sin^2(2b\ln n+b ) \big ) + \mathcal{o}(n^{-2 } ) { \label{eq : asymptote_rn}}\end{aligned}\ ] ] guided by the numerics on the difference equations and . the constant @xmath122 is given by @xmath123 and @xmath124 is a phase independent of @xmath4 . unfortunately , the formalism developed in this paper does not seem to shed any light on its delicate dependence on @xmath25 . it can be verified by a direct calculation that and satisfy and to order @xmath125 . the top panel of shows a comparison between the numerical results and the above asymptotes for suitably rescaled @xmath50 . in principle , @xmath50 and @xmath53 can be determined analytically to any order in @xmath4 because an approximation of @xmath50 to order @xmath126 gives rise to a difference equation for @xmath53 to order @xmath127 via . in turn , @xmath53 to order @xmath127 produces an equation for @xmath50 to order @xmath127 and so forth . this scheme breaks down for @xmath128 . for fixed @xmath128 , the numerics does not suggest an ansatz for the asymptotes . most remarkably , the effect of @xmath128 persists for very large @xmath4 , even for very small @xmath128 , as illustrated in . also shown in this figure as dashed ( dotted ) lines are the approximations of @xmath105 from the first three ( two ) terms of a taylor - series in @xmath69 around @xmath120 based on the iterative results for @xmath129 , and . in principle , the painlev iv , , provides a way to express @xmath130 in terms of lower order derivatives @xmath131 with @xmath132 , yet the results in suggest that for sufficiently large @xmath4 a finite taylor series eventually deviates wildly from the correct @xmath105 . note that both @xmath133 and @xmath134 are bounded in @xmath4 for large @xmath4 . shows the rescaled @xmath50 for fixed @xmath4 and varying @xmath69 . it resembles a hermite polynomial because of its direct relation to @xmath135 , and . for the same reason @xmath105 vanishes for @xmath136 . xxx e. l. basor and y. chen , `` the x - ray problem revisited '' , j. phys . a. : math . gen . * 36 * ( 2003 ) l175-l180 ; `` a note on the wiener - hopf determinants and the borodin - okunkov identity '' , integr . oper . theory , * 45 * ( 2003 ) 301 - 308 . h. widom , `` toeplitz determinants with singular generating functions '' , amer . , * 95 * ( 1973 ) 333 - 383 ; e. l. basor , `` asymptotic formulas for toeplitz determinants '' , trans . * 239 * ( 1978 ) 33 - 65 . a. p. bassom , p. a. clarkson , a. c. hicks , and j. b. mcleod , `` integral equations and exact solutions of the fourth painlev equation '' , proc . london ser . a * 437 * ( 1992 ) 1 - 24 . y. chen and m. e. h. ismail , `` jacobi polynomials from compatibility conditions '' , proc . amer . math . * 133 * ( 2005 ) 465 - 472 , 225 - 237 . y. chen and m. e. h. ismail , `` ladder operators and differential equations for orthogonal polynomials '' , j. phys . a. : math . gen . * 30 * ( 1997 ) 7817 - 7829 . p. j. forrester and n. s. witte , `` discrete painlev equations and random matrix averages '' , nonlinearity , * 16 * ( 2003 ) 1919 - 1944 . m. e. h. ismail and j. wimp , `` on differential equations for orthogonal polynomials '' , methods appl . anal . * 5 * ( 1998 ) 439 - 452 . m. jimbo , t. miwa , y. mri and m. sato , `` density matrix of an impenetrable bose gas and the fifth painlev transcendent '' , physica d * 1 * ( 1980 ) 80158 .	in this paper we present a brief description of a ladder operator formalism applied to orthogonal polynomials with discontinuous weights . the two coefficient functions , @xmath0 and @xmath1 , appearing in the ladder operators satisfy the two fundamental compatibility conditions previously derived for smooth weights . if the weight is a product of an absolutely continuous reference weight @xmath2 and a standard jump function , then @xmath0 and @xmath1 have apparent simple poles at these jumps . we exemplify the approach by taking @xmath2 to be the hermite weight . for this simpler case we derive , without using the compatibility conditions , a pair of difference equations satisfied by the diagonal and off - diagonal recurrence coefficients for a fixed location of the jump . we also derive a pair of toda evolution equations for the recurrence coefficients which , when combined with the difference equations , yields a particular painlev iv .
in the real world , one single object may have different representations in different domains . for example , the declaration of independence has versions translated into different languages . let @xmath0 denote the number of objects @xmath1 , and @xmath2 be the number of domains . then we have @xmath3 where the @xmath4th object @xmath5 has @xmath2 measurements @xmath6 ; @xmath7 is the representation for object @xmath5 in space @xmath8 . the problem explored in this paper is that for @xmath9 new objects @xmath10 , how to classify their representations @xmath11 given the representations @xmath12 with @xmath13 . for this task , @xmath14 , described above are needed to learn the relation between @xmath8 and @xmath15 so that we can map data from @xmath8 and @xmath15 to a common space @xmath16 . thus @xmath17 are the domain relation learning training data . in our scenario , we are interested in a particular setting that the data to be classified is in separated classes different from the data used to learn the low dimensional manifold . this is shown in figure [ classification ] , where disks represent the domain relation learning training data @xmath17 and squares denote the classifier training and testing data @xmath18 . a classification rule @xmath19 is trained on @xmath20 and applied on @xmath21 . we consider one domain relation learning method , canonical correlation analysis ( cca ) @xcite , which can be carried out using reduced - rank regression routines @xcite . we investigate classification performance in the common space @xmath16 obtained via cca , training the classifier on @xmath20 and testing on @xmath21 . the focus of this paper is not on optimizing the classifier ; rather , we investigate performance for a given clasifier ( 5-nearest neighbor ) as a function of the number of domain relation learning training data observation @xmath0 used to learn @xmath16 . the main contribution of this paper is an investigation of the notion of supplementing the training data of classifier by using data from other disparate sources / spaces . the structure of the paper is as follows : section [ background ] talks about related work . section [ method ] discusses the methods employed , including the manifold matching framework as well as embedding and classification details . experimental setup and results are presented in section [ results ] . section [ conclusion ] is the conclusion . different methods of transfer learning , multitask learning and domain adaptation are discussed in a recent survey @xcite . there are algorithms developed on unsupervised document clustering where training and testing data are of different kinds @xcite . the problem explored in this paper can be viewed as a domain adaptation problem , for which the training and testing data of the classifier are from different domains . when the classification is on the text documents in different languages , as described in the later sections of this paper , it is called cross - language text classification . there is much work on inducing correspondences between different language pairs , including using bilingual dictionaries @xcite , latent semantic analysis ( lsa ) features @xcite , kernel canonical correlation analysis ( kcca ) @xcite , etc . machine translation is also involved in the cross - language text classification , which translates the documents into a single domain @xcite . in this paper , we focus on manifold matching . the whole procedure can be divided into the following steps : 1 . for each single space @xmath8 , calculate the dissimilarity matrix for all domain relation learning training data observations @xmath5 . 2 . for each @xmath22 , use multidimensional scaling ( mds ) on the dissimilarity matrix to get a euclidean representation @xmath23 . 3 . run cca ( for @xmath24 ) or generalized cca ( @xmath25 ) to map the collection @xmath26 to a common space @xmath16 . 4 . pursue joint inference ( i.e. classification ) in the common space @xmath16 . this procedure combines mds and ( generalized ) cca in a sequential way . firstly mds is applied to learn low - dimensional manifolds , then ( generalized ) cca is used to match those manifolds to obtain a common space . this paper focuses on manifold matching and it demonstrates the classification improvement via fusing data from additional space to learn the common low dimensional manifold . it is interesting to investigate how to generate the low dimensional space using all data instead of matching separate manifolds . but this requires calculating the dissimilarity information for the objects representation in different spaces properly for the multi - dimensional scaling purpose . this issue had been investigated , e.g. , @xcite , but there had not been any clear answer . the framework structure for manifold matching is shown in figure [ model ] @xcite . for each of the @xmath0 objects @xmath27 , there are @xmath2 representations @xmath28 generated by the mappings @xmath29 . manifold matching works to find @xmath30 to map @xmath31 to a low - dimensional common space @xmath32 : @xmath33 after learning the @xmath34s , we can map a new measurement @xmath35 into the common space @xmath32 via : @xmath36 this allows joint inference to proceed in @xmath37 . the work described in this paper is based on dissimilarity measures . let @xmath38 denote the dissimilarity measure in the @xmath22th space @xmath8 , and @xmath39 be the euclidean distance in the common space @xmath37 . there are two kinds of mapping errors induced by the @xmath34s : fidelity error and commensurability error . fidelity measures how well the original dissimilarities are preserved in the mapping @xmath40 , and the fidelity error is defined as the within - condition squared error : @xmath41 commensurability measures how well the matchedness is preserved in the mapping , and the commensurability error is defined as the between - condition squared error : @xmath42 multidimensional scaling ( mds ) @xcite works to get a euclidean representation while approximately preserving the dissimilarities . given the @xmath43 dissimilarity matrix @xmath44 $ ] in space @xmath8 , multidimensional scaling generates embeddings @xmath45 for @xmath46 , which attempts to optimize fidelity , that is , @xmath47 . for the @xmath24 case , multidimensional scaling generates @xmath48 matrices @xmath49 from @xmath50 and @xmath51 from @xmath52 . the @xmath4th row vector @xmath53 of @xmath54 is the multidimensional scaling embedding for @xmath55 . canonical correlation analysis is applied to the multidimensional scaling results . canonical correlation works to find @xmath56 matrices @xmath57 and @xmath58 as the linear mapping method to maximize correlation for the mappings into @xmath59 , where two matices satisfy @xmath60 and @xmath61 . that is , for the @xmath62th ( @xmath63 ) dimension , the mapping process is defined by @xmath64 and @xmath65 , the @xmath62th column vector of @xmath66 and @xmath67 respectively . the orthonormal requirement on the columns of @xmath66 ( similarly @xmath67 ) implies that the correlation between different dimensions of the embedding is 0 . the correlation of the mapping data is calculated as @xmath68 which is equivalent to @xmath69 subject to @xmath70 and the constraint can be proved to be equivalent to @xmath71 for cca it holds @xmath72 . for new data @xmath73 , out - of - sample embedding for multidimensional scaling @xcite generates @xmath74 dimensional row vector @xmath75 . the final embeddings in the common space @xmath37 are given by @xmath76 and @xmath77 . canonical correlation analysis optimizes commensurability without regard for fidelity @xcite . for our work , first we use multidimensional scaling to generate a fidelity - inspired euclidean representation , and then we use canonical correlation analysis to enforce low dimensional commensurability . canonical correlation analysis is developed as a way of measuring the correlation of two multivariate data sets , and it can be formulated as a generalized eigenvalue problem . the expansion of canonical correlation analysis to more than two multivariate data sets is also available @xcite , which is called generalized canonical correlation analysis ( gcca ) . generalized canonical correlation analysis simultaneously find @xmath78 to map the multivariate data sets in @xmath2 spaces to the common space @xmath59 . similarly for the new data @xmath79 , we can get their representations in the common space @xmath37 as @xmath80 . similar to cca , the correlation of data in the @xmath62th mapping dimension is calculated as @xcite @xmath81 subject to @xmath82 gcca can be formulated as a generalized eigenvalue problem . different algorithms have been developed as the solution , e.g. least square regression . for the particular dataset used in our experiments , because it is not very large , we can perform eigenvalue decomposition on the respective matrices directly . given the measurements of @xmath9 new data points @xmath83 , ( generalized ) canonical correlation analysis in section [ embedding ] yields the embeddings @xmath84 in the common space @xmath59 . to classify @xmath84 , instead of using data points from the same space @xmath8 ( i.e. @xmath85 ) , we consider the problem in which we must borrow the embeddings from another space @xmath15 for training , that is , @xmath86 . this problem is motivated by the fact that in many situations there is a lack of training data in the space where the testing data lie . we investigate the effect of the number of domain relation learning training data observations on the classification performance . our experiments apply canonical correlation analysis and its generalization to text document classification . the dataset is obtained from wikipedia , an open - source multilingual web - based encyclopedia with around 19 million articles in more than 280 languages . each document may have links pointing to other documents in the same language which explain certain terms in its content as well as the documents in other languages for the same subject . articles of the same subject in different languages are not necessarily the exact translations of one another . they can be written by different people and their contents can differ significantly . english articles within a 2-neighborhood of the english article `` algebraic geometry '' are collected . the corresponding french documents of those english ones are also collected . so this data set can be viewed as a two space case : @xmath87 is the english space and @xmath88 is the french space . there are in total 1382 documents in each space . that is , @xmath89 , and @xmath90 . note that @xmath91 includes both domain relation learning training data @xmath92 and new data points @xmath93 ( @xmath94 ) used for classification training and testing . all 1382 documents are manually labeled into 5 disjoint classes ( @xmath95 ) based on their topics . the topics are category , people , locations , date and math things respectively . there are 119 documents in class 0 , 372 documents in class 1 , 270 documents in class 2 , 191 documents in class 3 , and 430 documents in class 4 . the documents in classes @xmath96 are the domain relation learning training data @xmath97 . there are in total @xmath98 documents in those 3 classes ( @xmath99 ) . the @xmath100 ( @xmath101 ) documents in classes @xmath102 are the new data @xmath103 . they are used to train a classifier and run the classification test . the method described in section [ embedding ] starts with the dissimilarity matrix . for our work two different kinds of dissimilarity measures are considered : text content dissimilarity matrix @xmath104 and graph topology dissimilarity matrix @xmath105 . both matrices are of dimension @xmath106 , containing the dissimilarity information for all data points @xmath107 . graphs @xmath108 can be constructed to describe the dataset ; @xmath109 represents the set of vertices which are the 1382 wikipedia documents , and @xmath23 is the set of edges connecting those documents in language @xmath22 . the ( @xmath110 ) entry @xmath111 is the number of steps on the shortest path from document @xmath4 to document @xmath112 in @xmath113 . in the english space @xmath87 , @xmath114 , where the 4 comes from the 2-neighborhood document collection . in the french space @xmath88 , @xmath115 is the document in french corresponding to the document @xmath116 , and @xmath117 depends on the french graph connections . it is possible that @xmath118 . at the extreme end , @xmath119 when @xmath120 and @xmath121 are not connected . we set @xmath122 for @xmath123 . @xmath124 is based on the text processing features for documents @xmath125 . given the feature vectors @xmath126 , @xmath127 is calculated by the cosine dissimilarity @xmath128 . for our experiments , we consider three different features for @xmath129 : mutual information ( mi ) features @xcite , term frequency - inverse document frequency ( tfidf ) features @xcite and latent semantic indexing ( lsi ) features @xcite . the wikipedia dataset used in the experiments are available online . see the paper @xcite for more details / description . to choose the dimension @xmath130 for the common space @xmath37 , we pick a sufficiently large dimension and embed @xmath104 and @xmath105 via multidimensional scaling . the scree plot for the mds embedding is shown in fig [ eigplot_all ] ( term frequency - inverse document frequency features are used for the text dissimilarity calculation ) . based on the plots in figure [ eigplot_all ] , we choose @xmath131 for the dimension of the joint space @xmath16 , which is low but preserves most of the variance @xcite . this model selection choice of dimension is an important issue in its own right ; for this paper , we fix @xmath131 throughout . for the canonical correlation analysis step , since it requires to multidimensional scale the dissimilarity matrices to @xmath74 at the beginning , as described in section [ embedding ] , when we choose different number @xmath132 of domain relation learning training documents , @xmath74 depends on @xmath132 . the choice of dimension is once again an important model selection problem ; for this paper , the values of @xmath74 with different @xmath132 are shown in table [ ccadim ] . we believe that the values of @xmath74 are chosen large enough to preserve most of the structure yet still small enough to avoid dimensions of pure noise which might deteriorate the following ( g)cca step . the second column indicates what percentage of the total manifold matching training data @xmath55 is used . .mds dimensions [ cols="^,^,^",options="header " , ] inferences regarding differences in the relative performance between competing methodologies ( as well as the seemingly non - monotonic performance across @xmath133 for a given methodology ) are clouded by the variability inherent in our performance estimates . however , these real - data experimental results nonetheless illustrate the general relative performance characteristics of cca and gcca and their regularized versions , as a function of @xmath133 . canonical correlation analysis and its generalization are discussed in this paper as a manifold matching method . they can be viewed as reduced rank regression , and they are applied to a classification task on wikipedia documents . we show their performance with manifold matching training data from different domains and different dissimilarity measures , and we also investigate their efficiency by choosing different amounts of manifold matching training data . the experiment results indicate that the generalized canonical correlation analysis , which fuses data from disparate sources , improves the quality of manifold matching with regard to text document classification . also , if we use regularized canonical correlation analysis and its generalization , we further improve performance . finally , increasing the amount of domain relation learning training data from @xmath134 to @xmath135 ( @xmath133 in the figures [ pic_selfidm_stopwordsrm_stem_lsi_cca ] , [ pic_selfidm_stopwordsrm_stem_tfidf_lsidim_cca ] , [ pic_selfidm_stopwordsrm_stem_mi_lsidim_cca ] , [ pic_selfidm_stopwordsrm_stem_lsi_halflsidim_cca ] , [ pic_selfidm_stopwordsrm_stem_tfidf_halflsidim_cca ] and [ pic_selfidm_stopwordsrm_stem_mi_halflsidim_cca ] ) of the available 819 documents yield approximately @xmath134 improvement in classification performance . this improvement is independent of the amount of training data available for the classifier . s. t. dumais , t. a. letsche , m. l. littman and t. k. landauer . _ automatic cross - language retrieval using latent semantic indexing_. 1em plus 0.5em minus 0.4emin aaai symposium on cross language text and speech retrieval , 1997 . b. fortuna and j. shawe - taylor . _ the use of machine translation tools for cross - lingual text mining_. 1em plus 0.5em minus 0.4emin proceedings of the icml workshop on learning with multiple views , 2005 . d. r. hardoon , s. r. szedmak , and j. r. shawe - taylor . _ canonical correlation analysis : an overview with application to learning methods_.1em plus 0.5em minus 0.4emneural computation 16(12 ) : 2639 , 2004 . d. karakos , j. eisner , s. khudanpur and c. e. priebe . _ cross - instance tuning of unsupervised document clustering algorithms_. 1em plus 0.5em minus 0.4emproceedings of the main conference human language technology conference of the north american chapter of the association for computational linguistics , 2007 . d. lin and p. pantel . _ concept discovery from text_.1em plus 0.5em minus 0.4emin proceedings of the 19th international conference on computational linguistics , ( morristown , nj , usa ) , pp . 1 - 7 , association for computational linguistics , 2002 . z. ma , d. marchette and c. e. priebe . _ fusion and inference from multiple data sources in a commensurate space_. 1em plus 0.5em minus 0.4emstatistical analysis and data mining , accepted for publication , january , 2012 . c. e. priebe , d. j. marchette , y. park , e. j. wegman , j. l. solka , d. a. socolinsky , d. karakos , k. w. church , r. guglielmi , r. r. coifman , d. lin , d. m. healy , m. q. jacobs , and a. tsao . _ iterative denoising for cross - corpus discovery_.1em plus 0.5em minus 0.4emin proceedings of the 2004 symposium on computational statistics ( invited talk ) , prague , august 23 - 27 , 2004 . c. e. priebe , z. ma , d. marchette , e. hohman and g. coppersmith . _ fusion and inference from multiple data sources_.1em plus 0.5em minus 0.4emthe 57th session of the international statistical institute , durban , august 16 - 22 , 2009 . c. e. priebe , d. j. marchette , z. ma and s. adali . _ manifold matching : joint optimization of fidelity and commensurability_.1em plus 0.5em minus 0.4embrazilian journal of probability and statistics , accepted for publication , february , 2012 . j. via , i. santamaria and j. perez . _ canonical correlation analysis ( cca ) algorithms for multiple data sets : application to blind simo equalization_. 1em plus 0.5em minus 0.4em13th european signal processing conference , antalya , turkey , 2005 .	manifold matching works to identify embeddings of multiple disparate data spaces into the same low - dimensional space , where joint inference can be pursued . it is an enabling methodology for fusion and inference from multiple and massive disparate data sources . in this paper we focus on a method called canonical correlation analysis ( cca ) and its generalization generalized canonical correlation analysis ( gcca ) , which belong to the more general reduced rank regression ( rrr ) framework . we present an efficiency investigation of cca and gcca under different training conditions for a particular text document classification task . manifold matching , canonical correlation analysis , reduced rank regression , efficiency , classification .
whispering - gallery mode ( wgm ) microcavities are at the frontier of research on subjects ranging from biosensing , nonlinear optics , and laser physics , to fundamental physics such as cavity quantum electrodynamics@xcite . contrary to its rapid experimental advances , numerical exploration of wgm s has been largely lagging behind with a limited number of available options@xcite . on the other hand , the beam propagation method ( bpm ) has a long history@xcite in modelling light propagation along both straight and curved waveguides as well as whispering - gallery microcavity eigenmode analyses@xcite . compared to boundary element@xcite , finite element@xcite , finite - difference time - domain@xcite , and free space radiation mode methods@xcite , bpm remains highly efficient without sacrificing substantial accuracy . by adopting a perfectly matched layer ( pml)@xcite to absorb the light which is otherwise reflected at the computation window and following the procedure formulated in @xcite to correct inaccuracies incurred at the refractive index discontinuities in high refractive index contrast waveguide structures , the finite - difference beam propagation method ( fd - bpm ) can achieve high accuracy with a rapid convergence rate . conventional fd - bpm formulations are based on the fresnel approximation , where light is assumed to propagate close to the propagation axis@xcite . to overcome this limitation for bent waveguide modelling , high - order algorithms known as wide - angle bpm@xcite or the conformal mapping approach@xcite are desirable . alternatively , bpm may be reformulated in cylindrical coordinates systems to analyze such structures@xcite . in this work , we simulated the light propagation in a wgm microcavity by implementing fd - bpm in a cylindrical system as shown in fig . [ cyl_coor ] . the field perturbation from a nanobead attached to the microcavity and quality factor degradation arising from cavity deformations were investigated . the computed field distribution correctly includes the radiative field component , which a mode analysis technique would fail to simulate . from the helmholtz equation , the field component @xmath0 of the te wave in a two - dimensional whispering - gallery microcavity satisfies @xmath1 where @xmath2 is the refractive index of the cavity , @xmath3 is the wave number in free space , and @xmath4 is the vacuum wavelength of the light circulating in the cavity . for a perfect wgm cavity with azimuthal symmetry , the refractive index is independent of @xmath5 ( @xmath6 ) . the electric field can be approximated as propagating in the form of @xmath7 where @xmath8 is the normalized @xmath9 azimuthal modal field distribution such that the squared norm @xmath10 represents the circulating power of the mode and @xmath11 is a complex constant whose real part @xmath12 represents the azimuthal mode order when the cavity is in resonance with the circulating light and its imaginary part @xmath13 characterizes the attenuation of the field in the azimuthal direction . note that , in general , the real part @xmath12 can be any real number for a given wavelength @xmath4 . when a certain wavelength @xmath14 yields an integer value for @xmath12 , resonance occurs . in addition , multiple wavelengths may yield an identical integer @xmath12 where eigensolutions @xmath15 correspond to resonance whispering - gallery modes with the same azimuthal order @xmath12 but different transverse modes . both quantities can be obtained from the nonzero solution of the mode equation , in turn described as an eigenequation with eigenvalue @xmath16 derived from eq . : @xmath17 { \hat \psi}_m = m^2{\hat \psi}_m\ ] ] if the aforementioned symmetry is broken due to an azimuthal angle dependent perturbation of the refractive index @xmath18 where the perturbation @xmath19 , one may reformulate @xmath0 as @xmath20 where @xmath21 is a reference value such that @xmath22 varies slowly along the azimuthal direction or , equivalently , the slowly varying envelope approximation ( svea ) holds . this is mathematically written as latexmath:[\[\label{svea } it is necessary to point out that the choice of @xmath21 is arbitrary as long as svea holds ; however , if the wavelength of the light is close to the resonance wavelength of the @xmath24 order unperturbed wgm , it is convenient to select @xmath25 . we will therefore drop the bar in the rest of the text for convenience . alternatively , one may treat @xmath26 as a @xmath5-dependent quantity where @xmath27 is obtained from solving eq . at each angle @xmath5 for higher accuracy . from eq . , we obtain the wave evolution along the azimuthal direction according to @xmath28 discretizing the computation window uniformly so that the coordinates @xmath29 of each grid @xmath30 can be expressed as @xmath31 , @xmath32 , and @xmath33 , one can evolve the field at @xmath34 from a previous azimuthal angle @xmath35 according to @xmath36 where @xmath37 here @xmath38 and @xmath39 are grid spacings along the @xmath40 and @xmath41 directions , as illustrated in fig . [ cyl_coor ] . also , @xmath42 is the refractive index of the waveguide structure at each point . collecting @xmath43 into a ket form @xmath44 and rearranging eq . into a matrix form , we obtain @xmath45 where @xmath46 and @xmath47 are two tridiagonal matrices . by adopting standard fd - bpm procedures , one may obtain the field evolution via eq . from the excitation field at @xmath48 . to characterize the bpm , we first tested it on a perfect silica microring resonator immersed in water . the refractive index of the silica ring was @xmath49 @xcite at a wavelength of @xmath50 and the surrounding water had a refractive index of @xmath51 @xcite . the resonator had a @xmath52-@xmath53 major radius and a @xmath54-@xmath53 minor diameter . to simplify the analysis , we reduced the three - dimensional waveguide structure to a two - dimensional one through the use of an effective index method ( eim)@xcite along the @xmath55-direction . the cavity was excited by a modal field obtained from the mode solver . to minimize the spurious reflection at the edges of the computation window , a @xmath56-@xmath53 pml@xcite was placed at the edge of the computation window . the pml is implemented by replacing the radial derivative with @xmath57 in which @xmath58 is defined as @xmath59 where @xmath60 is the inner edge of the pml layer and @xmath61 is a constant . to optimize the pml performance and determine the optimal value for @xmath61 , a simple experimental simulation was conducted . an optical beam was launched towards a @xmath62-@xmath53 pml through a straight waveguide and the reflected power was measured for different values of @xmath61 . the different values of reflected power versus @xmath61 are presented in fig . [ pml_opt ] . the insets show the field intensity for three different @xmath61 values , wherein the white lines are the inner pml edges . the @xmath61 value is 0 in fig . [ pml_opt](a ) , @xmath63 ( i.e. the optimal value ) in fig . [ pml_opt](b ) , and @xmath64 in fig . [ pml_opt](c ) . the optimal value was then utilized for the remaining simulations . -@xmath53 pml vs. different @xmath61 values . the insets are the field intensity for three different @xmath61 values : ( a ) @xmath65 , ( b ) @xmath63 , and ( c ) @xmath64 . the white lines indicate the inner pml edges.,title="fig : " ] + in fig . [ field_plots ] , we plot the intensity distribution in logarithmic scale by setting a @xmath66-@xmath53 window size , @xmath67 grids in the @xmath68 , and @xmath69 grids in the @xmath70 for @xmath71 , where we excited the ring with its fundamental mode . the field distribution is plotted in fig . [ field_plots ] . the resonance wavelength and the mode number calculated by the mode solver are @xmath72 nm and @xmath73 nm , respectively . the solid blue lines in fig . [ field_plots ] are showing the resonator edges and the dashed blue lines define the pml boundary . as can be seen in this figure , a small portion of the energy radiates towards the computation window edge yet the adopted pml efficiently prevents the otherwise spurious reflection from returning to the resonator . we further computed the quality factor @xmath74 for different grid schemes according to @xmath75 . @xmath76 is the total power at an azimuthal angle @xmath5 . [ q_vs_grid](a ) is the plot of quality factor vs. radial and azimuthal grid spacings . the computation window is set to @xmath77 in the @xmath78 and @xmath79 in the @xmath70 . as is depicted , q converges from @xmath80 towards @xmath81 by reducing the grid spacing from @xmath82 to @xmath83 in the radial direction and @xmath84 to @xmath85 in the azimuthal direction . we also computed the relative error by adopting the q calculated by richardson extrapolation as a reference value . as seen in fig . [ q_vs_grid](b ) this was reduced to @xmath86 from @xmath87 , accordingly . furthermore , in fig . [ q_vs_grid](c ) we plot the q as well as the relative error as a function of azimuthal grid spacing @xmath39 by setting @xmath88 . in fig . [ q_vs_grid](d ) the same quantities are plotted as a function of radial grid spacing @xmath38 . the least square fits to the relative error curves indicate a convergence rate of @xmath89 in the azimuthal direction and @xmath90 in the radial direction . clearly , this suggests that faster convergence is attainable by adopting higher - order finite - difference schemes . .49 , and ( d ) variations for the cross section at @xmath91 . in ( c ) , the last two points are omitted for the line of best fit.,title="fig : " ] 0.49 , and ( d ) variations for the cross section at @xmath91 . in ( c ) , the last two points are omitted for the line of best fit.,title="fig : " ] 0.47 , and ( d ) variations for the cross section at @xmath91 . in ( c ) , the last two points are omitted for the line of best fit.,title="fig : " ] 0.49 , and ( d ) variations for the cross section at @xmath91 . in ( c ) , the last two points are omitted for the line of best fit.,title="fig : " ] in fig . [ l_res ] , we plot the resonance wavelength ( blue cross markers ) and corresponding relative error ( red cross markers ) vs. radial grid spacing @xmath38 by setting @xmath92 . here , the resonance wavelength @xmath93 is obtained from the round trip total phase shift @xmath94 of the electrical field computed from the wavelength @xmath4 adopted in the bpm according to @xmath95 , where the resonance wavelength @xmath96 nm calculated via richardson extrapolation is used as a reference . a least square fit to the relative error indicates a @xmath97 convergence rate . direction.,scaledwidth=55.0% ] to further demonstrate the validity of the formulation , we launched an arbitrary gaussian field at the input of the same structure as illustrated in the insets ( red curve ) of fig . [ overlap ] . for comparison , we also plotted the fundamental mode profile obtained by the mode solver and displayed it as the blue curve in the insets . as shown in the insets of fig . [ overlap ] , after propagating @xmath98 , @xmath99 , and @xmath100 rounds in the resonator from left to right , the circulating field distribution gradually evolves into the mode profile . to quantitatively analyze the field evolution , we define a normalized overlap factor @xmath101 . @xmath102 is the normalized mode profile and @xmath103 is the normalized output field profile after each round trip of propagation . the magnitude of the overlap factor and its departure from unity are respectively plotted in fig . [ overlap ] as blue and red curves . as is shown , the overlap factor reaches @xmath104 at the @xmath105 round trip . , 25@xmath106 , and 125@xmath106 rounds of propagation are plotted in insets from left to right.,title="fig:",scaledwidth=100.0% ] + in fig . [ p_built_up ] , we excited the cavity in continuous wave ( cw ) mode and plot the accumulated power normalized to the input power at the input of the cavity @xmath107 as a function of the number of rounds light circulated in the cavity . as illustrated , @xmath107 saturates to @xmath108 ( blue curve ) when resonance occurs . the saturation power is in excellent agreement with the theoretic prediction of @xmath109 that can be obtained from a mode solver with a relative error around @xmath110 after circulating more than @xmath111 rounds , as indicated in the red curve . we further plotted the total power when the input light wavelength had reached the full wave at half maximum point @xmath112 . as expected , the power saturated to half of that corresponding to resonance . ( green ) , and the resonance power s relative error ( red).,scaledwidth=70.0% ] next , we applied the bpm to whispering - gallery mode nanodetection modelling@xcite . here we simulated the induced resonance wavelength shift caused by single polystyrene beads binding to the surface of a silica microtoroid immersed in water@xcite . we modeled the 3d microtoroid and bead structure in 2d using the effective index method ( eim)@xcite . the field evolution across a @xmath113-@xmath114 radius bead attached to the cavity is displayed in fig . [ bead](a ) , while a zoomed - in plot of field distribution around the bead is displayed as an inset . we further obtained the resonance wavelength shift from the additional phase shift of the electrical field attributed to the bead ( i.e. @xmath115 ) and plotted them as a function of bead radius in fig . [ bead](b ) as red cross markers . for comparison , we also plotted the corresponding shift predicted by the perturbation method@xcite . as shown , both are in good agreement aside from the fact that there is a identifiable departure due to the 2d simplification with eim . we believe that a three - dimensional full wave beam propagation method should model the shift with higher accuracy . .49 -@xmath114 bead is located . the inset shows the field distribution inside the bead . ( b ) resonance wavelength shift vs. nanobead radius , for @xmath116 , calculated by the bpm method ( red ) and perturbation method ( black).,title="fig : " ] 0.41 -@xmath114 bead is located . the inset shows the field distribution inside the bead . ( b ) resonance wavelength shift vs. nanobead radius , for @xmath116 , calculated by the bpm method ( red ) and perturbation method ( black).,title="fig : " ] finally , we applied the bpm to model the ellipticity effect . setting the major radius at @xmath117 to @xmath118 and sweeping the other radius from @xmath119 to @xmath120 , we calculated the quality factor as explained before using bpm . the window size is set to @xmath121 with @xmath122 grids in the @xmath68 and @xmath71 with @xmath69 grids in the @xmath70 . we excited the ring with its fundamental mode for a @xmath52-@xmath53 major radius and @xmath54-@xmath53 minor diameter ring resonator . [ ellipticity ] shows the quality factor as a function of ellipticity , in which the ellipticity is defined as @xmath123 where @xmath124 is the fixed radius and @xmath125 is the variable radius for @xmath126 . the inset in fig . [ ellipticity ] shows the field intensity for the extreme case of @xmath127 . and @xmath127 . in conclusion , we implemented a 2d finite - difference beam propagation method for simulating light propagation in whispering - gallery mode microcavities . with this method , we demonstrate the calculation of key optical properties such as resonance wavelength , quality factor in cases where the azimuthal symmetry is absent due to singular perturbations from nano particles , and ellipticity of the cavity . the field scattering that arises from asymmetry is clearly visible from our simulation . such properties are not readily obtainable through first - order perturbations on whispering - gallery modes as traditionally treated by other groups . therefore , cylindrical bpm addresses the need for in - depth study of areas such as biosensing with wgm s where azimuthal symmetry is perturbed . a full vectorial three - dimensional bpm approach will be investigated in future research .	we simulate light propagation in perturbed whispering - gallery mode microcavities using a two - dimensional finite - difference beam propagation method in a cylindrical coordinate system . optical properties of whispering - gallery microcavities perturbed by polystyrene nanobeads are investigated through this formulation . the light perturbation as well as quality factor degradation arising from cavity ellipticity are also studied . 10 m. l. gorodetsky , a. a. savchenkov , and v. s. ilchenko , `` ultimate q of optical microsphere resonators , '' optics letters * 21 * , 453455 ( 1996 ) . d. k. armani , t. j. kippenberg , s. m. spillane , and k. j. vahala , `` ultra - high - q toroid microcavity on a chip , '' nature * 421 * , 925928 ( 2003 ) . f. vollmer , d. braun , a. libchaber , m. khoshsima , i. teraoka , and s. arnold , `` protein detection by optical shift of a resonant microcavity , '' applied physics letters * 80 * , 40574059 ( 2002 ) . f. vollmer and s. arnold , `` whispering - gallery - mode biosensing : labelfree detection down to single molecules , '' nat . methods * 5 * , 591596 ( 2008 ) . t. lu , h. lee , t. chen , s. herchak , j .- h . kim , s. e. fraser , and k. vahala , `` high sensitivity nanoparticle detection using optical microcavities , '' pnas * 108 * , 59765979 ( 2011 ) . j. dominguez - juarez , g. kozyreff , and j. martorell , `` whispering gallery microresonators for second harmonic light generation from a low number of small molecules , '' nautre communications * 2 * , 18 ( 2010 ) . j. knittel , t. g. mcrae , k. h. lee , and w. p. bowen , `` interferometric detection of mode splitting for whispering gallery mode biosensors , '' applied physics letters * 97 * , 13 ( 2010 ) . y. sun and x. fan , `` optical ring resonators for biochemical and chemical sensing , '' anal bioanal chem * 399 * , 205211 ( 2011 ) . s. spillane , t. j. kippenberg , and k. j. vahala , `` ultralow - threshold raman laser using a spherical dielectric mcirocavity , '' nature * 415 * , 621623 ( 2002 ) . b. min , t. j. kippenberg , and k. j. vahala , `` compact , fiber - compatible , cascaded raman laser , '' optics letters * 28 * , 15071509 ( 2003 ) . m. cai and k. j. vahala , `` highly efficient hybrid fiber taper coupled microsphere laser , '' optics letters * 26 * , 884886 ( 2001 ) . a. polman , b. min , j. kalkman , t. j. kippenberg , and k. j. vahala , `` ultralow - threshold erbium - implanted toroidal microlaser on silicon , '' applied physics letters * 84 * , 10371039 ( 2004 ) . t. lu , l. yang , r. v. a. van loon , a. polman , and k. j. vahala , `` on - chip green silica upconversion microlaser , '' optics letters * 34 * , 482484 ( 2009 ) . t. lu , l. yang , t. carmon , and b. min , `` a narrow - linewidth on - chip toroid raman laser , '' ieee j. quantum electron . * 47 * , 320326 ( 2011 ) . s. m. spillane , t. j. kippenberg , k. j. vahala , k. w. goh , e. wilcut , and h. j. kimble , `` ultrahigh - q toroidal microresonators for cavity quantum electrodynamics , '' physical review a * 71 * , 013817 ( 2005 ) . m. oxborrow , `` traceable 2-d finite - element simulation of the whispering - gallery modes of axisymmetric electromagnetic resonators , '' ieee transactions on microwave theory and techniques * 55 * , 12091218 ( 2007 ) . j. k. s. poon , j. scheuer , s. mookherjea , g. t. paloczi , y. huang , and a. yariv , `` matrix analysis of microring coupled - resonator optical waveguides , '' opt . express * 12 * , 90103 ( 2004 ) . j. hong , w. p. huang , and t. makino , `` on the transfer matrix method for distributed - feedback waveguide devices , '' j. lightwave technol . * 10 * , 18601868 ( 1992 ) . x. du , s. vincent , and t. lu , `` full - vectorial whispering - gallery - mode cavity analysis , '' opt . express * 21 * , 2201222022 ( 2013 ) . m. d. feit and j. j. a. fleck , `` light propagation in graded - index optical fibers , '' appl . opt . * 17 * , 39903998 ( 1978 ) . d. yevick and b. hermansson , `` efficient beam propagation techniques , '' ieee j. quantum electron . * 26 * , 109 ( 1990 ) . j. saijonmaa and d. yevick , `` beam - propagation analysis of loss in bent optical waveguides and fibers , '' j. opt . . am . * 73 * , 1785 ( 1983 ) . w. huang , c. xu , s .- t . chu , and s. k. chaudhuri , `` the finite - difference vector beam propagation method : analysis and assessment , '' j. lightwave technol . * 10 * , 295305 ( 1992 ) . j. v. roey , j. van der donk , and p. e. lagasse , `` beam - propagation method : analysis and assessment , '' j. opt . . am . * 71 * , 803810 ( 1981 ) . w. huang , c. xu , and s. chaudhuri , `` a finite - difference vector beam propagation method for three - dimensional waveguide structures , '' ieee photon . technol . lett . * 4 * , 148151 ( 1992 ) . b. hermansson and d. yevick , `` propagating - beam - method analysis of two - dimensional microlenses and three - dimensional taper structures , '' opt . soc . am . a * 1 * , 663671 ( 1984 ) . h. rao , r. scarmozzino , and j. richard m. osgood , `` a bidirectional beam propagation method for multiple dielectric interfaces , '' ieee photon . technol . lett . * 11 * , 830832 ( 1999 ) . r. scarmozzino , a. gopinath , r. pregla , and s. helfert , `` numerical techniques for modeling guided - wave photonic devices , '' ieee j. sel . top . quantum electron . * 6 * , 150162 ( 2000 ) . a. ahmed , r. koya , o. wada , m. wang , and r. koga , `` eigenmode analysis of whispering gallery mmode of pillbox - type optical resonators utilizing the fe - bpm formulation , '' ieice tans . electron . * e78-c * , 1638 ( 1995 ) . w. yang and a. gopinath , `` a boundary integral method for propagation problems in integrated optical structures , '' ieee photon . . lett . * 7 * , 777779 ( 1995 ) . t. lu and d. yevick , `` boundary element analysis of dielectric waveguides , '' journal of optical society of america a * 19 * , 1197 1206 ( 2002 ) . t. lu and d. yevick , `` comparative evaluation of a novel series approximation for electromagnetic fields at dielectric corners with boundary element method applications , '' ieee / osa journal of lightwave technology * 22 * , 14261432 ( 2004 ) . t. lu and d. yevick , `` a vectorial boundary element method analysis of integrated optical waveguides , '' ieee / osa journal of lightwave technology * 21 * , 17931807 ( 2003 ) . h. deng and d. yevick , `` the nonunitarity of finite - element beam propagation algorithms , '' ieee photon . . lett . * 17 * , 14291431 ( 2005 ) . s .- t . chu and s. chaudhuri , `` a finite - difference time - domain method for the design and analysis of guided - wave optical structures , '' j. lightwave technol . * 7 * , 20332038 ( 1989 ) . m. reed , t. m. benson , p. c. kendall , and p. sewell , `` antireflectioncoated angled facet design , '' proc . inst . elect . eng . * 143 * , 214220 ( 1996 ) . berenger and jean - pierre , `` a perfectly matched layer for the absorption of electromagnetic waves , '' j. comput . phys . * 114 * , 185200 ( 1994 ) . w. p. huang , c. l. xu , w. lui , and k. yokoyama , `` the perfectly matched layer ( pml ) boundary condition for the beam propagation method , '' ieee photon . . lett . * 8 * , 649651 ( 1996 ) . g. r. hadley , `` high - accuracy finite - difference equations for dielectric waveguide analysis i : uniform regions and dielectric interfaces , '' j. lightwave technol . * 20 * , 12101218 ( 2002 ) . g. r. hadley , `` wide - angle beam propagation using pade approximant operators , '' opt . lett . * 17 * , 14261428 ( 1992 ) . s. lidgate , p. sewell , and t. benson , `` conformal mapping : limitations for waveguide bend analysis , '' sci . meas . technol . * 149 * , 262266 ( 2002 ) . m. rivera , `` a finite difference bpm analysis of bent dielectric waveguides , '' j. lightwave technol . * 13 * , 233 ( 1995 ) . h. deng , g. h. jin , j. harari , j. p. vilcot , and d. decoster , `` investigation of 3-d semivectorial finite - difference beam propagation method for bent waveguides , '' j. lightwave technol . * 16 * , 915922 ( 1998 ) . m. krause , `` finite - difference mode solver for curved waveguides with angled and curved dielectric interfaces , '' j. lightwave technol . * 29 * , 691699 ( 2011 ) . i. h. malitson , `` interspecimen comparison of the refractive index of fused silica , '' j. opt . . am . * 55 * , 12051208 ( 1965 ) . r. kitamura , l. pilon , and m. jonasz , `` optical constants of silica glass from extreme ultraviolet to far infrared at near room temperature , '' appl . opt . * 46 * , 81188133 ( 2007 ) . g. m. hale and m. r. querry , `` optical constants of water in the 200-nm to 200-m wavelength region , '' appl . opt . * 12 * , 555563 ( 1973 ) . k. s. chiang , `` performance of the effective - index method for the analysis of dielectric waveguides , '' opt . lett . * 16 * , 714716 ( 1991 ) . r. a. waldron , m.a . , and a.inst.p . , `` perturbation theory of resonant cavities , '' proceedings of the iee - part c : monographs * 107 * , 272274 ( 1960 ) .
the notion of lorentzian wormholes ( wh ) arose when morris and thorne explored the possibility of time travels for humans using the principles of general relativity ( gr ) @xcite . einstein s theory of gr predicts that the structure and geometry of spacetime in the presence of matter is not rigid but it is elastic and deformable . the more compact the object is , the more strong the curvature of space is , which essentially leads to the idea of black holes . however in the later case , the fabric of spacetime loses its meaning at the curvature singularity . if somehow the formation of singularity is avoided then it would be possible to travel in and out of the horizon . the possibility of such a solution to the einstein field equations was explored for the first time by flamm @xcite soon after the formulation of gr , but it was later shown that his solution was unstable . a typical wormhole is a tube - like structure which is asymptotically flat from both sides . the radius of the wormhole throat could be constant or variable depending on its construction and it is termed static or non - static respectively . gr predicts that to form a wh , an exotic form of matter ( violating the energy conditions ) must be present near the throat of the wh . the problem is the dearth of reasonable sources sustaining the wormhole geometry . one possible candidate is the phantom energy ( which is a cosmic dynamical scalar field with negative kinetic energy in its lagrangian ) and it is one of the candidates for explaining cosmic accelerated expansion as well @xcite . since the existence of phantom energy is questionable and no other suitable exotic matter candidate is available at the moment , an alternative approach is commonly followed : investigation if the modifications of laws of gravity ( i.e. gr ) , proposed primarily for explanation of accelerated expansion and avoiding singularities , can support the wh geometries . since the wh is a non - vacuum solution of einstein field equations , the presence of some form of energy - matter is necessary to construct a wh . in the framework of modified gravity , the matter content is assumed to satisfy the energy conditions near the wh throat , while higher curvature correction terms in the lagrangian are required to sustain the wh geometry . like @xmath0 ( where @xmath1 is ricci scalar ) gravity which is based on a symmetric connection , the @xmath2 ( where @xmath3 is torsion of spacetime ) gravity is based on a skew - symmetric connection @xcite . the latter theory is one of the many alternative ( or modified ) theories of gravity available in literature @xcite . @xmath2 theory is based on the tetrad formalism and the governing equations are derived by varying the gravitational lagrangian with respect to the tetrads . since the tetrad is not unique , hence the governing equations acquire different forms and hence different solutions in different tetrads ( see details in sec . ii ) . like any other theory , @xmath2 theory have several drawbacks : it does not obey local lorentz invariance ; violates the first law of thermodynamics ; no unique governing field equations @xcite . despite these problems , the theory efficiently describes the cosmic accelerated expansion and predicts the existence of new kinds of black holes and wormholes . it can also resolve the dark energy and dark matter conundrums @xcite . recently some attempts have been made to unify both @xmath0 and @xmath2 theories as @xmath4 gravity o by introducing a boundary term @xmath5 in @xmath6 gravity @xcite . here , we want to study wormholes in a theory where torsion is non - minimally coupled with a scalar field and includes a matter action . it is important to note that similar models have also been studied in literature @xcite . in this paper , we employ the noether symmetry approach @xcite and a wormhole metric ansatz to obtain the governing system of differential equations . after solving the equations , we get the form of metric coefficients , symmetry generators , invariants and the form of torsion and potential functions . this approach has been extensively studied in literature @xcite . the plan of the paper is as follows : in sec . ii , we provide a brief review of generalized teleparallel gravity and sketch relevant notations . in sec . iii , we present the model starting with the action of a non - minimally coupled scalar field with both torsion and the boundary term . we also derive the field equations and choose an ansatz for the wormhole metric . in sec . iv and its sub - sections , we construct a point - like lagrangian , search for noether symmetries and derive the wormhole solutions . discussion and conclusions are given in sec . in what follows we will follow conventions outlined in @xcite . here we present a brief review of the teleparallel formalism of gr . unlike gr , the dynamical variable in the teleparallel theory is the tetrad @xmath7 ( rather than the metric tensor ) , where greek and latin indices denote spacetime and tangent space indices respectively . here the metric tensor @xmath8 is related with the tetrads as @xmath9 where @xmath10 denotes the minkowski metric tensor . the inverse tetrad @xmath11 is defined as @xmath12 here @xmath13 is the determinant of the tetrad @xmath14 , which can be evaluated from the determinant of the metric tensor @xmath15 . gr is based on the symmetric levi - civita connection , whereas , teleparallel gravity relies on the anti - symmetric weitzenbck connection @xmath16 defined as @xmath17 the anti - symmetric nature of connection leads to the concept of torsion in geometry . the torsion tensor is the antisymmetric part of the weitzenbck connection @xmath18 the tensor @xmath19 , referred as the torsion vector , is defined through the contraction of the torsion tensor , i.e. @xmath20 to construct the field equations of teleparallel gravity , we consider the following lagrangian density and vary it with respect to the tetrad @xmath21 where @xmath22 the torsion scalar @xmath3 is defined as @xmath23 furthermore , to express teleparallel gravity as an equivalent of gr , we write the levi - civita connection @xmath24 in terms of the weitzenbck connection as @xmath25 where here @xmath26 is called the contortion tensor and it is defined as @xmath27 this contortion tensor is antisymmetric in its last two indices . now expressing the ricci scalar of the levi - civita connection in terms of the weitzenbck connection ; using ( [ levi ] ) , we get @xmath28 as the difference between the ricci scalar and the torsion scalar is simply a total derivative , the action ( [ taction ] ) gives rise to the same dynamics as the einstein hilbert action . this shows that teleparallel gravity is indeed equivalent to gr . defining the boundary quantity @xmath29 one then has simply the relation @xmath30 . note that one can write @xmath5 in terms of a levi - civita covariant derivative simply as @xmath31 . let us consider the following gravitational action which describes a non - minimally coupled scalar field to both torsion and the boundary term @xcite , @xmath32 e\ , d^4x.\label{action}\end{aligned}\ ] ] here , @xmath33 is the scalar field potential and @xmath34 and @xmath35 are smooth functions of the scalar field @xmath36 . this action is not a new theory , instead it is a generalisation or rather a unification of different theories into one action . this theory is very rich in the sense that one can recover very well - known scalar - tensor theories . in fact , let us for example choose @xmath37 where @xmath38 and @xmath39 are coupling constants . depending on the value of @xmath39 and @xmath38 , we can recover scalar tensor theories non - minimally coupled with the torsion scalar ( @xmath40 ) , with the boundary term ( @xmath41 ) , with the ricci scalar ( @xmath42 ) and also quintessence theory ( @xmath43 ) . the latter theories have been widely studied in the literature from cosmology to astrophysical sources like wormholes . traversable wormholes supported by non - minimally coupled scalar field with the ricci scalar ( @xmath42 ) were first studied in ref . @xcite . they found that depending on the value of the coupling constant , one can construct different kind of wormholes . wormholes supported by a kink - like configuration of a scalar field was studied in @xcite . moreover , in @xcite , the authors studied the stability of electrically charged and neutral wormholes within this theory . + additionally , canonical brans - dicke scalar tensory theory with @xmath44 can be recovered if we set @xmath45 with @xmath42 . wormholes configurations under this theory have been widely studied in the literature @xcite . one interesting feature of bans - dicke wormholes is that they can satisfy the energy conditions . for the reader interested on cosmology on all these theories mentioned above , see refs @xcite . in the context of teleparallel gravity , wormhole solutions have also been studied ( see @xcite and their referees therein ) , but to the best of our knowledge , wormholes in teleparallel scalar - tensor theories ( for example with @xmath46 ) have not been considered yet until now . note that action is very rich since we can study all these kind of theories and then make the corresponding choice of the coupling functions . moreover , using this approach , we can have a direct relation from the teleparallel and metric counterparts . + by varying this action with respect to the tetrad field we find the following field equations @xmath47-e^{\nu}_a \left[\frac{1}{2}\partial_\mu \phi \partial^\mu \phi -v(\phi)\right ] \nonumber\\ + e^{\mu}_a \partial^\nu \phi \partial_\mu \phi + 2(\partial_{\mu}f(\phi)+\partial_{\mu}g(\phi ) ) e^\rho_a s_{\rho}{}^{\mu\nu}+e^{\nu}_{a}\box g(\phi)-e^\mu_a \nabla^{\nu}\nabla_{\mu}g(\phi)=0\ , . \ ] ] by contracting this equation with the tetrad field @xmath48 , we can have this equation only in space - time index @xmath49-\delta^{\nu}_{\lambda } \left[\frac{1}{2}\partial_\mu \phi \partial^\mu \phi -v(\phi)\right ] \nonumber\\ + \partial^\nu \phi \partial_\lambda \phi + 2(\partial_{\mu } f(\phi)+\partial_{\mu}g(\phi ) ) s_{\lambda}{}^{\mu\nu}+\delta^{\nu}_{\lambda}\box g(\phi)- \nabla^{\nu}\nabla_{\lambda}g(\phi)=0\ , . \end{aligned}\ ] ] now , by varying the action ( [ action ] ) with respect to the scalar field we obtain the so - called , klein - gordon equation which reads @xmath50 let us consider a static spherically symmetric space - time in which the metric is @xmath51 where @xmath52 and @xmath53 are functions of the radial coordinate @xmath54 , and @xmath55 . + the corresponding off - diagonal tetrad field to the metric ( [ metric ] ) is @xmath56 this tetrad field has been used in the context of @xmath2 gravity since it does not produce a constraint , namely @xmath57 , as a diagonal tetrad . for this tetrad , the field equations read @xmath58 here , primes denotes differentiation with respect to the radial coordinate @xmath54 . it can be shown that if we choose @xmath59 we recover the case studied in @xcite . + additionally , the klein - gordon equation becomes @xmath60 where the torsion scalar and the boundary term are respectively given by @xmath61\,.\label{b2}\end{aligned}\ ] ] clearly , by substracting ( [ b2 ] ) with ( [ t2 ] ) we recover the ricci scalar @xmath62\,.\end{aligned}\ ] ] now , we have all the ingredients to study wormhole configurations in this framework . a spherically symmetric wormhole will be given by choosing @xmath63 where @xmath64 is the shape function of the wormhole and @xmath65 is known as the redshift function . thus , the field equations read @xmath66 in the following section , we will use the noether symmetry approach to find analytical wormhole solutions within this theory . the action ( [ action ] ) can be written as @xmath67 , where @xmath68 is the lagrangian density . in the background of the static spherically symmetric space - time ( [ metric ] ) , the point - like lagrangian density for the teleparallel theory of gravity is obtained as @xmath69 - g(\phi ) e^{a/2 } \left(m a ' + 2 m ' \right ) \nonumber \\ & & \qquad \quad - g_{,\phi } e^{\frac{a - b}{2 } } \left ( \frac{m^2}{2 } a ' \phi ' + 2 m m ' \phi ' \right ) - \frac{m^2}{2 } e^{\frac{a - b}{2 } } \phi'^2 - m^2 e^{\frac{a+b}{2 } } v(\phi ) \ , , \label{lagr1}\end{aligned}\ ] ] where @xmath70 . note that the hessian determinant of the above lagrangian is zero . this is clearly due to the absence of the generalized velocity @xmath71 in the point - like lagrangian . by varying the point - like lagrangian density ( [ lagr1 ] ) with respect to the redshift function @xmath72 , we get @xmath73 + 2 f_{,\phi } \frac{\phi'}{m } \left ( m ' - e^{b/2 } \right ) - g_{,\phi } \left ( \phi '' + 2 e^{b/2 } \frac{\phi'}{m } \right ) + \frac{1}{2 } \phi'^2 + e^{b } v_{,\phi } = 0\ , . \label{eq11}\end{aligned}\ ] ] variations with respect to the shape function @xmath74 gives us @xmath75 - g_{,\phi } \left ( \frac{a ' \phi'}{2 } + \frac { 2 m ' \phi'}{m } \right ) - \frac{1}{2 } \phi'^2 + e^{b } v ( \phi ) = 0\ , . \label{eq22}\end{aligned}\ ] ] now , if we vary the point - like lagrangian density with respect to @xmath76 , we find @xmath77 + f_{,\phi } \phi ' \left [ \frac{a'}{2 } + \frac{1}{m } \left ( m ' - e^{b/2 } \right ) \right ] \nonumber \\ & & \quad + g_{,\phi } \left [ \phi ' \left ( \frac{b'}{2 } - \frac { e^{b/2}}{m } \right ) - \phi '' \right ] + \left ( \frac{1}{2}- g_{,\phi \phi } \right ) \phi'^2 + e^{b } v_{,\phi } = 0\ , . \label{eq33}\end{aligned}\ ] ] finally , by varying the point - like lagrangian density with respect to @xmath36 yields @xmath78 + \phi '' + \phi ' \left [ \frac{1}{2 } ( a ' - b ' ) + \frac{2 m'}{m } \right ] - e^{b } v_{,\phi } = 0 , \label{eq44}\end{aligned}\ ] ] the latter equation is the klein - gordon equation . + the _ energy function _ associated with @xmath68 is defined by @xmath79 which is also the _ hamiltonian _ of the system . here , @xmath80 , are the generalized coordinates where @xmath81 , for the lagrangian density ( [ lagr1 ] ) of teleparallel theory of gravity . then @xmath82 has the following form @xmath83 , \label{energy2}\ ] ] which vanishes because of eq . ( [ eq22 ] ) due to the variation with respect to @xmath74 , i.e. @xmath84 . this can be explicitly solved in terms of @xmath74 as a function of the remaining generalized coordinates : @xmath85}{m^2 v(\phi ) - f(\phi ) } . \label{coord - b}\ ] ] then , using ( [ worm - b ] ) , the shape function @xmath64 of the wormhole takes the form @xmath86}{\left [ - f(\phi ) \left ( \frac{m'^2}{m^2 } + \frac{a ' m'}{m } \right ) + g_{,\phi } \left ( \frac{1}{2 } a ' \phi ' + \frac{2 m ' \phi'}{m } \right ) + \frac{1}{2 } \phi'^2 \right ] } . \label{shape-1}\ ] ] in its present form , @xmath87 is expressed in terms of several arbitrary functions . below , we shall determine explicitly @xmath87 as a function of radial coordinate and qualitatively investigate its behavior . let us consider a noether symmetry vector generator @xcite @xmath88 where @xmath81 are the generalized coordinates in the configuration space @xmath89 of the lagrangian , whose tangent space is @xmath90 . the components @xmath38 and @xmath91 of the noether symmetry generator @xmath92 are functions of @xmath54 and @xmath93 . the existence of a noether symmetry implies the existence of a vector field @xmath92 given in ( [ ngs - gen ] ) if the lagrangian @xmath94 satisfies @xmath95 } \mathcal{l } + \mathcal{l } ( d_r \xi ) = d_r g\ , , \label{ngs - eq}\ ] ] where @xmath96}$ ] is the first prolongation of the generator ( [ ngs - gen ] ) in such a form @xmath95 } = { \bf x } + \eta'^i \frac{\partial}{\partial q'^i},\ ] ] @xmath97 is a gauge function , @xmath98 is the total derivative operator with respect to @xmath54 , @xmath99 , and @xmath100 is defined as @xmath101 . it is important to give the following noether first integral to emphasize the significance of noether symmetry that if @xmath92 is the noether symmetry generator corresponding to the lagrangian @xmath102 , then @xmath103 which is also the hamiltonian or a conserved quantity associated with the generator @xmath92 . now we seek for the condition in order that the lagrangian density ( [ lagr1 ] ) would admit any noether symmetry . the noether symmetry condition ( [ ngs - eq ] ) for the lagrangian ( [ lagr1 ] ) gives rise to the following set of differential equations @xmath104 + f_{,\phi } \eta^4 - 2 g_{,\phi } m \eta^4_{,m } = 0 , \nonumber \\ & & g_{,\phi } \left ( m \eta^4_{,a } + \frac{1}{4 } m^2 \eta^4_{,m } \right ) - f \left [ \frac{1}{4 } m ( \eta^1 - \eta^2 ) + \frac{1}{2 } \eta^3 + \eta^3_{,a } + \frac{1}{2 } m ( \eta^1_{,a } + \eta^3_{,m } ) \right ] - \frac{1}{2 } f_{,\phi } \eta^4 = 0 , \nonumber \\ & & g_{,\phi } \left [ \frac{1}{2 } m ( \eta^1 - \eta^2 ) + 2 \eta^3 + 4 \eta^3_{,a } + m ( \eta^1_{,a } + \eta^4_{,\phi } - \xi_{,r } ) \right ] + g_{,\phi \phi } m \eta^4 + 2 m\eta^4_{,a } - 2 f \eta^3_{,\phi } = 0 , \label{ngs - eqs } \\ & & g_{,\phi } \left [ \frac{1}{2 } m ( \eta^1 - \eta^2 ) + \eta^3 + \frac{1}{4 } m^2 \eta^1_{,m } + m ( \eta^3_{,m } + \eta^4_{,\phi } - \xi_{,r } ) \right ] + g_{,\phi \phi } m \eta^4 - \frac{1}{2 } f m \eta^1_{,\phi}- f \eta^3_{,\phi } + \frac{1}{2 } m^2 \eta^4_{,m } = 0 , \nonumber\\ & & ( f + g ) \left [ \frac{1}{2 } m \eta^1 + \eta^3 + m \eta^1_{,a } + 2 \eta^3_{,a } \right ] + m ( f+g)_{,\phi } \eta^4 - f m e^{-b/2 } \eta^3_{,r } + \frac{1}{2 } g_{,\phi } m^2 e^{-b/2 } \eta^4_{,r } + e^{-a/2 } g_{,a } = 0 , \nonumber\\ & & ( f + g ) \left [ m \eta^1_{,b } + 2 \eta^3_{,b } \right ] + e^{-a/2 } g_{,b } = 0 , \nonumber \\ & & ( f + g ) \left ( \eta^1 + m \eta^1_{,m } + 2 \eta^3_{,m } \right ) + 2 ( f+g)_{,\phi } \eta^4 - f m e^{-b/2 } \eta^1_{,r } - f e^{-b/2 } \eta^3_{,r } + 2 g_{,\phi } m e^{-b/2 } \eta^4_{,r } + e^{-a/2 } g_{,m } = 0 , \nonumber\\ & & ( f + g ) \left [ m \eta^1_{,\phi } + 2 \eta^3_{,\phi } \right ] + g_{\phi } e^{-b/2 } \left ( \frac{1}{2 } m^2 \eta^1_{,r } + 2 m \eta^3_{,r } \right ) + m^2 e^{-b/2 } \eta^4_{,r } + e^{-a/2 } g_{,\phi } = 0 , \nonumber \\ & & ( m^2 v(\phi ) - f ) \left [ \frac{1}{2 } ( \eta^1 + \eta^2 ) - \xi_{,r } \right ] + 2 m v(\phi ) \eta^3 + ( m^2 v(\phi)_{,\phi } - f_{,\phi } ) \eta^4 \nonumber \\ & & \qquad + e^{-b/2}(f + g ) ( m \eta^1_{,r } + 2 \eta^3_{,r } ) + e^{-(a+b)/2 } g_{,r } = 0 . \nonumber\end{aligned}\ ] ] it is explicitly seen from the above equations that @xmath105 , and a trivial solution is @xmath106 and @xmath107 . this trivial solution means that @xmath108 is a noether symmetry for any form of the @xmath109 and @xmath33 . the corresponding noether constant of @xmath110 is @xmath111 which vanishes due to the eq . ( [ eq22 ] ) . * case ( i ) : * @xmath112 ( vanishing potential ) . relevant subcases : * ( i.a). take @xmath113 and @xmath114 , where @xmath115 and @xmath39 are constants . ( i.a.1 ) if @xmath116 , then we have teleparallel dark energy ( non - minimally coupled to @xmath3 ) . ( i.a.2 ) if @xmath117 , then it means that non - minimally coupled to the ricci scalar . ( i.a.3 ) if @xmath118 , then it gives non - minimally coupled to the boundary term . ( i.a.4 ) the condition @xmath119 gives rise to _ quintessence models_. ( i.b). if @xmath120 and @xmath121 , then the possibilities are ( i.b.1 ) @xmath116 , ( i.b.2 ) @xmath117 ( brans - dicke theory with @xmath122 ) , ( i.b.3 ) @xmath118 , and ( i.b.4 ) the condition @xmath119 gives rise to _ quintessence models _ and same as subcase ( i.a.4 ) . the noether symmetries for the lagrangian density ( [ lagr1 ] ) are obtained as @xmath123 for the subcases ( i.a.1 ) , ( i.a.2 ) , ( i.a.3 ) , ( i.b.1 ) and ( i.b.2 ) which yields that @xmath124 the corresponding first integrals of the above noether symmetries are @xmath125 where @xmath126 vanishes due to the @xmath84 . the second of the above first integrals gives @xmath127 $ ] , where @xmath128 . thus the eq . @xmath84 takes the form @xmath129 , \label{ode-1}\end{aligned}\ ] ] where @xmath130 is defined as @xmath131 , @xmath132 and @xmath133 are defined by @xmath134 the eq . ( [ ode-1 ] ) is a non - linear differential equation for @xmath130 , and it could be solved under some assumptions . in subcase ( i.b.3 ) , i.e. @xmath135 and @xmath136 , we have three noether symmetries @xmath137 and @xmath138 with @xmath139 . thus we found the first integrals in this case as @xmath140 and @xmath141 . ( i.c). when @xmath142 and @xmath135 ( const . ) , the components of noether symmetry generator @xmath92 are @xmath143 then the noether symmetries are @xmath137 given in ( [ ngsv-1 ] ) and @xmath144 with the non - vanishing lie brackets @xmath145 = \frac{1}{2 } { \bf x}_1 , \qquad \left [ { \bf x}_2 , { \bf x}_4 \right ] = - \frac{1}{2 } f_0 { \bf x}_3 , \qquad \left [ { \bf x}_2 , { \bf x}_5 \right ] = - \frac{1}{2 } { \bf x}_2 , \\ & & \left [ { \bf x}_3 , { \bf x}_4 \right ] = { \bf x}_2 , \quad \qquad \left [ { \bf x}_4 , { \bf x}_5 \right ] = - \frac{1}{2}\phi { \bf x}_2 - \frac{1}{2 } f_0 \left ( 1 + \ln m \right ) { \bf x}_3 .\end{aligned}\ ] ] the corresponding noether constants are @xmath146 where @xmath147 because of the fact @xmath84 , which yields @xmath148 using the first integral @xmath149 , we have @xmath150 where @xmath151 and @xmath131 . then , putting the above term into the the first integral @xmath152 , it becomes @xmath153 which has the solution @xmath154 , that is , @xmath155 or @xmath156 . taking @xmath157 , one can write that @xmath158 where @xmath159 . thus , together with the first integral @xmath160 , the eq . ( [ eb - f0 ] ) yields @xmath161 where @xmath162 . the shape function of the wormhole becomes @xmath163 } { ( 2 i_2 + 1 ) m'^2 } \right),\end{aligned}\ ] ] where @xmath164 and @xmath165 . the function @xmath53 is independent of @xmath166 and @xmath36 because of isotropy , and it must have the limit @xmath167 as @xmath168 . the form of the function @xmath53 can be determined by use of the geodesic deviation equation ( see page 50 of ref . @xcite ) , which gives that the solutions for @xmath53 with the appropriate limit behaviour @xmath167 as @xmath168 are @xmath169 if @xmath170 , @xmath54 if @xmath171 and @xmath172 if @xmath173 , where @xmath174 is the curvature parameter and @xmath175 is a dimensional constant . in the limit case @xmath176 , the metric coefficients , the shape function and the scalar field take the form @xmath177 , \label{w1-b } \\ & & \phi(r ) = - \frac{1}{(i_2 + 1 ) } \sqrt{\frac{2i_2 + 1}{2 f_0 } } \arctan \left [ \sqrt{\frac{1}{\ell } \left ( \frac{r}{r_0 } \right)^{2(i_2 + 1 ) } -1 } \right ] . \label{w1-phi}\end{aligned}\ ] ] the metric coefficients ( [ w1-a ] ) and the shape function ( [ w1-b ] ) are very similar to the wormhole solution found by cataldo et . al @xcite under the assumption of isotropic pressure . it can be noted that in this solution we must set @xmath178 to satisfy the wormhole condition at the throat @xmath179 . [ fig1 ] and [ fig2 ] show the behaviour of this solution for @xmath180 , @xmath65 and @xmath181 for different values of the parameters @xmath182 , @xmath183 and @xmath184 . in order to have a real and positive scalar field , the parameters must satisfy @xmath185 and @xmath186 . moreover , the scalar field must lie outside the wormhole in the region @xmath187 $ ] , otherwise it will become negative or complex . we can see from the pictures that the wormhole obeys the asymptotic flatness in both cases since as @xmath188 , a de sitter space - time is recovered . for @xmath169 , we find that @xmath189 , \label{w2-b}\end{aligned}\ ] ] which is a new wormhole solution as far as we know . one can find a similar solution for @xmath190 . + if @xmath191 , which corresponds to the sub cases ( i.a.1 ) and ( i.b.1 ) , then the noether symmetries are @xmath110 and @xmath192 given in ( [ ngsv-1 ] ) for arbitrary @xmath34 . ( i.d). if @xmath191 and @xmath193 , where @xmath194 , the solution of noether symmetry equations ( [ ngs - eqs ] ) are @xmath195 that is , the @xmath137 by ( [ ngsv-1 ] ) and @xmath196 are noether symmetries . here there is only non - vanishing lie bracket due to the @xmath137 and @xmath197 as @xmath198 = \frac{1}{2}\left ( 1- \frac{1}{m } \right ) { \bf x}_2.\end{aligned}\ ] ] here the first integrals for @xmath137 and @xmath197 are @xmath199 and @xmath200 if @xmath191 and @xmath201 , which corresponds to the sub case ( i.a.4 ) or ( i.b.4 ) , then the components of noether symmetry generator @xmath92 are same as ( [ ngsc-2 ] ) , but the gauge function @xmath202 is obtained by @xmath203 . therefore in this case the noether constant @xmath204 related with @xmath205 has the form @xmath206 case ( ii ) : * @xmath207 ( constant potential ) . when @xmath142 and @xmath135 or @xmath191 and @xmath135 ( constant ) , we have only two noether symmetries @xmath108 and @xmath208 which gives rise to the noether constants @xmath209 and @xmath210 . thus , the relation @xmath211 gives @xmath212 for @xmath213 , it follows that @xmath214 which yields the shape function of the wormhole as @xmath215}{e^a ( r a ' + 1)}. \label{shape-2}\end{aligned}\ ] ] considering the specific redshift function given by @xmath216 @xcite , and inserting this condition into the eq . ( [ shape-2 ] ) the shape function reduces to the form @xmath217 now , we take into account the noether first integral @xmath218 , and find that @xmath219 + \phi_0 , \label{phi-1}\end{aligned}\ ] ] where @xmath220 is an integration constant . + for @xmath221 , where @xmath222 , there is only one noether symmetry @xmath108 , which is trivial solution and corresponds to the first integral @xmath223 . teleparallel theories and their extensions have gained a lot of attention in recent years as alternative gravitation frameworks . additionally , in literature , these theories have been studied by their couplings , both minimal and non - minimal , with torsion . our choice of action generalizes and extends most of the earlier models of teleparallel gravity by assuming an additional coupling between the scalar field and a boundary term @xmath5 which is related to the divergence of torsion . moreover , this new coupling allows to formulate a theory that , under some constraints , allows to construct quintessence , teleparallel dark energy and non - minimally coupled scalar field to the ricci scalar theories . the latter has been used in the literature to study spherically symmetric configurations as wormholes ( see ref . @xcite ) obtaining the existence of traversable wormholes supported by non - minimally coupled scalars . here we derived new wormhole solutions according to the morris and thorne paradigm . in this connection , we adopt the noether symmetry approach solving the governing equations to obtain symmetry generators , gauge function and the metric coefficients of the wormhole geometry . to perform this analysis , we focused on two forms of potential function , i.e. @xmath224 and @xmath225 . in the former case , we obtained five symmetry generators while in latter case , only two . for the vanishing potential , we explicitly found two different wormhole solutions in the case @xmath226 , i.e. , within a quintessence model . the first wormhole solution is similar to that found by cataldo et . al @xcite under the assumption of isotropic pressure , whereas the second wormhole solution is , in our knowledge , a new wormhole solution . it can be seen that these solutions respect the asymptotic flatness as @xmath188 . in the case where a constant potential is assumed , we also found an explicit wormhole solution for the quitessence case ( @xmath227 and @xmath228 ) . for a power - law form of potential , only one generator exists which corresponds to a trivial solution . it might be interesting to perform a similar analysis by assuming a more general action as , for example , to include also coupling between the scalar field and the teleparallel torsion gauss bonnet term @xmath229 or a gauss bonnet boundary term @xmath230 ( see for example ref . . these will be the arguments of future work . s.c . wishes to acknowledge the support by istituto nazionale di fisica nucleare , sezione di napoli , italy , iniziative specifiche teongrav and qgsky . s. b. and s. c. also acknowledge the cost action ca15117 ( cantata ) , supported by cost ( european cooperation in science and technology ) . is supported by the comisin nacional de investigacin cientfica y tecnolgica ( becas chile grant no . 72150066 ) . s. nojiri and s. d. odintsov , int . j. geom . * 4 * , 115 ( 2007 ) s. capozziello and m. de laurentis , phys . rept . * 509 * , 167 ( 2011 ) s. nojiri and s. d. odintsov , phys . rept . * 505 * , 59 ( 2011 ) yi - fu cai , s. capozziello , m. de laurentis , e. n. saridakis , arxiv:1511.07586 ; m. jamil , d. momeni , r. myrzakulov , eur . j. c * 72 * , 1959 ( 2012 ) ; ibid , eur . j. c * 72 * , 2122 ( 2012 ) s. bahamonde , c. g. bhmer and m. wright , phys . d * 92 * ( 2015 ) no.10 , 104042 s. bahamonde and c. g. boehmer , arxiv:1606.05557 [ gr - qc ] . s. capozziello , m. de laurentis , r. myrzakulov , int.j.geom.meth.mod.phys . * 12 * , 1550095 ( 2015 ) . m. jamil , d. momeni , r. myrzakulov , eur . j. c * 72 * , 2075 ( 2012 ) ; ibid , eur . phys . j. c * 72 * , 2137 ( 2012 ) ; ibid , int . j. theor . phys . * 54 * , 1098 ( 2015 ) ; a. aslam , m. jamil , d. momeni , r. myrzakulov , canadian j. phys . , * 91 * , 93 ( 2013 ) s. capozziello , r. de ritis , c. rubano and p. scudellaro , riv . nuovo cim . * 19n4 * ( 1996 ) 1 . s. basilakos , s. capozziello , m. de laurentis , a. paliathanasis and m. tsamparlis , phys . rev . d * 88 * , 103526 ( 2013 ) ; s. capozziello and r. de ritis , class . * 24 * , 2153 ( 2007 ) . a. paliathanasis , s. basilakos , e. n. saridakis , s. capozziello , k. atazadeh , f. darabi and m. tsamparlis , phys . d * 89 * , 104042 ( 2014);i . hussain , m. jamil , f. m. mahomed , astrophys space sci . * 337 * , 373 ( 2012 ) ; m. jamil , f. m. mahomed , d. momeni , phys . lett . b * 702 * , 315 ( 2011 ) . s. v. sushkov and s. w. kim , class . * 19 * ( 2002 ) 4909 doi:10.1088/0264 - 9381/19/19/309 [ gr - qc/0208069 ] . k. a. bronnikov and s. grinyok , grav . cosmol . * 7 * , 297 - 300 ( 2001 ) . [ gr - qc/0201083 ] . k. a. bronnikov and s. v. grinyok , grav . cosmol . * 11 * ( 2005 ) 75 - 81 [ gr - qc/0509062 ] . e. f. eiroa , m. g. richarte and c. simeone , phys . lett . * a373 * , 1 ( 2008 ) ; * a373 * , 2399(e ) ( 2009);a . g. agnese and m. la camera , phys . d * 51 * , 2011 ( 1995);k . k. nandi , a. islam and j. evans , phys . d * 55 * , 2497 ( 1997 ) ; f. he and s .- w . kim , phys . rev . d * 65 * , 084022 ( 2002 ) ; a. bhadra and k. sarkar , mod . a * 20 * , 1831 ( 2005).k . a. bronnikov and a. a. starobinsky , jetp lett . * 85 * , 1 ( 2007 ) ; k. a. bronnikov and a. a. starobinsky , mod . a24 * , 1559 ( 2009 ) ; k. a. bronnikov , m. v. skvortsova and a. a. starobinsky , grav . cosmol . * 16 * , 216 ( 2010 ) . d. f. torres , phys . d * 66 * ( 2002 ) 043522;l . amendola , phys . rev . d * 62 * ( 2000 ) 043511 ; v. faraoni , int . j. mod . phys . d * 11 * ( 2002 ) 471 ; b. boisseau , g. esposito - farese , d. polarski and a. a. starobinsky , phys . * 85 * ( 2000 ) 2236 f. perrotta , c. baccigalupi and s. matarrese , phys . d * 61 * ( 1999 ) 023507 c. q. geng , c. c. lee , e. n. saridakis and y. p. wu , phys . b * 704 * ( 2011 ) 384 .	* abstract : * a gravitational theory of a scalar field non - minimally coupled with torsion and boundary term is considered with the aim to construct lorentzian wormholes . geometrical parameters including shape and redshift functions are obtained for these solutions . we adopt the formalism of noether gauge symmetry approach in order to find symmetries , lie brackets and invariants ( conserved quantities ) . furthermore by imposing specific forms of potential function , we are able to calculate metric coefficients and discuss their geometrical behavior .
cosmological observations , in particular the data from cosmic microwave background ( cmb ) , have been providing us knowledge of the history and constituents of the universe since the cosmic background explorer survey . over time we have been able to constrain our knowledge of the early universe with increasing precision . two full sky satellite surveys of cmb , the wilkinson microwave anisotropy probe ( wmap ) @xcite and planck @xcite , have released their data very recently ( last release of wmap data and first release of planck data ) . as the most precise cmb observation , planck has modified the derived cosmological parameters that we had obtained from the wmap and other cmb surveys including the atacama cosmology telescope @xcite and the south pole telescope @xcite . surprisingly , within the same framework of the standard concordance @xmath4cdm model , we find that the values of some of the key cosmological parameters derived from planck data are significantly different than the derived values from earlier cmb observations . for example , the values of the hubble parameter @xmath5 and the dark energy density are found to be considerably less from planck data compared to the values reported by wmap . two questions immediately arise from these differences : first , whether planck or earlier cmb observations have some unknown systematic that is reflected in their reported angular power spectra and second , whether the standard concordance @xmath4cdm model is a suitable and flexible model to explain different cmb data for large multipoles . in our two paper series we will try to address these two questions . we address the consistency of the concordance model of cosmology with planck data in the other paper of this series @xcite . in this paper we analyze and compare the reported angular power spectra from wmap and planck surveys , to search for any significant deviation . we should note that looking at individual cosmological parameters can not trivially imply inconsistency between the two observations . this is basically due to the fact that the standard six cosmological parameters of the concordance model are highly degenerate and not orthogonal . in this paper we use crossing statistic and its bayesian interpretation @xcite to compare the two datasets in the context of the standard model and independent of the individual cosmological parameters . the best fit model to each observation is used as a mean function along with a crossing function to fit the data from the other observation . this allows different smooth variations around a given mean function , allowing us to check whether we can improve the fit to the other data . we have used chebyshev polynomials as the crossing function , as used before in a different context @xcite . in fact chebyshev polynomials have properties of orthogonality and convergence which make them appropriate as a crossing function for smooth mean functions . using the crossing statistic , the consistency of the two datasets can be checked and it can be addressed whether , between the datasets , there lies unknown systematic effects . this paper is organized as follows . in section [ sec : formalism ] we shall describe in detail the framework of the comparison using the crossing statistic . in section [ sec : results ] we provide our results and sketch the conclusions . [ sec : formalism ] in this section we shall briefly discuss the crossing statistic and how we use the method to compare two observations . the crossing statistic was first proposed in @xcite followed by its bayesian interpretation @xcite and was subsequently used in , @xcite for reconstruction of the expansion history and in searching for systematics in data from supernovae and galaxy clusters . the main idea behind the crossing statistic is that given data based on an actual fiducial model and taking a proposed model , the actual model ( hence the data ) and the proposed model will cross each other at @xmath6 points . in the bayesian interpretation of the crossing statistic one can argue that two different models become virtually indistinguishable if one of them is multiplied by a suitable function . the coefficients of this function constitute the crossing hyperparameters and the functions themselves will be called crossing functions following @xcite . a bayesian interpretation of the crossing statistic can be used to test consistency of a proposed model and a given dataset without comparing the proposed model to any other model . in @xcite the crossing statistic has been used to compare two different datasets , searching for possible systematics , and in this paper we will follow a similar approach . similar to @xcite we use chebyshev polynomials of different orders as crossing functions and we multiply them to a given mean function to fit a dataset . if the given mean function is an appropriate choice to express the data , the crossing hyperparameters ( coefficients of the chebyshev polynomials ) would all be consistent with their fiducial values . this basically means that the given mean function does not need any significant modification to fit the data better . however , if the best fit derived crossing hyperparameters deviate significantly from zero , then one can conclude that the given mean function does not express the data well and including some modifications from the crossing function will improve the fit to the data significantly . the power and accuracy of the method has been shown in previous publications - it can be used for various purposes including regression and searching for systematics . + in this paper we consider two datasets , namely wmap 9 year and planck cmb data and we perform our analysis in the framework of the standard @xmath4cdm model as a pool of mean functions . to test the consistency of the two datasets our algorithm is as follows : + 1 . first we fit one of the data sets , let us say planck data assuming the standard concordance model . we call this best fit model @xmath7 . then we assume a crossing function and multiply it by @xmath7 to get @xmath8 , + c_^tt_modified^n = c_^tt_best fit model t_n(c_0,c_1,c_2, ... ,c_n , ) . [ eq : mainmain ] in this work we use chebyshev polynomials of different orders as crossing functions , + t_ii(c_0,c_1,c_2,x)=c_0+c_1 x+c_2(2x^2 - 1 ) [ eq : crossing - function2 ] t_iii(c_0,c_1,c_2,c_3,x)=c_0+c_1 x+c_2(2x^2 - 1)+c_3(4x^3 - 3x ) [ eq : crossing - function3 ] t_iv(c_0,c_1,c_2,c_3,c_4,x)=c_0+c_1 x+c_2(2x^2 - 1)+c_3(4x^3 - 3x)+c_4(8x^4 - 8x^2 + 1 ) [ eq : crossing - function ] where in our case , @xmath9 and @xmath10 is the maximum multipole moment covered by the survey ( the maximum multipole covered by whichever survey that has the broader range ) . to have an idea how crossing functions perform , one can look at the effects of different hyperparameters . considering chebyshev polynomials as the crossing function , @xmath11 clearly affects the overall amplitudes and shifts the mean function up and down . @xmath12 adds a tilt to the mean function and @xmath13 and higher order coefficients allow long range fluctuations with increasing frequencies . the crossing function can have different forms , but considering the shape of the data and our expectations that the angular power spectrum should exhibit smooth behavior in the context of the concordance model , we expect chebyshev polynomials of the first kind will perform satisfactorily . in this regard one can find some detailed discussions in @xcite . it will be explained in the results section that we use chebyshev polynomials only up to fourth order . due to convergence of the results there is usually no need to go to the higher orders . one can use higher orders of chebyshev polynomials but this would introduce more degrees of freedom , larger confidence limits that eventually restrict us from distinguishing between cases . now we fit @xmath14 to wmap 9 year data and derive the confidence limits of the crossing hyperparameters @xmath15 . these coefficients of the chebyshev polynomials ( crossing hyperparameters ) , perform as discriminators between true and false models . eq.[eq : mainmain ] generates many sample models , all based on the given mean function and each @xmath16 set produces variation with a particular likelihood fitting the data . this likelihood represents the `` probability of a particular variation given the data '' . @xmath17 and @xmath18 contours represent @xmath19 and @xmath20 confidence limits . as an example , in case of considering @xmath21 , any point ( in the marginalized @xmath22 hyperparameter space ) within @xmath18 probability contour , represents a variation that given this variation , the probability of the observed data would be more than @xmath23 . now , depending on where the @xmath24 point stands ( which is the variation associated to the mean function with no change ) in comparison to the @xmath25 confidence ball , one can state whether the given mean function can express the data to a given significance . let us recall that this mean function is the best fit model to the other dataset . + as we have stated previously @xmath26 is the best fit tt angular power spectra obtained from fitting a particular dataset ( say , planck or wmap ) . for a particular dataset this mean function is fixed as the parameters @xmath27 and @xmath28 along with @xmath29 and @xmath30 are fixed to their best fit values . while fitting the @xmath8 to the other dataset we perform markov chain monte carlo ( mcmc ) analysis to put constraints on the crossing hyperparameters . for mcmc we have used the publicly available software cosmomc @xcite . we should note that in our comparison between the two datasets we have kept the polarization sector untouched as planck has not yet published its polarization data and our analysis is based on the temperature data from wmap-9 and planck . since the reported best fit models from wmap-9 and planck both include wmap-9 low-@xmath31 ( up to @xmath32 ) polarization data , we have used same best fit models as our mean functions but beyond that we have only considered the temperature data from the two observations for our likelihood analysis . in this paper , we examine a two - way consistency check . first we use the planck best fit spectrum as the mean function and compare with the wmap-9 data , which clarifies whether wmap-9 data is consistent with planck best fit model . then we consider the wmap 9 year best fit model as the mean function and compare it with planck data to check the reverse ( whether the best fit wmap 9 year model is consistent to planck data ) . it should be mentioned that the two way consistency check is important as if the best fit model from wmap is consistent with planck data , it does not necessarily imply that the best fit model from planck is consistent with wmap-9 data . planck has much higher precision than wmap and has provided the data to much higher multipoles ( @xmath33 ) . hence a theoretical model that fits planck data may also fit wmap-9 data ( if there is no systematic ) since wmap-9 data span over a smaller range @xmath34 , but the converse may not be true . this can be easily understood if we consider the different volumes of the n - dimensional confidence balls ( n is associated to degrees of freedom of the assumed modified model ) when we fit a particular modified model to two datasets with different precision . fitting the same modified model , the volume of the confidence ball for the data with higher precision would be smaller than the volume of the confidence ball for the data with lower precision . hence , a point in the smaller confidence ball associated to the data with higher precision would be included in the confidence ball of the data with lower precision but the reverse may not necessarily be true . overall one can argue that it is always better to choose the mean function from the data with higher precision and broader range since the interpretation of the results would be straightforward as a mean function derived from the broader and more precise data covers the whole range of the other dataset . we would like to note here that in a same line of our study one could possibly use some other approaches too to compare the two data sets . for example gaussian processes can be used to model the two data sets using a single mean function and compare the confidence limits of the hyperparameters @xcite . information field theory with some adjustments can be possibly used as well @xcite . before going to the results section we should state that when we compare the modified models with wmap-9 data we have included the sunyaev - zeldovich effect and in comparing our modified models with the planck data we explored the parameter space of the foreground effects since the foregrounds are quite important on small scales . we have marginalized over 14 nuisance parameters for different foreground and calibration effects . in this section we shall present our results and address the consistency of the two recent cmb datasets . to begin , we test the robustness of the method and perform a self - consistency test . we fit the modified best fit wmap-9 model ( by modified we mean the best fit model , which is our mean function , multiplied by the crossing function ) to wmap-9 data itself and assess the fiducial point in the confidence contours -40pt of the hyperparameters . in fig . [ fig : cntrs9to9 ] we have plotted the marginalized confidence contours of @xmath35 ( left panel ) . it is evident that the mean function ( denoted by the intersection of the two perpendicular black lines at the center of each plot ) lies inside and almost at the center of the 1@xmath36 contour . this indicates that the best fit model from wmap-9 is consistent to wmap 9 year data . it should be noted that in this case any significant deviation from mean function could indicate inconsistency of the concordance model to the data . this result is in agreement with our previous analysis on wmap-9 data @xcite . moreover , as an additional test we have fixed the overall amplitude to its fiducial value ( @xmath37 ) and performed mcmc on @xmath12 and @xmath13 . we have plotted the result in the right panel of fig . [ fig : cntrs9to9 ] . as expected , the size of the contours decreased considerably as the degeneracies between the hyperparameters are lifted by fixing the overall amplitude . here too we find that the fiducial best fit model lies near to the center of 1@xmath36 limits of hyperparameters @xmath22 . these figures indicate that the best fit concordance model to wmap 9 year data is indeed in good agreement with the data . -40pt now we fit the modified mean functions of the two datasets ( best fit concordance models to wmap-9 and planck data multiplied by the crossing function ) to wmap-9 and planck data . in fig . [ fig : cntrsplanckto9 ] we exhibit the confidence contours of the crossing hyperparameters ( in this case the coefficients of the chebyshev polynomial of second order ) when we fit the modified mean function from planck data ( best fit concordance model to planck data multiplied by the crossing function ) to planck and wmap 9 year data . in the left panel the @xmath22 contours are marginalized over @xmath11 allowing the overall amplitudes to vary and right panel is for @xmath37 , fixing the overall amplitudes at their reported values . it is interesting to see that the best fit model to planck data ( mean functions are denoted by the intersection of the two perpendicular black lines at the center of each plot ) is consistent with both observations ( see the left panel ) . the large intersection area between the two confidence contours also indicates that the two observations are very well in agreement . one can interpret this as there being some hypothetical variations , each denoted by a point in the hyperparameter space , that can explain both observations simultaneously with high probabilities if we allow the overall amplitudes to vary . from the right panel of fig . [ fig : cntrsplanckto9 ] , it is evident that by fixing the amplitudes of the angular power spectra to their reported survey , there is a significant inconsistency - more than 3@xmath36 . this result suggests that by fixing the amplitudes , there is no hypothetical model that can simultaneously express both observations with high certainty . assuming that the proposed model(@xmath4cdm ) is correct , this indicates that there is a clear tension between the two datasets . we conclude that the overall amplitudes of the angular power spectra from planck and wmap play a very important role to test if these datasets are consistent . if we allow these amplitudes to vary , there is no tension , but if both observations insist on their reported values of the overall amplitudes there will be an unresolvable tension between the two . to check the extent of the mismatch we have compared the planck best fit model with wmap-9 data varying only the amplitude @xmath11 . the best fit value of @xmath11 in this case is 1.024 which implies the wmap-9 spectrum is about @xmath38 higher than planck . performing our crossing analysis once again fixing @xmath39 we found the that the mean function point @xmath40 lies now completely inside the 1@xmath36 contours of @xmath22 . -40pt in fig . [ fig : clsto9 ] we show the wmap 9 year binned tt spectrum with some theoretical @xmath41 s overplotted . specifically we show the best fit theoretical model from wmap-9 ( red ) and planck ( blue ) . these two curves clearly show that wmap power is higher than planck in high-@xmath31 and this is most evident near the first acoustic peak . however , the green curve , the planck best fit modified with best fit crossing function , shows that difference between planck and wmap-9 data can be addressed using a second order crossing function . we should mention here that a crossing function allows us to match the best fit models from the two datasets up to a statistically indistinguishable level ( depending on the quality of the data ) using an orthogonal basis independent of the cosmological parameters . this in turn allows us to understand how far these models are from each other statistically and if there is any hypothetical variation that can explain both datasets simultaneously . having described the consistency of planck and wmap-9 data using the best fit model from planck as the mean function , we shall demonstrate now our results when we use the best concordance model to wmap 9 year data as the mean function . we should note here that comparing two datasets , it is always better to choose a mean function from the data that has a higher precision and covers a broader range of scales . despite this , it is still of interest to perform the reverse test . we start with a modulation by chebyshev polynomials up to second order . we should emphasize that here we allow the 14 nuisance parameters for foreground and calibration to vary along with the crossing hyperparameters . the contours in fig . [ fig : cntrs9toplanck ] show that the wmap-9 mean function is 3@xmath36 away in the @xmath22 contour plane fitting planck data . when we fix the overall amplitude @xmath37 we find the smaller contour fitting the planck data pushes the mean function point further away . this result clearly indicates that the best fit concordance model to wmap 9 year data is inconsistent at not less than 3@xmath36 level from planck data . this result is completely consistent with the previous analysis . as we have noted earlier , when fitting the same modified model , the area of the marginalized confidence contours ( of the hyperparameters ) for the data with higher precision would be smaller than the marginalized confidence contours for the data with lower precision . hence , even if both data sets are different realizations of the same fiducial model ( and by default consistent ) , a point in the smaller confidence contour associated to the data with higher precision would be necessarily included in the confidence contour of the data with lower precision but the reverse may not be always true . if we look at the plots in fig . [ fig : cntrsplanckto9 ] and fig . [ fig : cntrs9toplanck ] it is evident that the contours fitting planck data is much smaller than the ones fitting wmap-9 . this is simply due to the higher precision and broader range of the planck data . -15pt using the second order chebyshev polynomials to modify the wmap-9 best fit model , we have not yet reached the likelihood which is comparable to planck best fit model ( though our analysis was sufficient to show that the mean function from wmap is not consistent to planck data ) . hence we now modify the spectrum further using the third and the fourth order chebyshev polynomials . while the best fit wmap-9 model had @xmath42 worse with respect to the best fit planck model to planck data , a significant difference , when we use second and third order crossing modifications the @xmath43 reduces to 16 and 7 respectively and with fourth order it reduces further to less than 5 . this indicates that by assuming fourth order chebyshev polynomials we are able to almost match the best fit model from wmap to the best fit model from planck . in fig . [ fig : clstoplanck ] we have plotted the planck data along with some theoretical @xmath41 s overlayed . the red and the blue lines are the wmap-9 and planck best fit models respectively . the green curve is for the modified wmap-9 best fit model with crossing function ( to second order ) set to its best fit form . it is clear that with the crossing function , we can address the wmap-9 and planck discrepancies . the planck high-@xmath31 binned spectrum is shown in the main plot and the low-@xmath31 data is plotted in the inset . it is clearly visible that the modified wmap-9 spectrum matches the planck best fit at all scales . in figure [ fig : cf ] we have plotted the samples of crossing functions within 2@xmath36 confidence contours of the crossing hyperparameters . through the red lines we have shown the samples of second order crossing function when modified mean function from wmap-9 is compared with planck data and in blue lines we have compared the former with the inverse crossing functions ( to second order ) when the modified mean function from planck is compared with wmap-9 data . the smallest scale probed by wmap ( @xmath44 ) is indicated by a vertical line . apart from an amplitude shift the plots successfully capture the scale dependent mismatch between wmap and planck . the agreement of the crossing functions and the inverse crossing functions in this plot reflects the robustness of our analysis . note that the blue curves contain a few lines which are nearly scale invariant , signifying that allowing for an overall amplitude shift the planck mean function can address the wmap-9 data very well without further modification . this agreement basically reflects the results in the left panel of fig . [ fig : cntrsplanckto9 ] . however , most importantly it can be seen that the wmap mean function requires further modifications than just an overall amplitude shift to address the planck data ( the red curves ) , especially , the preferred modifications ( beyond the overall amplitude shift ) are more evident in the smaller scales probed by planck ( @xmath45 ) . for the range of scales probed by wmap ( @xmath46 ) the crossing functions suggest an overall amplitude correction . in this paper , which is the first of the two consistency check papers , we have addressed the consistency of wmap-9 data to the recent planck dataset within the framework of the concordance @xmath4cdm model . in the context of crossing statistic , we adopted the best fit concordance model to wmap 9 year ( planck ) data as the mean function and multiplied it to a crossing function to fit planck ( wmap 9 year ) data . the marginalized confidence contours of the crossing hyperparameters have been used to check the consistency of the two observations . in our analysis we used chebyshev polynomials as the crossing functions . allowing the overall amplitude of the datasets to vary , which is an important assumption , we found that the best fit model to planck data is consistent with wmap 9 year data . however , contrary to this , we found that the best fit model to wmap-9 data is not consistent to planck data at worse than @xmath0 . this is not particularly unexpected since the precision of the planck data is considerably higher than wmap-9 data . since the area of the confidence contours of the hyperparameters using planck data are smaller than the area of the confidence contours using wmap 9 year data , not all points in the confidence contours of wmap-9 ( models consistent to wmap-9 ) should remain within smaller confidence contours of planck . looking at the large overlap of the confidence contours of the crossing hyperparameters ( fig . [ fig : cntrsplanckto9 ] ) we can conclude that allowing the overall shift of the amplitude of the two spectra , there is no strong statistical evidence that planck and wmap-9 data are inconsistent . our results show that there are some hypothetical models from which both data can be simultaneously drawn . in other words , there are some hypothetical models that given these variations , the probabilities of both observed data are high . in the plot of the crossing functions we have shown a family of modifications which can explain the possible discrepancies between the two datasets as a function of angular scales apart from an overall amplitude factor . this highlights the fact that comparing the derived individual cosmological parameters from different data sets might be misleading when the observables are complicated functions of these parameters . the basic six cosmological parameters of the concordance model have a very different nonlinear effect on the angular power spectrum and we use boltzmann codes to derive @xmath47 s . these parameters are degenerate and they are not orthogonal to each other . however , the chebyshev polynomials that we used as our crossing functions have properties of orthogonality which makes it suitable to test the consistency of the data from two observations . we should emphasize that the consistency of planck and wmap 9 year data holds only if we allow the overall amplitudes of the two spectra to vary . fixing the amplitudes at their reported values results in a clear and significant inconsistency of the two data sets at more than @xmath0 and the best fit models from both observations are ruled out strongly by the other observation . in this case , in fact there is no hypothetical variation that can express both datasets simultaneously with high probabilities . this is an important issue to be resolved . resolving this issue , one can claim that the two observations are consistent . d.k.h and a.s wish to acknowledge support from the korea ministry of education , science and technology , gyeongsangbuk - do and pohang city for independent junior research groups at the asia pacific center for theoretical physics . a.s would like to acknowledge the support of the national research foundation of korea ( nrf-2013r1a1a2013795 ) . we also acknowledge the use of publicly available cosmomc in our analysis . d.k.h and a.s would like to thank eric linder and tarun souradeep for useful discussions . we thank stephen a. appleby for his useful comments on the manuscript . we acknowledge the use of wmap-9 data and likelihood from legacy archive for microwave background data analysis ( lambda ) @xcite and planck data and likelihood from planck legacy archive ( pla ) @xcite . g. hinshaw , d. larson , e. komatsu , d. n. spergel , c. l. bennett , j. dunkley , m. r. nolta and m. halpern _ et al . _ , arxiv:1212.5226 [ astro-ph.co ] . p. a. r. ade _ et al . _ [ planck collaboration ] , arxiv:1303.5075 [ astro-ph.co ] . p. a. r. ade _ et al . _ [ planck collaboration ] , arxiv:1303.5076 [ astro-ph.co ] . see , http://www.princeton.edu/act/	within the context of the concordance model of cosmology we test the consistency of the angular power spectrum data from wmap and planck looking for possible systematics . the best fit concordance model to each observation is used as a mean function along with a crossing function with an orthogonal basis to fit the data from the other observation searching for any possible deviation . we report that allowing an overall amplitude shift in the observed angular power spectra of the two observations , the best fit mean function from planck data is consistent with wmap 9 year data but the best fit mean function generated from wmap-9 data is not consistent with planck data at the @xmath0 level . this is an expected result when there is no clear systematic / tension between two observations and one of them has a considerably higher precision . we conclude that there is no clear tension between planck and wmap 9 year angular power spectrum data from a statistical point of view ( allowing the overall amplitude shift ) . our result highlights the fact that while the angular power spectrum from cosmic microwave background observations is a function of various cosmological parameters , comparing individual parameters might be misleading in the presence of cosmographic degeneracies . another main result of our analysis is the importance of the overall amplitudes of the observed spectra from planck and wmap observations . fixing the amplitudes at their reported values results in an unresolvable tension between the two observations at more than @xmath0 level which can be a hint towards a serious systematic . @xmath1 and @xmath2 + _ @xmath1asia pacific center for theoretical physics , pohang , gyeongbuk 790 - 784 , korea + @xmath3department of physics , postech , pohang , gyeongbuk 790 - 784 , korea _ +
the annihilation line of positrons at 511 kev is the brightest gamma - ray line in the galaxy . first observed with a nai scintillator as a @xmath4 476 kev line coming from the galactic center ( gc ) region ( johnson , harnden & haymes , 1972 ; johnston & haymes , 1973 ) , it was subsequently unambiguously identified with a narrow ( @xmath8 kev ) @xmath9 annihilation line using germanium detectors ( leventhal , maccallum , stang , 1978 ) . since then many balloon flights and several space missions have measured the spatial distribution and spectral properties of the line . a summary of the high energy resolution observations of the 511 kev line prior to integral and the first spi / integral results can be found in jean et al . ( 2003 ) and teegarden et al . ( 2004 ) . positrons in the galaxy can be generated by a number of processes , including e.g. radioactive @xmath10 decay of unstable isotopes produced by stars and supernovae , jets and outflows from the compact objects , cosmic rays interaction with the interstellar medium ( ism ) , and annihilation or decay of dark matter particles . an important problem is to determine the total @xmath9 annihilation rate in the galaxy and to accurately measure the spatial distribution of the annihilation radiation . this is a key step in determining the nature of the positron sources in the galaxy . another problem is to measure the annihilation spectrum including the 511 kev line itself and the @xmath11 continuum arising from the decay of ortho - positronium . this information reveals the properties of the ism where positrons are annihilating . here we concentrate on the latter problem and report below the measurements of the @xmath9 annihilation spectrum ( including @xmath11 continuum ) based on spi / integral observations of the gc region over the period from feb . , 2003 through nov . the core of the data set is a deep 2 msec gc observation , carried out as part of the russian academy of sciences share in the integral data . previously reported results on the 511 kev line shape ( jean et al . 2003 ) are based on a significantly shorter data set . we use here a completely independent package of spi data analysis and for the first time report the results on the ortho - positronium continuum measurements based on the spi data ( fig.[fig : spraw ] ) . the imaging results will be reported elsewhere . the structure of the paper is as follows . in section 2 we describe the data set and basic calibration procedures . section 3 deals with the spectra extraction . in section 4 we present the basic results of spectral fitting . in section 5 we discuss constraints on the annihilation medium . the last section summarizes our findings . spi is a coded mask germanium spectrometer on board integral ( winkler et al . , 2003 ) , launched in october 2002 aboard a proton rocket . the instrument consists of 19 individual ge detectors , has a field of view of @xmath416(fully - coded ) , an effective area of @xmath12 and the energy resolution of @xmath42 kev at 511 kev ( vedrenne et al . , 2003 , attie et al . , 2003 ) . good energy resolution makes spi an appropriate instrument for studying the @xmath9 annihilation line . a typical integral observation consists of a series of pointings , during which the main axis of the telescope steps through a 5x5 grid on the sky around the position of the source . each individual pointing usually lasts s few ksec . a detailed description of the dithering patterns is given by winkler et al . for our analysis we use all data available to us , including public data , some proprietary data ( in particular , proposals 0120213 , 0120134 ) and the data available to us through the integral science working team . all data were taken by spi during the period from feb . , 2003 through nov . the choice of this time window was motivated by the desire to have as uniform a data set as possible . the first data used are taken immediately after the first spi annealing , while the last data used were taken prior to the failure of one of the 19 detectors of spi . while analysis of the gc data taken after nov . 2003 is possible , the amount of data ( in public access ) which can be used for background modeling is at present limited . prior to actual data analysis all individual observations were screened for periods of very high particle background . we use the spi anticoincidence ( acs ) shield rate as a main indicator of high background and dropped all observations with an acs rate in excess of 3800 cnts / s . several additional observations were also omitted from the analysis , e.g. those taken during cooling of spi after the annealing procedure . for our analysis we used only single and psd events and when available we used consolidated data provided by the integral science data center ( isdc , courvoisier et al , 2003 ) . as a first step all observations have been reduced to the same gain . trying to keep the procedure as robust as possible we assume a linear relation between detector channels and energies and use four prominent background lines ( ge@xmath13 at 198.4 kev ; zn@xmath14 at 438.6 ; ge@xmath14 at 584.5 kev and ge@xmath14 at 882.5 kev , see weidenspointner et al . , 2003 for the comprehensive list of spi background lines ) to determine the gain and shift for each revolution . while the linear relation may not be sufficient to provide the absolute energy calibration to an accuracy much higher than 0.1 kev over the spi broad energy band , the relative accuracy is high ( see fig.[fig : ecal ] ) . shown in the top panel is the energy of the background 511 kev line as a function of the revolution number . while the deviation from the true energy of the @xmath9 line is @xmath4 0.07 kev , the rms deviation from the mean energy is only 0.0078 kev . the best fit energy of the background line for the combined spectrum of all spi observations within 30of gc is 510.938 kev , compared to the electron rest energy of 510.999 kev . the energies quoted below were corrected for this systematic shift . in the bottom panel of fig.[fig : ecal ] we show the instrument resolution at 511 kev as a function of the revolution number . since the background ( internal ) 511 kev line is kinematically broadened we used two bracketing lines ( at 438 and 584 kev ) to calculate the resolution at 511 kev . the sawtooth pattern clearly seen in the plot is caused by the gradual degradation of the spi resolution due to the detector exposure to cosmic rays and due to the annealing procedure ( around revolution 90 ) which restores the resolution . the net result is that the mean resolution near 511 kev is @xmath4 2.1 kev ( fwhm ) and over the whole data set the resolution changes from @xmath4 2.05 kev to @xmath4 2.15 kev . all the data , reduced to the same energy gain , were then stored as individual spectra ( one per pointing and per detector ) with 0.5 kev wide energy bins . these spectra are used for subsequent analysis . once obvious spikes and flares are removed from the spi data , the background in the remaining `` clean '' data is rather stable and it typically does not vary by more than @xmath410% . however the 511 kev line observed from the gc region produces an excess signal at the level of 1 - 2 per cent of the background line and therefore variations of the background have to be taken into account . ideally one would prefer to have a background model which is based on some accurately measurable quantities ( like charged particles count rate ) so that the background subtraction does not introduce extra noise to the data . since no such model has been provided so far , we use the same data set to build a provisional background model . when doing so one has to bear in mind that the statistical significance of the accumulated data is limited ( when narrow energy bins are considered ) and the model has to be kept as simple as possible to provide a robust result . the simplest background model , which we found acceptable at the present stage of spi analysis , assumes that the background is linearly proportional to the ge detectors saturated event rate and time . an example of observed and predicted background for the 900 - 1200 kev range is shown in fig.[fig : back ] . such a broad band was selected to show variations of the background more clearly . the points in fig.[fig : back ] show actual measurements ( averaged over all 19 spi detectors ) , while the line connects predicted background values . one can see that most of the prominent background variations are well reproduced by the adopted model . however , for some revolutions further improvement of the model is possible once more data , especially blank - field observations , become publicly available . for the purpose of the gc data reduction we then regenerated a background model using all data excluding the central 30(radius ) region centered at gc . the total ( dead time corrected ) exposure of the background fields used is 3.7 msec . to verify the quality of the background model in the energy range of interest ( i.e. around 511 kev ) we used 0.5 msec spi observations of the coma region . while these particular observations were also part of the data set used for the generation of the background model , they contribute only @xmath410% to the total background exposure . applying exactly the same procedure ( described below ) used for the galactic center observations we extracted the spectrum assuming that the spatial distribution of the line flux has a shape of a gaussian with fwhm=6centered at the coma cluster ( fig.[fig : blank ] left panel ) . as for the galactic center observations we used an additional model consisting of a gaussian plus constant ( fig.[fig : blank ] right panel ) . by construction the left plot is basically the difference between the spectrum coming from the coma region and the mean spectrum over the entire background data set which contains many galactic plane pointings with bright sources . therefore , a small negative bias , seen in the left panel of fig.[fig : blank ] , is a natural result . for the right panel ( allowing for an additional free background component constant over all detectors ) this negative bias is not present . for both panels no significant spectral features are present near the 511 kev line . for comparison we show in fig.[fig : blank ] a line at 511 kev with parameters similar to those observed in the gc region . no evidence for spectral features near 511 kev is seen in the coma field , suggesting that the background is removed with a sufficient accuracy . for the gc spectra we use the data obtained when the main axis of the instrument was within 30of the gc direction with an overall exposure time of @xmath43.9 msec ( dead time corrected ) . the spatial model used is a simple gaussian with fwhm ranging from 2 to 26 . using available response files ( results of monte - carlo modeling , described by sturner et al . , 2003 ) the count rate was predicted for every pointing and every spi detector . two models are used to extract the spectra . in the first model the normalization of the model ( in a given energy band ) was then obtained from a simple linear @xmath15 fit to the data : @xmath16 where the summation is over the data set , @xmath17 is the normalization of the model ( free parameter ) , @xmath18 is the model predicted rate , @xmath19 is the observed rate in a single detector during a given pointing , @xmath20 predicted background rate , and @xmath21 is the standard deviation for the observed rate . in order to avoid bias due to the correlation of the observed rates @xmath19 and the errors @xmath21 ( see churazov et al . , 1996 ) we evaluate @xmath21 using the exposure time of the @xmath22th observation and the mean count rate ( in a given detector and given energy band ) averaged over a large number of observations . since the absolute values of the variations of the count rate are small this procedure provides an unbiased estimate of @xmath17 with statistical uncertainty very close to the theoretical limit . the first model is essentially the difference between the flux measured from the gc region ( weighted with a spatial gaussian of a given width ) and the mean spectrum over the data set used for the generation of the background model . for the second model we allow an additional background component ( constant in time and space , but variable in energy ) with a free normalization . i.e. in the second model the predicted count rate in the @xmath22th observation is @xmath23 , where both @xmath17 and @xmath24 are free parameters . while the second model has an obvious advantage compared to the first model ( see e.g. fig.[fig : blank ] ) , especially when trying to detect weak continuum radiation on top of the strong background line , the addition of the second free parameter increases ( for the present data set ) the statistical error in @xmath17 by a factor of 1.6 and 2.6 for the 2and 26gaussian , respectively . we therefore extracted spectra using both models ( hereafter model i and ii ) and checked the results for consistency . the choice of these simple models ( with the minimal number of free parameters ) is primarily driven by a desire to get maximum significance in the resulting spectra . observations during balloon flights ( e.g. harris et al . , 1998 ) and with osse / cgro ( kinzer et al . , 2001 ) and earlier spi imaging analysis ( kndlseder et al . , 2003 ) indicate that a gaussian is a reasonable first approximation for a 511 kev flux excess near gc even though the distribution might be more complicated than a simple gaussian . in principle the assumed spatial distribution mainly affects the absolute normalization of the flux and to a lesser degree the spectral shape ( unless the spectrum shape varies strongly across the studied area of the sky ) . the total flux in the 508 - 514 kev band obtained using both models is shown in fig.[fig : flux ] as a function of the gaussian width . one can see that the flux changes from @xmath25 to @xmath26 as fwhm changes from 2 to 26 . the dotted line shows the behavior of @xmath15 in the model with a free constant ( model ii ) as a function of the gaussian width . within this model the minimum @xmath15 is reached for the fwhm@xmath27 . the absolute value of the @xmath15 for 6gaussian is 0.9992 per degree of freedom ( for 38969 d.o.f . ) . given that the s / n for individual observation is very small , the absolute value of @xmath15 is not a very useful indicator of the acceptability of the model . however its closeness to unity shows that observed variations of count rates are very close to those expected from a pure statistical noise . the width of the distribution suggested by the above analysis ( @xmath46 ) is rather close to the value derived for the central bulge from osse observations ( kinzer et al . , 2001 ) , while the earlier analysis of a shorter spi data set suggested a somewhat broader distribution @xmath49(kndlseder et al . , 2003 ) although consistent with 6within the quoted uncertainties . the behavior of the curves in fig.5 indicates that the 6gaussian does not account for the total flux of 511 kev photons coming from the gc region and an extra component ( broader than 6gaussian ) is needed . this result is robust against various assumptions on the spi internal background . since the topic of this paper is the shape of the annihilation spectrum we will not elaborate on particular spatial models . subsequently we use only the spectra extracted using a 6gaussian . however , we verified that all the major spectral parameters ( except for the overall normalization ) are insensitive to the width of the spatial model . the spectra extracted with a 6gaussian , using two background models ( fixed and free background ) , are shown in fig.[fig : spraw ] and fig.[fig : spback ] . the significance of the narrow line detection is 54.1 and 23.8 @xmath28 respectively ( based on the 508 - 514 energy band ) . when fitting the spectra over the energy range near 511 kev it was assumed that the spi energy resolution is equivalent to the convolution of photon spectra with a gaussian having fwhm @xmath4 2.1 kev . for fitting spectral lines and the weak continuum on the red side of a line one has to verify the normalization and shape of the low energy tail in the spi spectral response . at present a monte - carlo simulated spi energy response matrix ( ` spi_rmf_grp_0003.fits ` , sturner et al . , 2003 ) is available , which is constructed for broad continuum channels . according to this matrix the total off - diagonal tail of the spi response for @xmath4511 kev photons amounts to @xmath430% of the flux . the tail however is very extended ( few hundred kev ) and between 450 kev ( lower energy boundary used for spectral fitting below ) and 511 kev only about 3 - 4% of the line flux are due to off - diagonal response . shown in fig.[fig : lowrsp ] is a typical situation one can expect for the gc spectra . the thin solid line shows a gaussian line with a total intensity of unity . the dashed line is an ortho - positronium continuum with the total flux 4.5 times larger than the line flux ( i.e. the case of annihilation through positronium formation ) . for comparison the dotted line shows the low energy tail of the narrow line estimated from the available response matrices . one can see that indeed the contribution of the tail is very minor ( few per cent relative to the positronium continuum above 400 kev ) . we however routinely included this component in the subsequent spectral fitting , linking its flux to the normalization of the narrow 511 kev line . the impact of the off - diagonal tail on the continuum is even smaller and we neglected the tail in the continuum modeling . e.g. for the positroniuum continuum ( the hardest continuum component used in the model ) an account for the tail contribution above 450 kev can change the normalization by less than 2% . the flux ratio of the positronium continuum and the narrow 511 kev line determines the positronium fraction @xmath29 , where @xmath30 and @xmath31 are the line and continuum flux , respectively . for large flux ratios ( see table 1 ) the positronium fraction is a weak function of the ratio and unless the monte - carlo simulated off - diagonal response is underestimated by a large factor ( larger than @xmath45 ) we do not expect any drastic changes in spectral parameters . for the ortho - positronium continuum we use the spectrum of ore & powell ( 1949 ) . to allow for broadening of the ortho - positronium continuum edge at 511 kev ( spi energy resolution and intrinsic broadening of the edge ) we convolved the continuum with a gaussian having a width from 2.1 to 7 . the fits were found to be fairly insensitive to the exact value of the edge broadening ( unless it is very large ) and instead of introducing an extra free parameter we fixed the width of the ortho - positronium edge at 4 kev . the best fit parameters to the spectra are shown in table [ tab : fit ] . for comparison we show in table [ tab : res ] the spectral parameters derived by various missions in the past . it is clear that the results are in broad agreement . deep integral observations provide the most stringent constraints on the line centroid and on the width of the 511 kev line . [ cols= " < , < , < " , ] [ tab:2gau ] using 2-gaussian models one can conclude that the best fit fraction of the broad gaussian in the total line flux is @xmath430% . one can compare this value with the expected contributions of the broad ( in flight ) and narrow ( thermalized ) components in the warm medium as a function of the ionization degree as shown in fig.[fig : ion ] . one can see that for 8000 k plasma one needs ionization degrees in the range of 0.07 - 0.17 to have an appropriate relation between the broad and narrow components . as is mentioned above the curves shown in fig.[fig : ion ] essentially reflect the changes of the positrons thermalized fraction as a function of the ionization state . the same 2-gaussian models can be used to get more quantitative constraints on the contribution of the cold neutral phase to the annihilation budget . the 90% confidence limit on the contribution of the broad ( 5.5 kev wide ) gaussian is @xmath4 39% . let us assume that this component is due to in - flight positronium formation in a cold neutral gas . about 6% of positrons in the cold gas fall below the positronium formation threshold and annihilate with bound electrons , thus contributing to the narrower component of the line . the conservative assumption that the rest of the positrons are thermalized in a warm ( @xmath32 k ) strongly ionized ( @xmath3350% ) medium and annihilate through the positronium formation implies an upper limit on the fraction of annihilations in the cold phase of 45% . thus an approximately 1:1 mixture of cold neutral and warm ionized phases would produce a spectrum resembling the one observed with integral . if instead a mixture of cold neutral and warm weakly ionized phases is considered , then the limits on the cold component contribution are stronger . if , for instance , the ionization degree of the warm phase is less than 10% then the cold component could not provide more than 10% of the annihilations . for warm phase with an ionization degree of 15% the limit on the cold component is @xmath430% . given the uncertainties in the determination of the mass fractions of various ism phases ( e.g. heiles , 2000 ) the `` natural '' mixture of annihilations in proportion to the phase masses could also be consistent with the data . a more detailed analysis of the integral data ( with an accurate separation of phases contributions and stringent constraints on the line components ) will become possible once more data are accumulated and the knowledge of the instrument background and calibration is improved . more deep observations of the gc region are already planned and it should be possible to achieve at least a factor of 2 - 4 longer exposure of the gc during the next few years . for completeness we estimate the limits imposed by observations on the motions of the medium where positrons are annihilating . first of all - a small statistical error on the line centroid achieved by spi ( @xmath40.075 kev ) corresponds to the uncertainty in velocity of @xmath4 44 km / s . the measured line energy coincides ( within the uncertainties ) with the unshifted line and therefore there is no evidence of bulk motion of the annihilating medium with velocities larger than @xmath440 km / s . note that further improvement in the line statistics will require an accurate account for the earth and solar system motion relative to the gc since the magnitude of the velocities is of the same order . the limits on the gas differential motions are not as strict . consider e.g. the envelope of a star or supernova isotropically expanding with the velocity @xmath34 . the intrinsically monochromatic line will then be observed as a boxy spectral feature with the full width of @xmath35 kev . one can then estimate the effective full width as @xmath36 kev . given that the observed width is @xmath42.4 kev and for some ism phases the intrinsic line can be relatively narrow @xmath41 - 1.5 kev , velocities larger than @xmath37 can be excluded . deep observations of the galactic center region by spi / integral have yielded the most precise parameters of the annihilation line to date . the energy of the line is consistent with the laboratory energy with an uncertainty of 0.075 kev . the width of the annihilation line is constrained to @[email protected] kev ( fwhm ) and the positronium fraction to 94@xmath06% . under a single phase annihilation medium assumption the most appropriate conditions are : the temperature in the range 7000 - 40000 k and the degree of ionization ranging from a few @xmath2 for low temperatures to almost complete ionization for high temperatures . e.g. annihilation of positrons in one of the canonical ism phases - 8000 k gas with an ionization fraction of @xmath410% - would produce an annihilation spectrum very similar to the one observed by integral . under the assumption of annihilation in a multi - phase medium , the contribution of a very hot phase ( @xmath5 k ) is constrained to be less than @xmath48% . neither moderately hot ( @xmath6 k ) ionized medium nor very cold ( @xmath7 k ) neutral medium can make a dominant contribution to the observed annihilation spectrum . further accumulation of the galactic center exposure with integral and improvements in the background knowledge and calibration should make possible detailed fits of the data with the composite annihilation spectrum . accurate separation of various components will place tight constraints on the width and the relative amplitude of annihilation features formed in flight and after thermalization . it will therefore be possible to make an accurate census of the distribution of annihilating positrons over ism phases . simultaneously accurate information on the spatial distribution of the annihilation flux ( especially in the latitudal direction ) can be efficiently combined with the data on the spatial distribution of various ism phases thus constraining the distribution of positron sources and the transport of positrons in the galaxy . we would like to thank spi pis v. schoenfelder , g. vedrenne , j .- roques and v.l . ginzburg , v.v . zheleznyakov , a.m. cherepashchuk , s.a . grebenev and c. winkler for their support . we are grateful to j.berakdar , a.lutovinov and l.vainshtein for useful discussions and the referee , b. j. teegarden , for a very helpful report . this work is based on observations with integral , an esa project with instruments and science data center funded by esa member states ( especially the pi countries : denmark , france , germany , italy , switzerland , spain ) , czech republic and poland , and with the participation of russia and the usa . dermer c. d. , murphy r. j. , 2001 , in : exploring the gamma - ray universe . proceedings of the fourth integral workshop , 4 - 8 september 2000 , alicante , spain . editor : b. battrick , scientific editors : a. gimenez , v. reglero & c. winkler . esa sp-459 , noordwijk : esa publications division , isbn 92 - 9092 - 677 - 5 , 115 guessoum n. , skibo j. g. , ramaty r. , 1997 , the transparent universe , proceedings of the 2nd integral workshop held 16 - 20 september 1996 , st . malo , france . edited by c. winkler , t. j .- courvoisier , and ph . durouchoux , european space agency , 113 heiles c. , 2001 , tetons 4 : galactic structure , stars and the interstellar medium , asp conference series , vol . 231 . edited by charles e. woodward , michael d. bicay , and j. michael shull . san francisco : astronomical society of the pacific , 294	the electron - positron annihilation spectrum observed by spi / integral during deep galactic center region exposure is reported . the line energy ( [email protected] kev ) is consistent with the unshifted annihilation line . the width of the annihilation line is [email protected] kev ( fwhm ) , while the strength of the ortho - positronium continuum suggests that the dominant fraction of positrons ( 94@xmath06% ) form positronium before annihilation . compared to the previous missions these deep integral observations provide the most stringent constraints on the line energy and width . under the assumption of an annihilation in a single - phase medium these spectral parameters can be explained by a warm @xmath1 k gas with the degree of ionization larger than a few @xmath2 . one of the wide - spread ism phases - warm ( @xmath3 k ) and weakly ionized ( degree of ionization @xmath4 0.1 ) medium satisfies these criteria . other single - phase solutions are also formally allowed by the data ( e.g. cold , but substantially ionized ism ) , but such solutions are believed to be astrophysically unimportant . the observed spectrum can also be explained by the annihilation in a multi - phase ism . the fraction of positrons annihilating in a very hot ( @xmath5 k ) phase is constrained to be less than @xmath48% . neither a moderately hot ( @xmath6 k ) ionized medium nor a very cold ( @xmath7 k ) neutral medium can make a dominant contribution to the observed annihilation spectrum . however , a combination of cold / neutral , warm / neutral and warm / ionized phases in comparable proportions could also be consistent with the data . epsf # 1= # 1=1.98 # 1#2=.99 = .99 # 1#2#3=.31 = .31 = .31 [ firstpage ] galaxy : center gamma rays : observations ism : general
the class of regular languages can be defined as the smallest class containing the finite languages and closed under union , concatenation and star . since regular languages are also closed under complementation , one can redefine them as the smallest class containing the finite languages and closed under boolean operations , concatenation and star . in this new formulation , a natural question is that of the _ generalized star height _ of a regular language , which is the minimum number of nested stars required to define the language when boolean operations are allowed . it is not clear who first considered the problem of generalized star height , but mcnaughton and papert reported in their 1971 monograph @xcite that this problem had been open `` for many years '' . there exist regular languages of star height 0 and 1 , but it is not even known whether there exists a language of star height 2 . see http://liafa.jussieu.fr/~jep/problemes/starheight.html . we consider regular languages of star height 0 , which are also called _ star - free_. in 1965 , schtzenberger proved @xcite that a language is star - free if and only if its syntactic monoid is _ group - free _ , that is , has only trivial subgroups . an equivalent condition is that the minimal deterministic automaton of a star - free language is _ permutation - free _ , that is , has only trivial permutations . another point of view is that these automata are _ counter - free _ , since they can not count modulo any integer greater than 1 . they can , however , _ count to a threshold _ , that is @xmath1 _ or more_. such automata are called _ aperiodic , _ and this is the term that we use . the _ state complexity _ of a regular language @xcite is the number of states in the minimal deterministic finite automaton accepting that language . we prefer the equivalent concept of _ quotient complexity _ @xcite , which is the number of distinct left quotients of the language , because quotient complexity has some advantages . the _ quotient complexity of an operation _ in a subclass of regular languages is the maximal quotient complexity of the language resulting from the operation , as a function of the quotient complexities of the operands when they range over all the languages in the subclass . the complexities of basic operations in the class of regular languages were studied by maslov @xcite and yu , zhuang and salomaa @xcite . the complexities of operations were also considered in several subclasses of regular languages : unary @xcite , finite @xcite , ideal @xcite , closed @xcite , prefix - free @xcite , suffix - free @xcite , bifix- , factor- , and subword - free @xcite , and convex @xcite . the complexity of operations can be significantly lower in a subclass of regular languages than in the general case . we prove that this is _ not _ the case for star - free languages , which meet the bounds for regular languages , with small exceptions . it was shown in @xcite that the tight bound for converting an @xmath2-state aperiodic nondeterministic automaton to a deterministic one is @xmath3 . in section [ sec : terminology ] we define our terminology and notation . boolean operations , concatenation , star , and reversal are studied in sections [ sec : boolean][sec : reversal ] , respectively . unary languages are treated in section [ sec : unary ] , and section [ sec : conclusions ] concludes the paper . if @xmath4 is a finite non - empty alphabet , then @xmath5 is the set of all words over this alphabet , with @xmath6 as the empty word . for @xmath7 , @xmath8 , let @xmath9 be the length of @xmath10 , and @xmath11 , the number of @xmath12 s in @xmath10 . a language is any subset of @xmath5 . we use the following set operations on languages : _ complement _ ( @xmath13 ) , _ union _ ( @xmath14 ) , _ intersection _ ( @xmath15 ) , _ difference _ ( @xmath16 ) , and _ symmetric difference _ ( @xmath17 ) . we also use _ product _ , also called _ ( con)catenation _ ( @xmath18 ) and _ star _ ( @xmath19 ) . the reverse @xmath20 of a word @xmath7 is defined by : @xmath21 , and @xmath22 . the _ reverse _ of a language @xmath23 is @xmath24 . _ regular _ languages are the smallest class of languages containing the finite languages and closed under boolean operations , product and star . _ star - free _ languages are the languages one can construct from finite languages using only boolean operations and concatenation . some examples of star - free languages are @xmath25 , @xmath26 , @xmath27 over @xmath28 , and @xmath29 over @xmath30 . we do not write such expressions for star - free languages , but denote them as usual . the _ ( left ) quotient _ of a language @xmath23 by a word @xmath10 is defined as @xmath31 . the number of distinct quotients of a language is called its _ quotient complexity _ and is denoted by @xmath32 . a quotient @xmath33 is _ accepting _ if @xmath34 ; otherwise it is _ rejecting_. a _ deterministic finite automaton ( dfa ) _ is a quintuple @xmath35 , where @xmath36 is a finite set of _ states , _ @xmath37 is a finite _ alphabet , _ @xmath38 is the _ transition function , _ @xmath39 is the _ initial state , _ and @xmath40 is the set of _ final _ or _ accepting states_. as usual , the transition function is extended to @xmath41 . a dfa @xmath42 accepts @xmath43 if @xmath44 , and the _ language _ accepted by @xmath42 is @xmath45 . the _ language of a state _ @xmath46 of @xmath42 is the language @xmath47 accepted by the automaton @xmath48 . if the language of a state is empty , that state is _ empty_. let @xmath49 if @xmath50 , and @xmath51 , otherwise . the _ quotient automaton _ of a regular language @xmath23 is @xmath52 , where @xmath53 , @xmath54 , @xmath55 , @xmath56 , and @xmath57 . since this is the minimal dfa accepting @xmath23 , the quotient complexity of @xmath23 is equal to the state complexity of @xmath23 , and we call it simply _ complexity_. a _ transformation _ of a set @xmath58 into itself is a mapping @xmath59 where @xmath60 for @xmath61 . each word in @xmath5 performs a transformation of the set @xmath36 of states of a dfa @xmath42 . a dfa is _ aperiodic _ if no word performs a permutation , other than the identity permutation , of a subset of @xmath36 . since testing if a dfa is aperiodic is pspace - complete @xcite , we use a subclass of aperiodic automata . without loss of generality , we assume that @xmath62 . a transformation is _ non - decreasing _ if @xmath63 implies @xmath64 . a non - decreasing transformation can not have a non - trivial permutation , and the composition of non - decreasing transformations is non - decreasing . hence a dfa with non - decreasing input transformations is aperiodic . a _ nondeterministic finite automaton ( nfa ) _ is defined as a quintuple @xmath65 where @xmath36 , @xmath37 , and @xmath66 are as in a dfa , @xmath67 is the _ transition function _ and @xmath68 is the _ set of initial states_. if @xmath69 also allows @xmath6 , that is , @xmath70 , we call @xmath71 an _ @xmath6-nfa_. we now consider the quotient complexity of union , intersection , symmetric difference , and difference in the class of star - free languages . the upper bound for these four operations in the class of regular languages is @xmath72 @xcite . [ thm : boolean ] for each of the operations union , intersection , symmetric difference , and difference , there exist binary star - free languages @xmath73 and @xmath23 with quotient complexities @xmath74 and @xmath75 , respectively , that meet the bound @xmath72 . let @xmath28 . we examine union first . for @xmath76 , let @xmath77 and let @xmath23 be any binary star - free language with @xmath78 . then @xmath79 . similarly , if @xmath80 , let @xmath81 and let @xmath73 be any binary star - free language with @xmath82 . then @xmath83 . for @xmath84 , let @xmath85 , and @xmath86 ; then @xmath82 and @xmath78 , and both @xmath73 and @xmath23 are star - free . the quotient automata of @xmath73 and @xmath23 are in fig . [ fig : unionkl ] for @xmath87 and @xmath88 , and their direct product for @xmath14 , in fig . [ fig : union ] . unionkl.eepic union.eepic let @xmath89 , and consider the quotients of @xmath90 by the @xmath72 words @xmath91 , @xmath92 , and @xmath93 ; these quotients @xmath94 correspond to states @xmath95 in the direct - product automaton for @xmath90 . we begin with the rejecting quotients of @xmath90 . first , @xmath96 , and all the other quotients are non - empty . next , if @xmath97 and @xmath98 ( rows 1 to @xmath99 , columns 1 to @xmath100 ) , then the pair @xmath101 of non - empty words belongs to @xmath94 and to no other rejecting quotient . if @xmath102 , then @xmath103 ( rows 1 to @xmath99 , column @xmath2 ) contains @xmath104 , but has no words from @xmath105 . if @xmath106 , then @xmath107 ( row @xmath108 , columns 1 to @xmath100 ) contains @xmath109 , but has no words from @xmath110 . so all rejecting quotients are distinct . now turn to the accepting quotients . for @xmath111 , quotient @xmath112 ( rows 1 to @xmath113 , column @xmath114 ) contains @xmath115 , and this word is not contained in any other quotient @xmath116 with @xmath117 , and @xmath118 has no words from @xmath119 . thus all the quotients in column @xmath114 are distinct . for @xmath120 , @xmath121 ( row @xmath113 , columns 1 to @xmath114 ) contains @xmath122 , and this word is not contained in any other quotient @xmath123 with @xmath124 , and @xmath125 has no words from @xmath126 . thus all the quotients in row @xmath113 are distinct . excluding @xmath127 , each quotient in column @xmath114 contains @xmath12 but not @xmath128 , each quotient in row @xmath113 contains @xmath128 but not @xmath12 , and @xmath127 contains both @xmath12 and @xmath128 . hence all accepting quotients are distinct , and our claim holds for union . for difference , we can use @xmath129 and @xmath23 , where @xmath73 and @xmath23 meet the bound @xmath72 for union , because @xmath130 . for intersection , it was shown in @xcite that the languages @xmath131 and @xmath132 meet the bound @xmath72 . since both languages are star - free , our claim holds for intersection . these languages also meet the bound @xmath72 for symmetric difference @xcite . the tight bound for product of regular languages @xcite is @xmath133 . we show that this bound can be met by star - free languages , with some exceptions . in subset constructions , we use the notation @xmath134 to mean that subset @xmath135 under input word @xmath10 moves to subset @xmath136 . [ thm : product ] there exist quaternary star - free languages @xmath73 and @xmath23 with quotient complexities @xmath74 and @xmath137 , respectively , such that @xmath138 . let the quotient automaton for @xmath73 be @xmath139 , where @xmath140 , @xmath141 , @xmath142 , @xmath143 , and @xmath144 next , let the quotient automaton for @xmath23 be @xmath145 , where @xmath146 , @xmath141 , @xmath147 , @xmath148 , and @xmath149 the automaton @xmath150 for @xmath87 is shown in fig . [ fig : product ] , where the transition labeled @xmath6 should be ignored for now . the automaton @xmath151 for @xmath88 is also shown in fig . [ fig : product ] . if the transition labeled @xmath6 is taken into account and @xmath152 is made a rejecting state , then we have an @xmath6-nfa for @xmath153 . here the initial state is @xmath154 , the set of accepting states is @xmath155 , and the transitions are as shown . product.eepic for @xmath156 , @xmath157 , @xmath158 , @xmath159 , and @xmath160 , denote @xmath161 by @xmath162 . similarly , for @xmath163 , and @xmath164 , denote @xmath165 by @xmath166 . we first show by induction on the size of @xmath135 that all @xmath167 subsets of the form @xmath168 , where @xmath169 , @xmath170 , and @xmath171 , are reachable . when @xmath172 , the set @xmath173 is reached by @xmath174 , for @xmath175 . now suppose we want to reach @xmath176 , where @xmath177 , @xmath178 , @xmath179 , and @xmath180 . let @xmath157 ; by the induction assumption , @xmath168 is reachable . then @xmath181 . thus @xmath176 is also reachable . next , we prove that the @xmath182 subsets of the form @xmath183 , where @xmath135 is any subset of @xmath184 , are reachable . if @xmath76 , then @xmath185 is the initial subset . let @xmath135 and @xmath136 be be as above . then @xmath186 . if @xmath187 , there are two cases . if @xmath188 , then start with @xmath189 , which has already been shown to be reachable we then have @xmath190 if @xmath191 , then start with @xmath192 . now @xmath193 . finally , we show that the @xmath167 subsets of the form @xmath194 , where @xmath195 , and @xmath171 are reachable . we have @xmath196 . in summary , @xmath133 different subsets are reachable . we now prove that all these subsets are pairwise distinguishable . for @xmath197 , state @xmath198 of @xmath199 accepts the word @xmath200 , and state @xmath2 accepts the word @xmath201 ; moreover , each of these words @xmath202 is accepted by only that one state @xmath203 of @xmath199 , and none of these words is accepted by state @xmath204 , if @xmath177 . hence , if @xmath203 is in @xmath205 or in @xmath206 , then @xmath135 and @xmath136 are distinguished by @xmath202 . first , let @xmath207 , and consider @xmath168 and @xmath208 , where @xmath209 , and @xmath135 and @xmath136 differ by state @xmath203 . then @xmath168 and @xmath208 are distinguished by @xmath202 . next , let @xmath210 and take @xmath168 and @xmath211 , where @xmath212 . first apply @xmath213 ; then we reach @xmath214 and @xmath215 , where @xmath216 . then @xmath215 accepts @xmath217 , whereas @xmath214 rejects this word . second , suppose @xmath218 and @xmath135 and @xmath136 differ by state @xmath203 ; then @xmath219 and @xmath220 are distinguished by @xmath202 . third , consider @xmath168 , where @xmath221 and @xmath220 , where @xmath222 and @xmath195 . then @xmath223 is accepted by @xmath220 but not by @xmath224 . since all reachable sets are pairwise distinguishable , the bound is met . [ cor : prod ] there exists a ternary star - free language @xmath23 with quotient complexity @xmath75 , such that @xmath225 . if @xmath226 , the dfa @xmath150 has one state , which is both initial and accepting . now @xmath128 is not needed in the proofs of reachability and distinguishability . right ( left ) ideal _ @xcite is a language @xmath23 satisfying @xmath227 ( @xmath228 ) . if @xmath229 ( @xmath230 ) , then @xmath90 is the right ( left ) ideal generated by @xmath73 . corollary [ cor : prod ] shows that the bound @xmath182 on the quotient complexity of the left ideal generated by a regular language can also be met by a star - free language . if @xmath80 in theorem [ thm : product ] , then either @xmath231 and @xmath232 , or @xmath233 is the right ideal generated by @xmath73 . in the second case , it is known @xcite that @xmath108 is a tight bound for @xmath234 , and that the language @xmath235 is a witness @xcite . since that witness is star - free , the general bound holds also for star - free languages . the case @xmath236 and @xmath237 remains . for @xmath238 , the best bound for product of regular languages is 6 , whereas it is 4 for star - free languages . this was verified with the _ gap _ package _ automata _ @xcite by enumerating all products of 2-state aperiodic automata . there are only three types of inputs possible for a 2-state aperiodic dfa : the input that takes both states to state 1 , the input that takes both states to state 2 , and the identity input . if 1 is the accepting state , then subsets @xmath239 and @xmath240 are not distinguishable . therefore a rejecting quotient of @xmath150 can appear with only three subsets of quotients of @xmath151 in the dfa of @xmath153 instead of @xmath241 , and an accepting quotient , only with one subset instead of two . the complexity is maximized when there is only one accepting quotient of @xmath73 . hence @xmath242 . if 2 is the accepting state , then @xmath243 and @xmath240 are not distinguishable . hence @xmath244 in this case . [ thm : m2 ] there exist ternary star - free languages @xmath73 and @xmath23 with quotient complexities @xmath187 and @xmath245 , respectively , such that @xmath246 . let @xmath247 be the dfa in the proof of theorem [ thm : product ] restricted to input alphabet @xmath248 . let @xmath249 , where @xmath250 for @xmath177 , subset @xmath173 is reached by @xmath174 , @xmath251 , by @xmath252 , and @xmath253 , by @xmath254 . finally , @xmath255 is reached by @xmath256 . this gives @xmath257 subsets . for @xmath177 , @xmath173 accepts no words from @xmath105 , @xmath251 accepts @xmath6 , and @xmath253 accepts @xmath128 but not @xmath6 . hence subsets @xmath168 and @xmath176 with @xmath258 , @xmath259 , and @xmath260 , are distinguishable . next , @xmath168 and @xmath211 with @xmath261 are distinguished by @xmath262 . also , @xmath173 and @xmath253 are distinguished from @xmath255 by @xmath6 , and @xmath251 from @xmath255 by @xmath213 . therefore all @xmath257 subsets are distinguishable . we do not know whether the bound @xmath263 can be reached . however , we have verified with _ gap _ that it can not be reached if @xmath264 . the following dfa plays a key part in finding bounds on the quotient complexities of stars of star - free languages . let @xmath137 , and @xmath265 , where @xmath266 and @xmath267 since all the inputs perform non - decreasing transformations , @xmath268 is aperiodic . in fig . [ fig : star ] , if we ignore state 0 and its outgoing transitions , and also the @xmath6 transition , then the figure shows the automaton @xmath269 . with state 0 and the @xmath6 transition it depicts the @xmath6-nfa of @xmath270 . star.eepic we first study @xmath271 , the restriction of @xmath272 to the alphabet @xmath273 . [ lem : star ] if @xmath137 , and @xmath23 is the star - free language accepted by @xmath271 , then @xmath274 . consider the subsets of @xmath275 in the subset construction of the dfa for @xmath270 . since 0 can only appear in @xmath276 , the remaining reachable subsets are subsets of @xmath36 . the empty subset can not be reached because there is a transition from each state under every letter . since state @xmath114 can not occur without state 1 , we eliminate @xmath277 subsets . because state @xmath114 always appears with state @xmath278 , and state @xmath2 can only be reached from state @xmath114 by @xmath12 , the subset @xmath279 first appears with state 2 , and afterwards , always with a state from @xmath280 ; hence @xmath279 can not be reached . also , @xmath278 and @xmath2 can not appear together without @xmath114 , because @xmath2 can not be reached by @xmath128 , and 1 can not be reached by @xmath12 without including @xmath114 . this eliminates another @xmath281 subsets . so @xmath282 subsets are unreachable , and @xmath283 . now turn to the reachable subsets , and note that subsets @xmath276 and @xmath239 are reached by @xmath6 and @xmath128 , respectively . first , let @xmath284 . all singleton sets @xmath285 are reached by @xmath174 from @xmath239 . now let @xmath157 , @xmath178 , where @xmath286 , @xmath287 , and @xmath288 ; then @xmath289 thus any @xmath290 can be reached from a smaller @xmath291 , and so all subsets in @xmath292 are reachable . second , let @xmath293 ; then @xmath294 , as above , and all subsets in @xmath295 are reachable . third , let @xmath296 . if @xmath172 , then @xmath297 is reachable from @xmath239 by @xmath298 . now suppose @xmath291 is not empty . if @xmath299 , then @xmath300 . so @xmath301 . now , if @xmath302 , then @xmath303 if @xmath304 , then @xmath305 in either case , @xmath306 and all @xmath281 subsets in @xmath307 are reachable . fourth , let @xmath308 . we have shown that @xmath309 , if @xmath310 . since also @xmath311 , we have @xmath312 . hence all @xmath313 subsets @xmath314 in @xmath315 are reachable . altogether , @xmath316 subsets are reachable . it remains to be shown that all the reachable subsets are pairwise distinguishable . state 0 does not accept @xmath317 , while @xmath114 accepts it . each state @xmath318 with @xmath319 accepts @xmath320 and each of these words is accepted by only that one state , and @xmath2 accepts @xmath128 . so any two subsets @xmath135 and @xmath321 are distinguishable . [ thm : stars ] for @xmath322 there exists a quaternary star - free language @xmath23 with @xmath78 such that @xmath323 . for @xmath80 , the tight upper bound is 2 . for @xmath80 , there are only two languages , @xmath25 and @xmath5 , and both are star - free . we have @xmath324 , and @xmath325 . for @xmath237 , there are two star - free unary languages , @xmath6 and @xmath326 , and the bound can not be met if @xmath327 . if @xmath28 , then @xmath328 meets the bound @xmath329 . for @xmath330 , we analyzed all 3-state aperiodic automata using _ gap_. the bound 6 is met by @xmath331 defined above , and bounds 5 and 4 are met by @xmath332 and @xmath333 , respectively . these bounds can not be improved . we now turn to the general case . we will show that the following sets of states are reachable in the nondeterministic automaton @xmath71 ( see fig . [ fig : star ] ) from the initial state 0 : the set @xmath276 , all subsets of @xmath36 containing @xmath334 , and all non - empty subsets of @xmath335 . by lemma [ lem : star ] , we can reach all these subsets by words in @xmath336 , except @xmath279 and the subsets of @xmath335 containing @xmath337 . we have @xmath338 ; hence @xmath337 is reachable . now consider @xmath339 , where @xmath340 . let @xmath341 ; then using @xmath342 we move to @xmath343 , and by @xmath344 we reach @xmath345 . since @xmath339 is reachable by lemma [ lem : star ] , @xmath136 is also reachable . thus we can reach all the subsets of @xmath335 containing @xmath337 by words in @xmath346 . the only set missing now is @xmath279 , and it is reached by @xmath347 . in lemma [ lem : star ] , we have already shown that any two subsets @xmath348 such that @xmath321 are distinguishable by words in @xmath336 . table [ tab : starsummary ] summarizes our results for the quotient complexity of @xmath270 in case @xmath23 is star - free . for unary languages , see section [ sec : unary ] . the figures in boldface type are known to be tight upper bounds . for @xmath349 , we analyzed all 4-state automata with non - decreasing input transformations . automata @xmath350 , @xmath351 , and @xmath352 meet the bounds 12 , 11 , and 9 , respectively . the bounds 11 and 9 can not be improved in the class of automata with non - decreasing input transformations . for the rest , the bounds for @xmath353 and @xmath354 are met by @xmath355 , and @xmath356 , respectively . for regular binary languages , the tight bound for reversal @xcite is @xmath3 . for star - free languages the bound @xmath0 can be met , but with @xmath358 letters . [ thm : reversal ] for each @xmath75 there exists a star - free language @xmath23 with quotient complexity @xmath2 such that @xmath359 . for @xmath80 , the bound is met if @xmath360 , for @xmath237 , if @xmath361 , and for @xmath137 , if @xmath362 . for @xmath80 and @xmath30 , @xmath110 is a witness . for @xmath237 and @xmath28 , @xmath363 is a witness . we have verified using _ gap _ that all star - free languages @xmath23 with @xmath237 satisfy @xmath364 ; hence this bound can not be increased . now let @xmath137 , and let @xmath365 where @xmath266 , @xmath366 , @xmath367 , and @xmath368 since all the inputs perform non - decreasing transformations , @xmath268 is aperiodic . figure [ fig : reversalodd ] shows the nfa @xmath71 which is the reverse of dfa @xmath369 . reversalodd.eepic assume initially that @xmath2 is odd . let @xmath157 be a subset of @xmath36 , and let @xmath370 . then nfa @xmath71 has the following properties : * p1 * if @xmath371 , @xmath372 and @xmath373 , then input @xmath374 deletes state @xmath375 from @xmath135 without changing any of the other states . * p2 * if @xmath371 , @xmath376 , and @xmath377 , then input @xmath374 adds state @xmath375 to @xmath135 without changing any of the other states . we now examine the sets of reachable states in @xmath71 . the set @xmath378 of all the odd states can not be reached . for suppose that it is reached from some set @xmath135 . if it is reached by @xmath12 , then @xmath135 must be a subset of @xmath379 . however , the successor under @xmath12 of such a set @xmath135 also contains @xmath114 if it contains @xmath2 . if we use @xmath128 , then @xmath135 must be a subset of @xmath380 . but then the successor of @xmath135 also contains 2 if it contains 1 . if we use @xmath381 with @xmath318 odd , then @xmath135 must be a subset of @xmath382 , and @xmath135 must also have @xmath383 . but then the successor of @xmath135 also contains @xmath384 , which is even , if it contains @xmath318 . if we use @xmath381 with @xmath318 even , then we also get @xmath318 . if @xmath330 , there are no @xmath381 inputs . set @xmath243 is initial , @xmath239 can be reached by @xmath12 and @xmath385 by @xmath128 . we can get @xmath25 by @xmath386 , @xmath240 and @xmath387 by @xmath317 and @xmath388 , respectively , and @xmath389 by @xmath390 . set @xmath391 is unreachable . so assume @xmath392 . first , consider subsets @xmath135 of @xmath90 , the set of _ middle states _ ; these are subsets of @xmath36 containing neither 1 nor @xmath2 . if @xmath191 start with @xmath393 . by using inputs @xmath381 , delete @xmath114 or not , add @xmath100 or not , etc . , until we reach 2 , which can not be removed by any @xmath381 . if @xmath188 , then @xmath394 has 2 , is a subset of @xmath90 , and so is reachable ; then @xmath135 is reached by @xmath395 from @xmath394 . second , consider subsets @xmath135 of @xmath36 containing 1 but not @xmath2 . if @xmath191 , start with @xmath396 and apply @xmath317 to reach @xmath397 . each state in @xmath396 , except 2 , is without a predecessor in @xmath397 . hence , by using inputs @xmath381 , we can construct any such @xmath135 . if @xmath188 , start with @xmath396 and apply @xmath12 to reach @xmath398 , where @xmath378 of all the odd states . by using inputs @xmath381 , we can construct any such set @xmath135 . third , examine subsets @xmath135 of @xmath36 containing @xmath2 but not 1 . if @xmath191 , start with @xmath396 and apply @xmath128 to reach @xmath399 , and then apply @xmath12 to get @xmath400 . construct any such set @xmath135 using inputs @xmath381 . if @xmath188 , then @xmath135 is a subset of @xmath401 containing @xmath2 . since the set @xmath402 is a subset of @xmath90 , it is reachable ; then @xmath135 is reached by @xmath128 from @xmath402 . finally , consider subsets @xmath135 containing both 1 and @xmath2 . apply @xmath403 to @xmath396 to reach @xmath404 . from this set we can reach any set containing @xmath405 . now assume that @xmath188 . we now show that @xmath406 is reachable for every even @xmath407 in @xmath36 . apply @xmath408 to @xmath396 to reach @xmath409 . if @xmath410 , we are done ; otherwise , delete @xmath100 and @xmath114 by @xmath411 and @xmath412 in that order . then insert @xmath413 and @xmath100 by @xmath414 and @xmath411 in that order . if @xmath415 , we are done ; otherwise , continue in this fashion . if we reach @xmath416 , then @xmath417 , and the process stops . if @xmath88 , then we can reach @xmath418 . from @xmath418 we can get @xmath419 , @xmath420 , and @xmath418 . we are missing only @xmath421 , which is unreachable . if @xmath422 , from @xmath409 we can reach by @xmath381 inputs all the subsets containing @xmath337 but not @xmath243 , except those subsets containing @xmath100 without @xmath114 . from now on , we are interested only in the missing subsets , which are with @xmath337 , without @xmath245 , and have @xmath100 without @xmath114 . then take @xmath423 . from here we can reach all subsets containing @xmath424 without @xmath425 , except those containing @xmath426 without @xmath413 . if @xmath427 , then @xmath428 , and we are missing only @xmath429 , which is unreachable . continuing in this fashion , we can reach all the subsets containing @xmath337 but not 2 , except @xmath378 . together with the case where @xmath191 , we have all the states containing @xmath337 , except @xmath378 . reversaleven.eepic the case where @xmath2 is even is similar . the nfa @xmath71 is shown in fig . [ fig : reversaleven ] for @xmath430 . by an argument similar to that for @xmath2 odd , @xmath378 can not be reached . any subset of @xmath431 can be reached as follows . if @xmath191 , apply @xmath128 to @xmath396 to get @xmath432 , and then @xmath12 to get to @xmath433 . now any subset of @xmath90 containing @xmath245 can be reached by inputs @xmath381 . if @xmath188 , then any subset of @xmath434 can be reached from @xmath432 by inputs @xmath381 . second , consider subsets @xmath135 of @xmath36 containing 1 but not @xmath2 . if @xmath191 , start with @xmath396 and apply @xmath388 to reach @xmath435 . then apply @xmath317 to get @xmath436 . now any subset of @xmath437 containing @xmath240 can be reached by inputs @xmath381 . if @xmath188 , start with @xmath396 and apply @xmath408 to reach @xmath438 . by using inputs @xmath381 , we can construct any subset @xmath135 of @xmath437 containing 1 and not 2 , except the subsets that have @xmath439 without @xmath100 . in case @xmath349 , we can reach @xmath240 , @xmath389 , and @xmath239 , but not @xmath391 . from now on , we are interested only in the missing subsets . as in the even case , we can get subsets containing @xmath439 without @xmath100 by deleting @xmath413 and @xmath100 , adding @xmath426 , and re - inserting @xmath413 . now we are unable to reach states having @xmath440 without @xmath426 . we verify that @xmath406 is reachable for every even @xmath318 with @xmath441 , and continue as in the odd case . we can keep moving this problem to the left , until we reach @xmath416 . then state 4 can not be removed because @xmath378 is not reachable . third , examine subsets @xmath135 of @xmath36 containing @xmath2 but not 1 . if @xmath191 , all such subsets are reachable by inputs @xmath381 from @xmath396 . if @xmath188 , then @xmath135 is a subset of @xmath401 containing @xmath2 . since @xmath402 is a subset of @xmath90 , it is reachable ; then @xmath135 is reached by @xmath128 from @xmath402 . finally , consider subsets @xmath135 containing both 1 and @xmath2 . if @xmath191 , apply @xmath317 to reach @xmath442 . from here we can reach any set containing @xmath405 by inputs @xmath381 . if @xmath188 , we reach @xmath443 from @xmath396 by @xmath12 . from here we can reach any set containing @xmath337 but not 2 by inputs @xmath381 . we still need to verify that all the reachable subsets are pairwise distinguishable . state @xmath318 , and only state @xmath318 , accepts @xmath174 . hence , if @xmath348 and @xmath135 and @xmath136 differ by state @xmath318 , then they are distinguishable by @xmath174 . the case of unary languages is special . for regular unary languages , the tight bounds for each boolean operation @xmath444 , product @xmath153 , star @xmath270 , and reversal @xmath445 are @xmath72 , @xmath72 , @xmath446 , and @xmath2 , respectively @xcite . with the exception of the bound for reversal , these bounds can not be met by star - free unary languages . [ thm : unary ] let @xmath73 and @xmath23 be unary star - free languages with quotient complexities @xmath108 and @xmath2 , respectively . + 1 . for each boolean operation @xmath447 , @xmath448 and the bound is tight . + 2 . for product , @xmath449 , and the bound is tight . + 3 . for the star , the tight bound is @xmath450 + 4 . for reversal , @xmath451 . if a unary star - free language @xmath23 is finite and @xmath78 , its longest word has length @xmath100 ; if it is infinite , the longest word not in @xmath23 has length @xmath100 . one verifies that @xmath448 . the witness languages are @xmath452 and @xmath453 for union and symmetric difference , @xmath454 and @xmath455 for intersection , and @xmath456 and @xmath457 for difference , since @xmath458 . one verifies that @xmath449 , and @xmath459 and @xmath460 are witnesses . + 3 . if @xmath23 is infinite , then @xmath461 , and @xmath462 ; hence @xmath463 . for @xmath464 , the bounds actually met in the infinite case are 1 , 1 , 3 , 4 , 5 , respectively . if @xmath23 is finite , it must contain @xmath298 , and if it has @xmath12 , then @xmath465 . the tight bounds for finite unary star - free languages are 2 , 2 , 1 , 2 , 3 , respectively . hence the tight bounds for all unary star - free languages for the first five values of @xmath2 are 2 , 2 , 3 , 4 , 5 , and the witnesses are @xmath25 , @xmath6 , @xmath466 , @xmath467 , and @xmath468 , respectively . it was shown in @xcite that for a finite unary language @xmath23 , @xmath469 for @xmath392 . for @xmath470 , this bound applies here , and a witness is @xmath471 . + 4 . for unary languages , we have @xmath472 ; hence @xmath473 . we have shown that all the commonly used regular operations in the class of star - free languages meet the quotient complexity bounds of arbitrary regular languages . the only exceptions are in the product for @xmath237 , reversal , and operations on unary languages . brzozowski , j. : quotient complexity of regular languages . in dassow , j. , pighizzini , g. , truthe , b. , eds . : proceedings of the 11th international workshop on descriptional complexity of formal systems , magdeburg , germany , otto - von - guericke - universitt ( 2009 ) 2542 . brzozowski , j. : complexity in convex languages . in dediu , fernau , h. , martin - vide , c. , eds . : proceedings of the 4th international conference on language and automata theory @xmath474lata@xmath475 . volume 6031 of lncs , springer ( 2010 ) 115 brzozowski , j. , jirskov , g. , li , b. : quotient complexity of ideal languages . in lpez - ortiz , a. , ed . : proceedings of the 9th latin american theoretical informatics symposium , @xmath474latin@xmath475 . volume 6034 of lncs , springer ( 2010 ) 208211 brzozowski , j. , jirskov , g. , zou , c. : quotient complexity of closed languages . in ablayev , f. , mayr , e.w . , eds . : proceedings of the 5th international computer science symposium in russia , @xmath474csr@xmath475 . volume 6072 of lncs , springer ( 8495 ) 208211 cmpeanu , c. , culik ii , k. , salomaa , k. , yu , s. : state complexity of basic operations on finite languages . in boldt , o. , jrgensen , h. , eds . : revised papers from the 4th international workshop on automata implementation , @xmath474wia@xmath475 . volume 2214 of lncs , springer ( 2001 ) 6070 han , y.s . , salomaa , k. , wood , d. : operational state complexity of prefix - free regular languages . in sik , z. , flp , z. , eds . : automata , formal languages , and related topics , university of szeged , hungary ( 2009 ) 99115	the quotient complexity , also known as state complexity , of a regular language is the number of distinct left quotients of the language . the quotient complexity of an operation is the maximal quotient complexity of the language resulting from the operation , as a function of the quotient complexities of the operands . the class of star - free languages is the smallest class containing the finite languages and closed under boolean operations and concatenation . we prove that the tight bounds on the quotient complexities of union , intersection , difference , symmetric difference , concatenation , and star for star - free languages are the same as those for regular languages , with some small exceptions , whereas the bound for reversal is @xmath0 . * keywords : * aperiodic , automaton , complexity , language , operation , quotient , regular , star - free , state complexity
the space density of galaxy clusters as a function of cluster mass is a well - known cosmological probe ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and ranks among the best observational tools for constraining @xmath8 , the normalization of the matter power spectrum in the low redshift universe ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? is formally defined as the variance of the linear matter density averaged over spheres with radius @xmath9 . ] the basic idea is this : in the high mass limit , the cluster mass function falls off exponentially with mass , with the fall - off depending sensitively on the amplitude of the matter density fluctuations . observing this exponential cutoff can thus place tight constraints on @xmath8 . in practice , however , the same exponential dependence that makes cluster abundances a powerful cosmological probe also renders it susceptible to an important systematic effect , namely uncertainties in the estimated masses of clusters . because mass is not a direct observable , cluster masses must be determined using observable mass tracers such as x - ray emission , sz decrements , weak lensing shear , or cluster richness ( a measure of the galaxy content of the cluster ) . of course , such mass estimators are noisy , meaning there can be significant scatter between the observable mass tracer and cluster mass . since the mass function declines steeply with mass , up - scattering of low mass systems into high mass bins can result in a significant boost to the number of systems with apparently high mass @xcite . if this effect is not properly modeled , the value of @xmath8 derived from such a cluster sample will be overestimated . one approach for dealing with this difficulty is to employ mass tracers that have minimal scatter , thereby reducing the impact of said scatter on the recovered halo mass function . for instance , @xcite introduced a new x - ray mass estimator , @xmath10 , which in their simulations exhibits an intrinsic scatter of only @xmath11 , independent of the dynamical state of the cluster . use of a mass estimator with such low scatter should lead to improved estimates of @xmath8 from x - ray cluster surveys @xcite . such tightly - correlated mass tracers are not always available . in such cases , determination of the scatter in the mass - observable relation is critical to accurately inferring the mass function and thereby determining cosmological parameters . of course , in practice , it is impossible to determine this scatter to arbitrary accuracy , but since the systematic boost to the mass function is proportional to the square of the scatter @xcite ( i.e. the variance ) , even moderate constraints on the scatter can result in tight @xmath8 constraints . in this paper , we use optical and x - ray observations to constrain the scatter in the mass richness relation for the maxbcg cluster catalog presented in @xcite . specifically , we use observational constraints on the mean mass richness relation , and on the mean and scatter of the @xmath12richness relation , to convert independent estimates of the scatter in the @xmath13 relation into estimates of the scatter in the mass richness relation . an interesting byproduct of our analysis is a constraint on the correlation coefficient between mass and x - ray luminosity at fixed richness . to our knowledge , this is the first time that a correlation coefficient involving multiple cluster mass tracers has been empirically determined . the layout of the paper is as follows . in section [ sec : notation ] we lay out the notation and definitions used throughout the paper . section [ sec : data ] presents the data sets used in our analysis . in section [ sec : rough ] we present a pedagogical description of our method for constraining the scatter in the richness - mass relation , while section [ sec : formalism ] formalizes the argument . our results are found in section [ sec : results ] , and we compare them to previous work in section [ sec : other_work ] . in section [ sec : mf ] , we use our result to estimate the halo mass function in the local universe at @xmath7 , the median redshift of the maxbcg cluster sample , and we demonstrate that our recovered mass function is consistent with the latest cosmological constraints from wmap @xcite . a detailed cosmological analysis of our results will be presented in a forthcoming paper ( rozo et al . , in preparation ) . our summary and conclusions are presented in section [ sec : conclusions ] . we summarize here the notation and conventions employed in this work . given any three cluster mass tracers ( possibly including mass itself ) @xmath14 and @xmath15 , we make the standard assumption that the probability distribution @xmath16 is a bivariate lognormal . the parameters @xmath17 , @xmath18 , and @xmath19 are defined such that @xmath20 note the slopes of the mean and logarithmic mean are the same , as appropriate for a log - normal distribution . the scatter in @xmath21 at fixed @xmath15 is denoted @xmath22 , and the correlation coefficient between @xmath21 and @xmath23 at fixed @xmath15 is denoted @xmath24 . _ we emphasize that all quoted scatters are the scatter in the natural logarithm , not in dex . _ note these parameters are simply the elements of the covariance matrix specifying the gaussian distribution @xmath25 . under our lognormal assumption for @xmath16 , the parameters @xmath17 and @xmath18 are related via @xmath26 in this work , the quantities of interest are cluster mass @xmath27 , x - ray luminosity @xmath28 , and cluster richness @xmath29 . unless otherwise specified , cluster mass is defined as @xmath30 , the mass contained within an overdensity of 500 relative to critical . @xmath28 is the total luminosity in the rest - frame @xmath31 band , and @xmath29 is the maxbcg richness measure @xmath3 , the number of red sequence galaxies with luminosity above @xmath32 within an aperture such that the mean density within said radius is , on average , @xmath33 times the mean galaxy density assuming @xmath34 . likewise , unless otherwise stated all parameters governing the relations between @xmath27 , @xmath28 , and @xmath29 assume that @xmath27 is measured in units of @xmath35 , @xmath28 is measured in units of @xmath36 , and @xmath29 is measured in `` units '' of @xmath37 galaxies . for instance , including units explicitly , the mean relation between cluster mass and richness reads @xmath38 a hubble constant parameter @xmath39 is assumed through out . , our weak lensing masses scale as @xmath40 and the x - ray luminosities as @xmath41 . ] in addition , the weak lensing data presented in this analysis assumed a flat @xmath42cdm cosmology with @xmath43 . the recovered mass function has the standard hubble parameter degeneracy . in this work we use the public maxbcg cluster catalog presented in @xcite , which is an optically selected volume limited catalog of close to @xmath44 clusters over the redshift range @xmath45 $ ] . these clusters were found in @xmath46 of imaging data from the sloan digital sky survey ( sdss , * ? ? ? * ) using the maxbcg cluster finding algorithm @xcite . this algorithm identifies clusters as overdensities of red sequence galaxies . all clusters are assigned a redshift based on the sdss photometric data only , and these redshifts are known to be accurate to within a dispersion @xmath47 . every cluster is also assigned a richness measure @xmath3 , which is the number of red sequence galaxies above a luminosity cut of @xmath32 and within a specified scaled aperture , centered on the brightest cluster galaxy ( bcg ) of each cluster . only clusters with @xmath48 are included in the final catalog . interested readers are referred to @xcite and @xcite for further details . in the interest of economy of notation , from now on we denote the maxbcg richness measure simply as @xmath29 . the relationship between cluster richness and various well known mass tracers has been studied in large , homogeneous samples , such as 2mass @xcite and sdss @xcite . of particular interest to us are the weak lensing measurements of the mean mass as a function of richness , and the x - ray measurements of the mean and scatter of the x - ray luminosity as a function of richness . the former analysis has been carried out by @xcite based on the weak lensing data presented in @xcite , and independently by @xcite . in short , @xcite stacked maxbcg clusters within narrow richness bins , and measured the average weak lensing shear profile of the clusters . these shear profiles were turned into surface mass density contrast profiles using the redshift distribution of background sources estimated with the methods of @xcite and the neural net photometric redshift estimators described in @xcite . then , @xcite fit the resulting profiles using a halo model scheme to obtain tight constraints on the mean mass of maxbcg clusters for each of the richness bins under consideration . the @xcite analysis is very similar in spirit to the one described above . the main differences are the way the source redshift distribution is estimated , and the details of the model fitting use to recover the masses . the differences in the results between these two analysis are discussed in appendix [ sec : m - n_priors ] , where we use them to set priors on the mass richness relation . the measurement of the mean x - ray luminosity of maxbcg clusters has been carried out by @xcite following an approach similar to that pioneered in @xcite . the necessary x - ray data is readily available from the rosat all - sky survey ( rass , * ? ? ? in short , @xcite stacked the rass photon maps @xcite centered on maxbcg clusters in narrow richness bins . the background subtracted stacked photon counts within a @xmath49 aperture were used to estimate the mean x - ray luminosity @xmath28 in the @xmath50 rest frame of the clusters . in addition , @xcite measured the scatter in x - ray luminosity at fixed richness by individually measuring @xmath28 for all maxbcg clusters with @xmath51 . it is worth noting that due to the shallowness of rass , many of the maxbcg clusters are not x - ray luminous enough to be detected individually . however , non - detection and upper limits for @xmath28 for individual systems were properly taken into consideration using the bayesian approach detailed in @xcite , and the recovered mean x - ray luminosity from this baysian analysis was fully consistent with the stacked means . in addition to the data sets above , we use the constraints on the @xmath13 relation from @xcite . these constraints are based on the 400d cluster x - ray survey , a flux limited cluster survey based on rosat pointed observations with an effective sky coverage of 397 @xmath52 @xcite . briefly , @xcite measured both the total soft band x - ray luminosity and the cluster mass for each cluster in the sample . x - ray luminosities are estimated from rosat data , and measure the luminosity in the rest - fram @xmath31 band , extrapolated to infinity assuming standard @xmath53 profiles . cluster masses are estimated based on the values of @xmath54 derived from followup chandra observations , though they note that the results they obtain using different mass tracers such as x - ray temperature and total gas mass are very similar . the @xmath55 relation is itself calibrated based on hydro - static mass estimates . importantly , @xcite explicitly correct for the malmquist bias expected for a flux limited cluster sample , so the @xmath13 relation they derive can be interpreted as the relation one would obtain using a mass limited cluster sample . for this work , we have repeated the analysis in @xcite with a slightly different definition for @xmath28 . in particular , we measure the x - ray luminosity in the rest - frame @xmath31 band within a @xmath56 aperture . the change in band is tailored to match the energy band used by @xcite , which we used to place priors on the @xmath13 relation . it is worth noting that @xcite do not use a @xmath57 aperture , as we do . we have , however , carefully calibrated the scaling between our @xmath28 definition and that of @xcite so as to be able to use their results in our analysis.end a detailed description of our measurements can be found in appendix [ app : lx - n_priors ] . the problem we are confronted with is the following : we have four pieces of observational data , namely * the abundance of galaxy clusters as a function of richness . * the mean relation between cluster richness and mass . * the mean and variance of the relation between cluster richness and x - ray luminosity . * the mean and variance of the relation between cluster x - ray luminosity and mass . from this data , we wish to determine the scatter in mass at fixed richness for the cluster sample under consideration . the basic idea behind our analysis is as follows . consider the probability @xmath58 , which we take to be gaussian in @xmath5 and @xmath4 . this probability distribution is completely specified by the mean and variance of both @xmath27 and @xmath28 at fixed richness , and by the correlation coefficient between @xmath27 and @xmath28 . of these , there are only two quantities that are not already observationally constrained : @xmath59 , the scatter in mass at fixed richness , and @xmath60 , the correlation coefficient between mass and @xmath28 at fixed richness . suppose now that we guessed values for these two quantities , so that the probability distribution @xmath61 is fully specified . given the abundance function @xmath62 , we can use @xmath61 to randomly assign a mass and an x - ray luminosity to every cluster in the sample . we can then select a mass limited sub - sample , and measure the corresponding @xmath13 relation , comparing it to the @xmath13 measurement from @xcite . since the @xmath13 relation we predict depends on our assumptions about @xmath61 , there should only be a small region in parameters space where our predictions are consistent with independent observational constraints on the @xmath13 relation . figure [ fig : contours ] illustrates this idea . to create the figure , we have set every observed parameter of the distribution @xmath61 to the central value of the priors described in appendix [ app : priors ] and summarized in table [ tab : priors ] . we then defined a grid in the two dimensional space spanned by @xmath59 and @xmath60 , and carried through the argument described above . the resulting predictions for the amplitude , slope , and scatter of the @xmath13 relation as a function of @xmath59 and @xmath60 are shown in the figure . we plot contours of constant amplitude , slope , and scatter of the @xmath13 relation as solid , dashed , and dotted lines respectively . the thicker curves correspond to the central values of the priors , while thinner curves demark the corresponding @xmath1 confidence limits . as we can see , all three contours intersect in a finite region of parameter space , indicating good agreement between our weak lensing and x - ray data , and the independent determination of the @xmath13 relation . based on figure [ fig : contours ] , we expect a detailed analysis should constrain our parameters to @xmath63 , and @xmath64 . the rest of this paper is simply a way of formalizing the argument described above in order to place errors on both @xmath59 and @xmath60 . we wish to formalize the above argument in order to place quantitative constraints on the scatter in mass at fixed richness . details of how we go about doing so are presented below . readers interested only in our results can move directly to section [ sec : results ] . as we mentioned above , the key point in our analysis is our ability to compute the amplitude and slope of the mean relation @xmath65 , and the scatter about this mean , as a function of our two parameters of interest : the scatter in mass at fixed richness and the correlation coefficient between @xmath27 and @xmath28 at fixed @xmath29 . let us define @xmath66 , and let @xmath67 denote our parameters of interest . our predictions for the @xmath13 relation as a function of our parameters of interest can be summarized simply as @xmath68 . now , adopting a bayesian framework , a set of priors on @xmath69 is simply a probability distribution @xmath70 . since @xmath69 is a function of @xmath71 , the priors immediately define a probability distribution over @xmath71 given by @xmath72 since we know how to compute both @xmath70 and @xmath68 , we can find any confidence regions for our parameters of interest . the problem we are confronted with , however , is slightly more complicated , in that the functions @xmath69 depend not only on @xmath71 , but also on additional nuisance parameters @xmath73 . indeed , our predictions for the observable parameters of the @xmath13 relation depend on both the abundance function of clusters and @xmath61 . the abundance function can be accurately described by a schechter function ( we explicitly checked a schechter function is statistically acceptable ) , @xmath74 given a schechter fit , our prediction for the @xmath13 relation will also depend on the value of the parameters @xmath75 and @xmath76 . likewise , the distribution @xmath61 also depends on the amplitude and slope of the means @xmath77 and @xmath78 , as well as the scatter in @xmath28 at fixed @xmath29 . all in all , we have six additional nuisance parameters @xmath79 . let @xmath80 denote the full set of parameters . the priors from the @xmath13 relation define a probability distribution over @xmath81 given by @xmath82 since we have a total of 8 parameters , and only three observables from the @xmath13 relation , it is obvious that the above likelihood function will result in large degeneracies because the parameters are under - constrained . if one has priors @xmath83 in the nuisance parameters , however , the probability distribution @xmath84 in the parameters of interest is given by @xmath85 this equation allows us to compute @xmath84 , and therefore place constraints on our parameters of interest . in practice , we will ignore the determinant term in the probability distribution defined in equation [ eq : prob ] . this is because the function @xmath86 is estimated using a monte carlo approach , implying that accurate numerical estimates of the jacobian @xmath87 would be too computationally intensive to be performed . fortunately , the determinant typically introduces only slight modulations of the likelihood , so we do not expect our results to be adversely affected by this . we estimate the probability distribution @xmath84 using a monte carlo approach . ignoring an overall normalization constant and setting @xmath88 , we have @xmath89 where @xmath90 for @xmath91 through @xmath92 are random draws of the nuisance parameters @xmath90 , drawn from the prior distribution @xmath93 . we set @xmath94 as our default value ( see below for further discussion ) . the prior distributions for our nuisance parameters are characterized by a statistical and a systematic error . the former is modeled as gaussian and the latter using a top - hat distribution . thus , given a prior of the form @xmath95 a random draw is obtained by setting @xmath96 where @xmath97 is drawn from a gaussian of zero mean with a covariance matrix defined by the statistical errors , and @xmath98 is drawn from a top hat distribution that is non - zero only for @xmath99 . the probability distribution @xmath100 used in equation [ eq : probestimator ] is the product of the likelihoods @xmath101 for each of the @xmath13 parameters @xmath102 . the probability for each @xmath13 parameter is given by the convolution of the top - hat and gaussian distributions defined by the statistical and systematic errors of @xmath103 , so that @xmath104\ ] ] where @xmath105 note that the above equations are appropriate only when the various @xmath13 parameters are uncorrelated , so it is important to place the priors at the pivot point of the @xmath13 relation ( @xmath106 ) . this explains why table [ tab : priors ] quotes a prior on @xmath107 rather than on @xmath108 alone . we also need to specify how the function @xmath109 is evaluated . we do this using a monte carlo approach . given @xmath71 and @xmath73 , we generate @xmath110 mock clusters in the richness range @xmath111 $ ] . we then randomly draw mass and x - ray luminosity values for each of these clusters based on the distribution @xmath61 , and select a mass limited subsample of clusters using a mass cut @xmath112 with @xmath113 ( the reason for this particular value is explained below ) . using a least squares fitting routine , we find the best fit line between @xmath4 and @xmath5 . this defines both @xmath114 and @xmath115 . the scatter @xmath116 is defined as the root mean square fluctuation about the best fit line . using equation [ eq : probestimator ] and the function @xmath109 defined above , we evaluate the probability distribution @xmath84 along a grid of points in @xmath117 $ ] and @xmath118 $ ] with @xmath119 grid points per axis . a full run of our code then requires we perform @xmath120 monte carlo integrals with @xmath94 points in each integration . each draw also requires us to evaluate the function @xmath109 , which in turn requires generating a mock catalog with @xmath110 clusters , so the procedure as a whole is computationally expensive . to increase computational efficiency , for each monte carlo evaluation of @xmath84 we generate a single cluster catalog that is used to estimate the likelihood at every grid point . this correlates the values of @xmath121 along our grid , but does not otherwise adversely affect our results . our monte carlo approach requires that both the number of clusters in the random catalogs @xmath122 and the number of times the likelihood function is evaluated @xmath92 is sufficiently large to achieve convergence . our default values for @xmath122 and @xmath92 were selected to ensure the recovered likelihood is accurate to within a dispersion of @xmath123 inside high likelihood regions . the error in the recovered likelihood increases with decreasing likelihood , but even in the tails of the distributions our estimates are accurate to about @xmath124 . this was explicitly tested by running a coarse grid with our default values for @xmath92 and @xmath122 , and by repeating the analysis with both of these parameters increased by a factor of two . , one needs to generate cluster catalogs with @xmath125 clusters in order for the contours to appear smooth by eye . however , @xmath110 is a sufficient number of clusters for our analysis , since we only require that the noise in the likelihood be much smaller than the width of the priors . since the latter are quite wide , even relatively noisy estimates of the @xmath13 relation are sufficient for constraining the marginalized distribution . ] finally , we emphasize that it is necessary to explicitly check whether our results are sensitive to the @xmath126 cut applied to the maxbcg clusters sample . in particular , when selecting a mass limited subsample of clusters , we need to ensure that the mass limit @xmath127 be sufficiently large that the number of clusters with @xmath128 and @xmath112 is insignificant . we have explicitly checked that for our adopted low mass cut @xmath129 our results are robust to the richness cut @xmath126 by repeating the analysis in a coarse grid using an @xmath130 richness cut instead . we find that the likelihood estimates in both cases are in agreement to within the expected accuracy of our monte carlo approach . the priors used in our analysis are summarized in table [ tab : priors ] . we follow the notation @xmath131 where @xmath132 is the central value , @xmath133 is the 1@xmath134 statistical error on the parameter @xmath135 marginalized over all other parameters , and @xmath136 is the systematic error . in all cases , we model statistical errors as gaussian , and we include known covariances between different parameters . systematic errors are assumed to follow top - hat distributions , and the final prior distribution is given by the convolution of these two functions . @xmath137 & @xmath138 + @xmath75 & @xmath139 + @xmath140 & @xmath141 + @xmath142 & @xmath143 + @xmath144 & @xmath145 + @xmath146 & @xmath147 + @xmath148 & @xmath149 + @xmath150 & @xmath151 + @xmath152 & @xmath153 + @xmath154 & @xmath155 [ tab : priors ] we believe that the priors contained in table [ tab : priors ] are fair , that is , they are neither overly aggressive nor overly conservative . a detailed discussion of our priors can be found in appendix [ app : priors ] . figure [ fig : lkhd ] shows the @xmath156 and @xmath1 probability contours for the parameters @xmath59 and @xmath60 . the likelihood peak occurs at @xmath157 and @xmath158 . the marginalized means are @xmath159 and @xmath160 . we wish to determine whether the breadth of the likelihood region in figure [ fig : lkhd ] is limited by uncertainties in the scaling relations of maxbcg clusters , or by uncertainties in the @xmath13 relation . to do so , we repeat our analysis with two new sets of priors : for the first , we use a tight @xmath161 statistical prior on all nuisance parameters , but let the @xmath13 parameters float . the second set of priors uses a tight @xmath161 prior on each of the @xmath13 parameters , but floats all other nuisance parameters with the original priors . we find that using tight priors on our nuisance parameters has negligible impact on the likelihood regions recovered from our analysis . on the other hand , the confidence regions obtained with the tight @xmath13 priors , shown in figure [ fig : lkhd ] as dashed curves , are tighter than those derived from our original analysis . thus , the dominant source of error in our analysis is the uncertainty in the values of the @xmath13 parameters . this can be easily understood based on figure [ fig : contours ] . we can see from the figure that the uncertainty in @xmath60 is largely due to the prior on the scatter in @xmath28 at fixed @xmath27 , which is already tight and thus does not change between our fiducial prior and our tight priors . on the other hand , we can see that both the amplitude and slope priors cut - off regions with high scatter . tightening these priors excludes a larger section of parameter space , and results in the tighter contours observed in figure [ fig : lkhd ] . figure [ fig : marg ] shows the marginalized probability distributions for @xmath59 and @xmath60 . the solid curves correspond to our original analysis , while the dashed curves illustrate the results one expects assuming our hypothetical tight priors for the @xmath13 relation parameters . we find that the logarithmic scatter in mass at fixed richness and the correlation coefficient between @xmath5 and @xmath4 are @xmath162 assuming our hypothetical tight @xmath13 priors , the constraints become @xmath163 and @xmath6 . we emphasize that these latter constraints are only meant as a guide to the accuracy one could achieve with this method if the @xmath13 relation were known to about @xmath164 accuracy . it is evident from our results that cluster richness is not as effective a mass tracer as x - ray derived masses . indeed , even total ( i.e. not core - core excluded ) x - ray luminosity is a more faithful mass tracer than the adopted richness measure of the maxbcg catalog , as demonstrated both by the smaller scatter and the very large correlation coefficient . note that the latter indicates that , at fixed richness , over - luminous clusters are almost guaranteed to also be more massive than average . this is an important result which forms the basis for a concurrant paper in which we improve our richness estimates by demanding tighter correlations in the @xmath12richness relation @xcite . there are not many previous results against which our measurements of scatter in mass at fixed richness may be compared . one possible reference point is the upper limit based on the error bar in the weak lensing mass estimates of @xcite . more specifically , assuming that the error in @xmath77 is entirely due to the intrinsic scatter in @xmath27 at fixed @xmath29 , it follows that the error in the mass is simply @xmath165 where @xmath166 is the observed error and @xmath62 is the number of clusters with richness @xmath29 . for the richest bin , which provides the tightest constraint , @xcite find @xmath167 . the bin contains @xmath168 clusters , so an upper limit to the scatter in mass at fixed richness is @xmath169 . figure [ fig : marg ] shows that our results easily satisfy this upper limit on the scatter . the only other measurement of the scatter in mass at fixed richness for maxbcg clusters is that found in @xcite . these scatter estimates are obtained as follows : first , @xcite select all maxbcg clusters whose central galaxy has a spectroscopic redshift . they then bin the clusters in richness , and compute the velocity relative to the bcg of every galaxy member with spectroscopic data . the recovered velocity distribution of galaxies is found to be non - gaussian . assuming that the velocity distribution of galaxies of halos of fixed mass is exactly gaussian , and that the observed non - gaussianity is entirely due to mass - mixing within a richness bin , @xcite estimate the scatter in mass at fixed richness based on the observed non - gaussianity of the velocity distribution . an updated version of the results from @xcite can be seen in figure [ fig : compare ] . the only difference between this plot and the corresponding figure in @xcite is that here we have made used of the additional spectroscopic data from the sdss data release 6 @xcite , which results in tighter error bars . also shown in the figure as a horizontal band is the @xmath1 confidence region from our analysis . as we can see , our scatter estimate appears to be systematically lower than that of @xcite , a discrepancy first noted in ( * ? ? ? * more on the relation between our work and theirs below ) . such a bias is not entirely unexpected , as we now know that a significant fraction of cluster have their bcgs miss - identified , a problem that was not yet known and was therefore unaccounted for at the time the @xcite results came out . to get a better understanding of how our results and those of @xcite compare , we can use our results along with the miscentering probability model from @xcite to predict the scatter that @xcite observed given this miscentering systematic . we proceed as follows . first , we use our best fit model for the abundance distribution to generate a mock catalog with @xmath170 clusters with @xmath126 . each of these clusters is assigned a mass by drawing from the @xmath171 distribution defined by the values of @xmath59 corresponding to the two @xmath1 confidence limits on @xmath59 . these assigned masses are then turned into velocity dispersions using the scaling relation from @xcite . at this point , we have a cluster catalog where each cluster has a richness and a velocity dispersion . if a cluster is miscentered , we expect that in most cases the new center will be a cluster galaxy . assuming this is the case , and that bcgs are at rest at the center of a cluster , the velocity dispersion of cluster galaxies relative to random satellites will be a factor of @xmath172 high than relative to the bcg . using the miscentering model described in @xcite for @xmath173 , the probability that a cluster of richness @xmath29 be correctly centered , we randomly label clusters as properly centered or miscentered , and boost their `` observed '' velocity dispersion for those clusters labeled as miscentered by the expected amount . the clusters are assigned a new mass based on their `` observed '' velocity dispersions , and the corresponding scatter in the @xmath174 relation is estimated . we repeat this procedure @xmath175 times in order to compute the mean systematic correction due to miscentering . our predictions for the scatter values observed by @xcite are shown in figure [ fig : compare ] with dashed lines , and correspond to the @xmath1 confidence interval from our analysis . we see that miscentering introduces a richness dependent correction that boosts the scatter in the recovered velocity dispersion and places it in significantly better agreement with the data from @xcite . the agreement with the @xcite data is an interesting result . perhaps the single most difficult systematic effect that had to be addressed in the @xcite analysis is the validity of the assumption that non - gaussianities in the velocity distribution of stacked clusters are entirely due to mass - mixing is a valid . the reasonable agreement between our results and those of @xcite suggests that their assumption is indeed justified , though a robust conclusion will have to wait until a more detailed analysis is performed , especially given the possibility of velocity bias of the galaxy population ( i.e. if satellite galaxies have a velocity dispersion different from that of the dark matter ) . the analysis in this work is also very closely related to that of @xcite . @xcite sought to constrain the @xmath13 relation of clusters by fitting the scaling of @xmath78 with @xmath77 . however , as recognized in @xcite , in order to fully interpret their result in terms of the traditional definition of the @xmath13 relation , i.e. the mean x - ray luminosity at fixed mass , one needs to know both the scatter in mass at fixed richness , and the corresponding correlation coefficient with @xmath28 . given that these two quantities are unknown , but that the @xmath13 relation is already constrained from x - ray surveys , it seems reasonable to suggest that a better use of the lensing and x - ray data of maxbcg clusters is to use our knowledge of the @xmath13 relation to constrain the scatter in mass at fixed richness and the corresponding correlation coefficient , as was done in this work . our work differs from the ideas presented in @xcite in another significant way . while our analysis employs only @xmath176 and @xmath62 , @xcite used the halo mass function @xmath177 and the probability distribution @xmath178 to interpret their measurements . this has the important drawback that in doing so , one needs to assume a cosmological model in order to compute the halo mass function , rendering their interpretation cosmology dependent . by focusing on the quantities that are directly observable , i.e. @xmath62 and @xmath176 , we are able to avoid this difficulty . the price we pay for this is that rather than constraining the scatter in richness at fixed mass , which is the more directly relevant quantity from a cosmological perspective , we constrain instead the scatter in mass at fixed richness . while this makes implementing such a constraint a little more cumbersome in a cosmological analysis , the fact that the constraint itself is cosmology independent is obviously of paramount importance . as mentioned in the introduction , to obtain an unbiased estimate of the halo mass function based on the observed cluster richness function requires that we understand the scatter between cluster richness and halo mass . given our lognormal assumption , and the fact that the mean mass richness relation is already known from weak lensing , our measurement of the scatter in this scaling relation fully determines the probability distribution @xmath171 . thus , we are now in a position to determine the halo mass function of the local universe with the maxbcg cluster catalog . let us define then @xmath179 as the number of halos within a logarithmic mass bin of width @xmath180 centered about @xmath181 , @xmath182 given our cluster catalog and @xmath171 , we construct an estimator @xmath183 for @xmath184 by randomly drawing a mass from @xmath171 for each halo in the cluster catalog , and then counting the number of halos within the logarithmic mass bin centered about @xmath181 . note that since the mass of each cluster is a random variable , our mass function estimator @xmath183 is itself a random variable . the mean and correlation matrix of @xmath183 can easily be obtained by making multiple realizations of @xmath183 , and averaging the resulting mass functions . in practice , we also need to marginalize our results over uncertainties in @xmath171 and over uncertainties in the richness function @xmath62 . to do so , we randomly draw the parameters @xmath185 , and then resample of the cluster richness function to obtain a new estimate of @xmath184 . the whole procedure is iterated @xmath186 times , and the mean and covariance matrix of the number counts in each of our logarithmic mass bins is computed . is large enough for our results to be insensitive to the maxbcg richness cut @xmath126 . ] figure [ fig : mf ] shows the mass function recovered through our analysis . to turn our number counts into a density , we assumed a wmap5 cosmology @xcite , with @xmath43 and @xmath187 , and a photometric redshift error @xmath188 ( used for computing the effective volume of the sample ) . the diamonds correspond to our estimated means , and the error bars are the square root of the diagonal elements of the correlation matrix . we emphasize that the error bars are heavily correlated . the mean and covariance matrix of the recovered halo mass function can be found in appendix [ app : mfdata ] . also shown in figure [ fig : mf ] with dotted lines are the halo mass functions at @xmath7 predicted by wmap5 assuming the @xcite mass function . for both curves , we set all cosmological parameters to the central values reported in @xcite , except for @xmath8 , which is set to @xmath189 for the upper curve and @xmath190 for the lower curve . these two values define the @xmath1 confidence interval for @xmath8 in @xcite . as we can see , the mass function recovered from our analysis is fully consistent with the wmap5 cosmology , though it seems to push for values of @xmath8 on the high end of their allowed region . a detailed cosmological analysis of our data will be presented in a subsequent paper ( rozo et al , in preparation ) . we have shown that by combining the information in the maxbcg richness function , the mean richness - mass relation , the mean and scatter of the @xmath12richness relation , and the mean and scatter of the @xmath13 relation , we can constrain both the scatter in mass at fixed richness for maxbcg clusters , as well as the correlation coefficient between mass and @xmath28 at fixed richness . we find @xmath162 these constraints are dominated by uncertainties in the @xmath13 relation , and can be significantly tightened if our understanding of the @xmath13 relation improves . we also found our results are consistent with those presented in @xcite once miscentering of maxbcg clusters is taken into account . our lower limit on the correlation between @xmath27 and @xmath28 at fixed richness constitutes the first observational constraint on a correlation coefficient involving two different halo mass tracers . note that the large correlation between @xmath28 and @xmath27 implies that @xmath28 - even without core exclusion - is a significantly better mass tracer than the maxbcg richness estimator ( i.e. at fixed richness , over - luminous cluster are nearly always more massive ) . this is an important result , which we use in a concurrent paper to help us define new richness estimators that are better correlated with cluster mass @xcite . using our results , and assuming @xmath43 and @xmath39 , we have estimated the halo mass function at @xmath7 , corresponding to the median redshift of the cluster sample . we find that our recovered mass function is in good agreement with the mass function predicted by @xcite for the wmap5 cosmology @xcite . a detailed cosmological analysis will be presented in a forthcoming paper ( rozo et al , in preparation ) . our work sheds new light on the interrelationship of bulk properties of massive halos . we have used weak lensing , x - ray luminosities , and optical richness estimates to constrain the scatter in the richness - mass relation , which can lead to improved cosmological constraints . in principle , one could also turn this question around , and , assuming cosmology , we could constrain the scatter in the richness - mass relation , which would then allow us to place constraints on the amplitude , slope , and scatter of the @xmath13 relation . such an analysis would be interesting in that , by doing so , one could compare the predicted amplitude of the @xmath13 relation to that derived from hydrostatic mass estimates , thereby directly probing the amount of non - thermal pressure support in galaxy clusters . note that even though this question can also be directly addressed by comparing weak lensing and x - ray mass estimates of individual clusters , the analysis suggested here would benefit from having small uncertainties , whereas projection effects result in rather noisy weak lensing mass estimates for individual systems . the authors would like to thank alexey vikhlinin for providing them with the full covariance of the @xmath13 parameters from their analysis of the 400d cluster sample . er thanks david weinberg and chris kochanek for useful discussions and their careful reading of the manuscript . esr would like to thank the tabasgo foundation . tm and jh gratefully acknowledge support from nsf grant ast 0807304 and doe grant de - fg02 - 95er40899 . ae gratefully acknowledges support from nsf grant ast-0708150 . rhw was supported in part by the u.s . department of energy under contract number de - ac02 - 76sf00515 and by a terman fellowship at stanford university . this project was made possible by workshop support from the michigan center for theoretical physics . 53 natexlab#1#1 , j. k. et al . 2008 , , 175 , 297 , s. , kneib , j .- p . , czoske , o. , soucail , g. , smail , i. , ebeling , h. , & smith , g. p. 2005 , , 434 , 433 , s. , soucail , g. , kneib , j .- , czoske , o. , ebeling , h. , hudelot , p. , smail , i. , & smith , g. p. 2007 , , 470 , 449 . 2007 , , 470 , 449 , m. r. et al . 2007 , , 669 , 905 , r. a. , vikhlinin , a. , hornstrup , a. , ebeling , h. , quintana , h. , & mescheryakov , a. 2007 , , 172 , 561 , h. 2006 , , 653 , 954 , x. , kochanek , c. s. , & morgan , n. d. 2007 , , 658 , 917 , j. et al . 2008 , arxiv e - prints , 803 , astro - ph/0803.0586 , a. e. 1990 , , 363 , 349 , a. e. et al . 2008 , , 672 , 122 , c. s. , white , s. d. m. , efstathiou , g. , & davis , m. 1990 , , 351 , 10 , m. d. , yee , h. k. c. , majumdar , s. , barrientos , l. f. , hoekstra , h. , hall , p. b. , & infante , l. 2007 , , 655 , 128 , z. , mohr , j. j. , & holder , g. p. 2001 , , 553 , 545 , j. p. 2004 , , 609 , 603 , j. p. , & arnaud , k. a. 1991 , , 372 , 410 , h. 2007 , , 379 , 317 , g. , haiman , z. , & mohr , j. j. 2001 , , 560 , l111 , w. , & kravtsov , a. v. 2003 , , 584 , 702 , d. e. et al . 2007 , arxiv e - prints , 709 , astro - ph/0709.1159 , b. c. 2007 , , 665 , 1489 , b. p. et al . 2007 , , 660 , 239 . 2007 , , 660 , 221 , a. v. , vikhlinin , a. , & nagai , d. 2006 , , 650 , 128 , m. , cunha , c. e. , oyaizu , h. , frieman , j. , lin , h. , & sheldon , e. s. 2008 , arxiv e - prints , 801 , astro - ph/0801.3822 , m. , & hu , w. 2004 , , 70 , 043504 . 2005 , , 72 , 043006 , a. , hoekstra , h. , babul , a. , & henry , j. p. 2008 , , 384 , 1567 , r. , seljak , u. , & hirata , c. m. 2008 , arxiv e - prints , 805 , astro - ph/0805.2552 , r. et al . 2008 , , 386 , 781 , b. j. 2007 , , 668 , 772 , d. , vikhlinin , a. , & kravtsov , a. v. 2007 , , 655 , 98 , h. , lima , m. , cunha , c. e. , lin , h. , frieman , j. , & sheldon , e. s. 2008 , , 674 , 768 , e. , scott , d. , & white , m. 2001 , , 325 , 77 , e. et al . 2006 , , 369 , 2013 , t. h. , & bhringer , h. 2002 , , 567 , 716 , e. , dodelson , s. , & frieman , j. a. 2004 , , 70 , 083008 , e. , wechsler , r. h. , koester , b. p. , evrard , a. e. , & mckay , t. a. 2007 , arxiv astrophysics e - prints , astro - ph/0703574 , e. et al . 2007 , arxiv astrophysics e - prints , astro - ph/0703571 , e. , zentner , a. r. , bertone , g. , & chen , j. 2006 , , 639 , 573 , e. , et al . 2008 , arxiv e - prints , 809 , astro - ph/0809.2797 , e. s. et al . 2008 , , 387 , l28 . 2008 , , 675 , 1106 , p. , bhringer , h. , collins , c. a. , & guzzo , l. 2003 , , 398 , 867 , e. s. et al . 2007 , arxiv e - prints , 709 , astro - ph/0709.1153 , r. , evrard , a. e. , bhringer , h. , schuecker , p. , & nord , b. 2006 , , 648 , 956 , j. l. , kravtsov , a. v. , klypin , a. , abazajian , k. , warren , m. s. , yepes , g. , gottlober , s. , & holz , d. e. 2008 , arxiv e - prints , 803 , astro - ph/0803.2706 , a. et al . 2008 , arxiv e - prints , 805 , astro - ph/0805.2207 , w. et al . 1999 , , 349 , 389 , w. , boller , t. , englhauser , j. , freyberg , m. , & supper , r. 2001 , in astronomical society of the pacific conference series , vol . 225 , virtual observatories of the future , ed . r. j. brunner , s. g. djorgovski , & a. s. szalay , 234+ , d. g. , adelman , j. , anderson , j. e. , anderson , s. f. , annis , j. , & the sdss collaboration . 2000 , , 120 , 1579 , y .- y . , finoguenov , a. , bhringer , h. , kneib , j .- p . , smith , g. p. , czoske , o. , & soucail , g. 2007 , , 467 , 437 , y .- y . , finoguenov , a. , bhringer , h. , kneib , j .- , smith , g. p. , kneissl , r. , okabe , n. , & dahle , h. 2008 , , 482 , 451 our estimates of the @xmath13 parameters depend on the abundance function of maxbcg clusters , which is observationally determined , but not known to infinite precision . here , we fit the observed abundance function using a schechter function , such that the mean number of clusters @xmath191 of richness @xmath29 is @xmath192 the amplitude @xmath193 is chosen such that the total number of clusters exactly equals the observed number of clusters . we set this normalization condition because we are interested only in the shape of the richness function , and not in its amplitude . the fits are done by maximizing the likelihood of the observed distribution , binned in bins of width @xmath194 . we assume that the probability distribution of observed @xmath195 clusters in a bin of richness @xmath29 is poisson , with @xmath196 for numerical purposes , we cut the distribution at @xmath197 , which is sufficiently large to not affect our fits . we emphasize that we use the above likelihood only to define estimators for @xmath76 and @xmath75 , since , as discussed below , both goodness of fit and errors in the parameter estimation are obtained through monte carlo simulation . the richness distribution is fit over the range @xmath126 by maximizing the log - likelihood function using an amoeba routine . to estimate our errors , we follow a monte carlo approach and resample the observed richness function @xmath198 times . we find that the parameters @xmath76 and @xmath75 are significantly correlated , with the probability distribution being gaussian in @xmath75 and @xmath137 . the best fit parameters are @xmath199 with a correlation coefficient @xmath200 to assess goodness of fit , we generate @xmath198 mock catalogs with as many clusters as the real data from the probability distribution specified by @xmath201 and @xmath202 . we compute the likelihood for each of these mock catalogs , and compare the corresponding likelihood distribution to that observed in the real data . we find that our fit is statistically acceptable . the most significant systematic error affecting our measurements of the shape of the richness function are completeness and purity variations in the cluster catalog . @xcite have shown that the maxbcg catalog is over @xmath203 pure and complete for @xmath126 . here , we take a conservative approach , and consider the change in the best fit parameters assuming the observed counts are rescaled by a completeness / purity correction factor @xmath204 given by @xmath205 this corresponds to a @xmath124 decrease in the observed counts at @xmath206 while holding the counts at @xmath207 constant . upon refitting the data after this correction we find systematic offsets @xmath208 which we adopt as our systematic error . note the systematic offsets are allowed to be both positive and negative , since the correction multiplier @xmath204 above could easily be larger than unity rather than smaller than unity . our priors on the @xmath174 relation are based on the results presented in @xcite , @xcite , and @xcite . to assign our priors , we first compare the results of these two works as a means of assessing systematic uncertainties in the mass parameters . we then focus exclusively on the @xcite results to place our final priors on the @xmath174 relation . the latter choice reflects the fact that @xcite report weak lensing mass estimates for several mass definitions , among them @xmath30 , the relevant quantity in the @xmath13 relation of @xcite . let us then begin by discussing the @xcite results first . while @xcite quote a power - law fit for the mean mass at fixed @xmath77 , this fit is based a non - public version of the maxbcg catalog that extends to a richness of @xmath209 ( the catalog for clusters with @xmath210 is not public ) . since the maxbcg catalog is only known to be highly complete and pure in the range @xmath126 , we have refit the @xcite masses restricting ourselves to the range @xmath211 . this slightly lower cut is necessary due to the richness binning in @xcite . we find that the mass @xmath212 within a 180 overdensity threshold relative to mean matter density is @xmath213 with a correlation coefficient @xmath214 between the amplitude and slope parameters . @xcite preformed a similar but independent weak lensing analysis of the maxbcg clusters , though using @xmath215 as their mass variable . they find @xmath216 to compare against the @xcite values , we use the @xcite mass conversion formulae to find an approximate power law relation between @xmath215 and @xmath212 over the range @xmath217 . we find @xmath218 , which is only a @xmath219 correction . applying this correction , we find that the corresponding @xmath174 parameters from @xcite are @xmath220 we find that the slopes of the @xcite and @xcite results are nearly identical , but that the masses of @xcite are systematically higher by @xmath221 . this difference can be traced back to how the lensing critical surface density for each of the two works is estimated . in general , lensing masses are proportional to the quantity @xmath222 , where @xmath223 is the lensing critical surface density , and the average is to be computed over the source redshift distribution . given multi - band photometric data @xmath224 for each galaxy , one way to compute @xmath225 is to use a photometric redshift estimator @xmath226 , and then assume that the true source redshift distribution is identical to the photometric redshift distribution . @xcite have shown that such a simple approach typically results in biased lensing mass estimates , but they also demonstrate that it is possible to achieve unbiased results using the probability distribution @xmath227 . the weak lensing analysis in @xcite , on which the results from @xcite are based , falls somewhere in between these two approaches . while @xcite does in fact make use of photometric redshifts , they do not simply assume that the source redshift distribution is identical to the photometric redshift distribution . rather , they construct a probability distribution @xmath228 , and use this probability to estimate @xmath225 . as it turns out , evaluating @xmath225 in this way leads to results that are nearly identical to those obtained by simply setting @xmath229 . thus , even though the approach used in @xcite is more sophisticated than the simple case considered in @xcite , we expect the @xcite results to be biased but correctable as prescribed in @xcite . this correction amounts to a boost of the lensing masses by a factor of @xmath230 . the statistical error bar in the correction is added in quadrature to the statistical error bar from our fit , which results in @xmath231\times ( n/20)^{1.18\pm 0.09 } \label{eq : johnston_mass}\ ] ] these new values for the @xcite data are in considerably better agreement with those of @xcite . there remains , however , a systematic @xmath164 difference between the two amplitudes , as well as a small difference @xmath232 between the two slopes . a possible culprit for this systematic @xmath164 offset is the difference in how miscentering is accounted for in the data models . the word miscentering refers to the fact that when finding clusters , one will inevitably find clusters that are improperly centered , either due to failures of the cluster finding algorithm , or simply because there is no obvious center of the cluster based on its optical image . such offsets between the true and assigned centers are problematic because if a cluster is miscentered , the corresponding lensing signal is weakened , resulting in systematically low mass estimates . to determine whether the remaining offset between @xcite and @xcite is consistent with differences in the miscentering model , we refit our data assuming no errors on the miscentering corrections . we find @xmath233 with a correlation coefficient @xmath234 . note that these errors are smaller than the errors quoted before , as they should be , given that this new fit does not marginalize over a wide range of miscentering models . by subtracting the two sets of errors in quadrature , we find that the miscentering priors adopted in @xcite correspond to an error @xmath235 in the amplitude and @xmath161 in the slope . thus , the @xcite mass measurements are well within the centering error included in the analysis of @xcite . nevertheless , it is unclear whether miscentering can in fact account for the difference between the @xcite and @xcite results . more specifically , @xcite also performed their analysis including the @xcite model for miscentering , and find after applying the centering correction their best fit @xmath236 relation becomes @xmath237 comparing this to equation [ eq : johnston_mass ] , we find including a miscentering correction in the @xcite analysis increases the tension between the two results . moreover , it suggests that the difference between the two results is due to some other form of systematic difference between the two analysis pipelines . in light of this , we opt for introducing a systematic correction to the @xcite results of @xmath238 and @xmath239 for the amplitude and slope respectively . we also introduce systematic errors of the same magnitude as this systematic correction , so that our final result is @xmath240(n/20)^{1.13\pm 0.09\ ( stat ) \pm 0.05\ ( sys)}.\ ] ] note the central values of the original @xcite analysis ( corrected for photometric redshift bias ) as well as the @xcite results both with and without miscentering corrections are all encompassed by our systematic error . now , in this work we are interested more in the @xmath241 ( henceforth simply @xmath174 ) relation than in the @xmath242 relation , since it is the former mass which is accessible to x - ray studies . to constrain the @xmath174 relation we use the quoted @xmath30 mass measurements from @xcite , re - scaling their @xmath243 errors to @xmath30 by assuming the relative errors are constant . a fit to the data results in @xmath244 ( n/40)^{1.11 \pm 0.08}\ ] ] with a correlation coefficient @xmath245 . we now boost this expression by factor @xmath246 due to the photometric redshift bias correction , and add the systematic corrections @xmath238 and @xmath239 to the amplitude and slope respectively as per our discussion of the @xmath236 relation . we also include a systematic error on the amplitudes and slopes of this same magnitude . we obtain @xmath247 the final systematics we consider here are the purity and completeness of the sample . now , as long as the completeness is not correlated with mass , completeness should not in any way bias the recovered parameters of the @xmath174 relation , though it obviously affects the error bars due to lower statistics . the same can not be said of purity . if only a fraction @xmath248 of the clusters are actually good matches to real halos in the universe , then a fraction @xmath249 of the clusters will have a lensing signal that is significantly different from the mean signal . as an extreme case , we can consider what happens if a fraction @xmath249 of the clusters had no mass associated with them . in that case , the observed mean mass is simply @xmath250 where @xmath251 is the true mean , so one should boost the observed masses by a factor of @xmath252 to obtain an unbiased estimate . for @xmath253 , this amounts to an increase in @xmath140 of magnitude @xmath254 . now , @xcite showed that the purity of the maxbcg cluster sample is expected to be above @xmath203 over the range or richnesses considered here , and the increase in @xmath140 quoted above is undoubtedly an overestimate of the necessary correction since even false cluster detections will have excess mass associated with them . in light of this , we have adopted a one - sided systematic error bar @xmath255 to take into account the impact of purity in the recovered @xmath174 relation . the error bar is one sided since we expect impurities will tend to decrease the observed mean mass . we can , however , turn this prior into a normal double - sided prior by including a systematic correction @xmath256 to the central value , and setting the systematic error bar to the same magnitude as the central value shift . we can also get a rough estimate for the systematic error on the purity by assuming that the quoted systematic error in the amplitude should be made only in the limit of high or low richness . if that were the case , using the fact the slope is measured over a decade of richness values , the corresponding slope would be @xmath257 which amounts to a systematic offset @xmath258 . these systematic error bars are added linearly to our previous systematic error . our final set of priors for the @xmath174 relation is @xmath259 with a correlation coefficient @xmath245 between the two statistical errors . the priors in the @xmath260 relation come from repeating the analysis described in @xcite , but with @xmath28 defined as the x - ray luminosity in the 0.5 - 2.0 kev band , and corrected for aperture effects . as in @xcite , we restrict this analysis to clusters with @xmath261 . we begin by measuring the stacked mean @xmath260 relation and scatter on a fixed @xmath262 scale @xmath263 where we have measured @xmath28 in units of @xmath264 , with a pivot point of @xmath265 . we emphasize that the scatter determined above is the total scatter in the observed @xmath260 relation that can not be attributed to poisson uncertainties in the rosat photon counts . in particular , the quoted scatter is affected by possible point source contamination , agn activity , cool cores , cluster mergers , etc . there are multiple systematic errors that can affect the derived parameters for the @xmath260 relation . these include photometric redshift errors , evolution of the richness parameter @xmath29 , uncorrelated point sources , cluster mis - centering , and cluster agn and cool cores . in addition , we need to account for the fraction of cluster flux lost due to our finite aperture and the rass psf , in order to compare our results with the luminosity measurements of @xcite . we shall now discuss each of these possible systematic effects . @xcite find that the accuracy of the maxbcg photo - z estimates is high enough such that any biases are insignificant relative to the statistical uncertainty of the parameter determinations , and can thus be safely ignored . however , @xcite did find significant redshift evolution in the @xmath260 relation , well above the expected self - similar evolution . similar redshift evolution is found in @xcite ; the reason for the systematic undercounting of cluster members at high redshift is explained in @xcite . we have estimated the effect of this redshift evolution on our derived scatter parameter via a simple monte carlo , and confirm that although the apparent evolution is strong , it is insignificant relative to the intrinsic scatter . therefore , we may also safely ignore this possible systematic effect . we now take a combined approach to the systematic effects due to cluster mis - centering , a finite aperture , the rass psf and uncorrelated point sources . the first three effects are strongly related , in that they all tend to scatter cluster photons out of our initial fixed @xmath262 aperture , and these may affect the normalization , slope , and scatter in the @xmath260 relation . uncorrelated point sources should not affect the mean relation because the large number of stacked sources smooths out the foreground and background . however , when uncorrelated point sources are aligned with individual clusters they may increase the measured scatter by boosting the apparent @xmath28 . we have estimated the effects of these systematics by running a monte carlo with simulated rass data on top of random backgrounds selected from the area of the rass photon map that overlaps with the maxbcg mask . we first resample the maxbcg richness function 100 times . each cluster is given a redshift drawn from the maxbcg redshift distribution , as well as a random postion on the sky selected from the area of the rass survey that overlaps with the maxbcg mask . after we select the richest 1000 clusters in each realization , each cluster is given a luminosity based on the mean relation from eqn . [ eqn : lxnmean ] and an input intrinsic scatter , @xmath266 . each cluster luminosity is then converted to a number of photon counts according to the rass exposure at the given point , and scattered by poisson uncertainties . then , each cluster is given a position offset according to the maxbcg miscentering distribution described in ( * ? ? ? * see 4.3 ) . the cluster profiles are assumed to follow a @xmath53 model , @xmath267 . to ensure we are on similar footing as @xcite , we randomly assign each cluster @xmath53 model parameters uniformly in the range @xmath268 and @xmath269 . finally , the photons are scattered according to the rass psf , following the method of ( * ? ? ? we then calculate the stacked mean relation and scatter as described in @xcite . figure [ fig : lnsystematics ] summarizes the results from our systematic tests . the x - axis shows the input intrinsic scatter , @xmath270 . the y - axis shows the ratio of the input parameter to output parameter for the normalization @xmath144 ( circles ) , slope @xmath146 ( diamonds ) , and scatter @xmath148 ( squares ) . we note that when @xmath271 then @xmath272 , which can not be displayed on the plot . this is consistent with our expectation that uncorrelated sources may boost the observed scatter , especially with low intrinsic scatter . overall , we find that ( a ) the slope @xmath146 is not significantly biased ; ( b ) at moderate to large scatter ( @xmath273 ) the intrinsic scatter @xmath148 is not significantly biased ; and ( c ) the output normalization @xmath144 must be boosted by a factor of @xmath274 to account for the flux lost to miscentering , the finite aperture , and rass psf effects . our priors become then @xmath275 in addition to these corrections , we also need to take into account systematic uncertainties due to purity and completeness in the sample . just as with the weak lensing mass estimates , completeness should not affect the measured @xmath260 relation , whereas purity will tend to suppress the x - ray luminosity at fixed richness . following the same procedure as in appendix [ sec : m - n_priors ] , we derive systematic errors @xmath276 and @xmath277 , which we add linearly to our previous systematic error estimates . finally , we have repeated our scatter analysis using not just the 1000 richest clusters , but also the 2000 richest clusters , in which case we find @xmath278 . to take into account this variation in our analysis , we also introduce a systematic error @xmath279 . our final set of priors is @xmath280 as discussed in section [ sec : rough ] , our analysis hinges on the fact that we can use prior knowledge about the @xmath13 relation to constrain the @xmath174 relation . here , we use the results of @xcite to put priors on the @xmath13 relation , which may be summarized as , as appropriate for the maxbcg sample . ] @xmath281 we report a prior on @xmath107 because at @xmath282 the @xmath13 parameters derived from the @xcite sample are correlated . to decouple them , one needs to shift to the statistical pivot point @xmath283 and introduce the scatter dependence quoted above ( vikhlinin , private communication ) . these constraints are derived from chandra observations of clusters in the 400d cluster catalog @xcite , which allowed @xcite to measure @xmath54 and thereby infer cluster mass using the @xmath55 relation . this relation was itself calibrated on a cluster subsample for which masses were derived using the standard hydrostatic equilibrium argument . this last point is very important , since simulations suggest that hydrostatic mass estimates of clusters may be biased low by @xmath284 ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? one way to calibrate such uncertainties is to compare weak lensing mass estimates to hydrostatic mass estimates . there are several examples of this type of approach . for instance , @xcite have performed such an analysis using the weak lensing mass estimates of @xcite , and find @xmath285 a similar analysis has been carried out by @xcite , who used the weak lensing mass estimates of @xcite and their own analysis of chandra public data to obtain @xmath286 . finally , using xmm x - ray observations and the weak lensing data of @xcite , @xcite , and @xcite , @xcite find @xmath287 . @xcite also note , however , that a histogram of @xmath288 peaks at a ratio of @xmath289 , and that clusters in the tails of the distribution tend to have tight error bars , possibly biasing the error weighted ratio . in light of this , we have opted for a `` middle of the road '' approach , and introduce a correction factor @xmath290 . our corresponding prior is @xmath291 estimating systematic errors in @xmath152 and @xmath154 is difficult . for instance , comparisons with weak lensing masses are not an effective way of assessing systematics because weak lensing mass estimates are so noisy : trying to fit a power law relation between @xmath292 and @xmath293 results in very large errors for the slope of the relation . one alternative is to consider multiple studies of the @xmath13 relation in order to asses how sensitive the recovered parameters are to the analysis pipeline . unfortunately , such an excercise is far from trivial . one difficulty is the fact that there is very little agreement on the meaning of @xmath28 , with many works focusing on core - excised and/or core - corrected bolometric x - ray lumunisoties ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? even among those works that also explore the @xmath13 relation when @xmath28 is a soft x - ray band luminosity ( e.g. * ? ? ? * ; * ? ? ? * ) , there are still important differences in the aperture used to estimate @xmath28 . in principle , we could attempt to convert between the various definitions of @xmath28 to try to compare the works against each other , but many of these @xmath13 measurements are affected by malmquist bias , making comparisons to the @xcite results difficult . one work that does constrain the the soft x - ray band , non - core excised , malmquist bias corrected @xmath13 relation is @xcite . unfortunately , the energy band they use is slightly different from that of of @xcite , so even here comparison is not trivial . we expect , however , that at least the scatter and slopes of the @xmath13 relation will not be strongly affected by the minor differences between the two @xmath28 definitions . given our purposes , the interesting thing about the @xcite results is that they use a very different methodology for constraining the @xmath13 relation . in particular , they assume knowledge of cosmological parameters , and then use the observed cluster x - ray luminosity function to constrain @xmath294 . assuming their `` compromise cosmology '' , which they argue gives the best results , they find @xmath295 and @xmath296 . these values are in excellent agreement with those of @xcite , and suggest that placing additional systematic errors in the @xmath13 parameters is not really necessary at this point . table [ tab : mfdata ] presents the mean and covariance matrix of the mass function data derived from our analysis . these results represent the state of the art mass function measurements at low redshift from optically derived cluster catalogs . we emphasize we assumed @xmath43 and @xmath39 , so appropriate rescaling must be applied if the results are to be compared against significantly different cosmologies . note that the covariance matrix data in table [ tab : mfdata ] is normalized such that the diagonal entries are the fractional error @xmath297 , while the off diagonal entries are the correlation coefficients @xmath298 . we present the data in this way since it is easier to understand when expressed this way . the actual values for the covariance matrix are easily reconstructed from the data in the table . @xmath299 & 3.22 & 3.70 & 4.26 & 4.91 & 5.65 & 6.50 & 7.49 & 8.62 & 9.92 & 11.42 & 13.14 & 15.12 & 17.41 & 20.04 & 23.07 & 26.55 & 30.56 + 3.22 & 7.90e-7 & 0.22 & 0.82 & 0.77 & 0.72 & 0.66 & 0.61 & 0.55 & 0.50 & 0.44 & 0.39 & 0.35 & 0.30 & 0.25 & 0.21 & 0.18 & 0.15 & 0.12 + 3.70 & 5.61e-7 & 0.82 & 0.24 & 0.79 & 0.76 & 0.71 & 0.67 & 0.62 & 0.57 & 0.52 & 0.46 & 0.41 & 0.36 & 0.31 & 0.27 & 0.23 & 0.19 & 0.16 + 4.26 & 3.92e-7 & 0.77 & 0.79 & 0.27 & 0.77 & 0.74 & 0.70 & 0.66 & 0.61 & 0.57 & 0.52 & 0.46 & 0.41 & 0.36 & 0.31 & 0.26 & 0.22 & 0.18 + 4.91 & 2.70e-7 & 0.72 & 0.76 & 0.77 & 0.30 & 0.75 & 0.72 & 0.68 & 0.64 & 0.59 & 0.55 & 0.50 & 0.44 & 0.38 & 0.34 & 0.29 & 0.24 & 0.20 + 5.65 & 1.82e-7 & 0.66 & 0.71 & 0.74 & 0.75 & 0.35 & 0.72 & 0.69 & 0.65 & 0.61 & 0.57 & 0.52 & 0.46 & 0.41 & 0.36 & 0.30 & 0.26 & 0.22 + 6.50 & 1.21e-7 & 0.61 & 0.67 & 0.70 & 0.72 & 0.72 & 0.41 & 0.68 & 0.65 & 0.62 & 0.57 & 0.52 & 0.47 & 0.42 & 0.37 & 0.32 & 0.27 & 0.22 + 7.49 & 7.93e-8 & 0.55 & 0.62 & 0.66 & 0.68 & 0.69 & 0.68 & 0.47 & 0.64 & 0.60 & 0.57 & 0.52 & 0.47 & 0.42 & 0.37 & 0.32 & 0.27 & 0.23 + 8.62 & 5.11e-8 & 0.50 & 0.57 & 0.61 & 0.64 & 0.65 & 0.65 & 0.64 & 0.55 & 0.59 & 0.55 & 0.51 & 0.47 & 0.41 & 0.37 & 0.32 & 0.27 & 0.23 + 9.92 & 3.24e-8 & 0.44 & 0.52 & 0.57 & 0.59 & 0.61 & 0.62 & 0.60 & 0.59 & 0.65 & 0.53 & 0.49 & 0.45 & 0.40 & 0.36 & 0.31 & 0.26 & 0.22 + 11.42 & 2.03e-8 & 0.39 & 0.46 & 0.52 & 0.55 & 0.57 & 0.57 & 0.57 & 0.55 & 0.53 & 0.76 & 0.47 & 0.43 & 0.38 & 0.34 & 0.30 & 0.25 & 0.21 + 13.14 & 1.25e-8 & 0.35 & 0.41 & 0.46 & 0.50 & 0.52 & 0.52 & 0.52 & 0.51 & 0.49 & 0.47 & 0.92 & 0.40 & 0.36 & 0.32 & 0.28 & 0.24 & 0.20 + 15.12 & 7.61e-9 & 0.30 & 0.36 & 0.41 & 0.44 & 0.46 & 0.47 & 0.47 & 0.47 & 0.45 & 0.43 & 0.40 & 1.11 & 0.33 & 0.30 & 0.26 & 0.22 & 0.19 + 17.41 & 4.53e-9 & 0.25 & 0.31 & 0.36 & 0.38 & 0.41 & 0.42 & 0.42 & 0.41 & 0.40 & 0.38 & 0.36 & 0.33 & 1.36 & 0.26 & 0.24 & 0.20 & 0.17 + 20.04 & 2.63e-9 & 0.21 & 0.27 & 0.31 & 0.34 & 0.36 & 0.37 & 0.37 & 0.37 & 0.36 & 0.34 & 0.32 & 0.30 & 0.26 & 1.74 & 0.21 & 0.19 & 0.16 + 23.07 & 1.48e-9 & 0.18 & 0.23 & 0.26 & 0.29 & 0.30 & 0.32 & 0.32 & 0.32 & 0.31 & 0.30 & 0.28 & 0.26 & 0.24 & 0.21 & 2.22 & 0.17 & 0.14 + 26.55 & 8.29e-10 & 0.15 & 0.19 & 0.22 & 0.24 & 0.26 & 0.27 & 0.27 & 0.27 & 0.26 & 0.25 & 0.24 & 0.22 & 0.20 & 0.19 & 0.17 & 2.88 & 0.12 + 30.56 & 4.41e-10 & 0.12 & 0.16 & 0.18 & 0.20 & 0.22 & 0.22 & 0.23 & 0.23 & 0.22 & 0.21 & 0.20 & 0.19 & 0.17 & 0.16 & 0.14 & 0.12 & 3.88	we measure the logarithmic scatter in mass at fixed richness for clusters in the maxbcg cluster catalog , an optically selected cluster sample drawn from sdss imaging data . our measurement is achieved by demanding consistency between available weak lensing and x - ray measurements of the maxbcg clusters , and the x - ray luminosity mass relation inferred from the 400d x - ray cluster survey , a flux limited x - ray cluster survey . we find @xmath0 ( @xmath1 cl ) at @xmath2 , where @xmath3 is the number of red sequence galaxies in a cluster . as a byproduct of our analysis , we also obtain a constraint on the correlation coefficient between @xmath4 and @xmath5 at fixed richness , which is best expressed as a lower limit , @xmath6 . this is the first observational constraint placed on a correlation coefficient involving two different cluster mass tracers . we use our results to produce a state of the art estimate of the halo mass function at @xmath7 the median redshift of the maxbcg cluster sample and find that it is consistent with the wmap5 cosmology . both the mass function data and its covariance matrix are presented .
ii pegasi is an sb1 binary system for which the primary component was determined to be of spectral class k2 - 3 iv - v by @xcite . it was classified as an rs cvn system by @xcite . the first photometric light curves were obtained by @xcite , who found variability with a period of approximately 6.75 d and interpreted the asymmetric light curve in terms of rotational modulation due to large , cool starspots . in 1986 september , the difference between maximum and minimum light for the @xmath1 filter reached 0.5 mag , implying a projected spot area coverage of the visible hemisphere at minimum light on the order of 50% @xcite . on the basis of high - quality radial velocity measurements , @xcite determined the revolution period of the binary to be @xmath3 d. the same authors performed a detailed model atmosphere analysis of high - resolution and high signal - to - noise ccd spectra , obtaining values for the photospheric temperature and surface gravity of the primary star of @xmath4 k and @xmath5 , with @xmath6 expressed in cgs units . these values correspond to a k2 iv star of mass @xmath7 . they estimated the radius of the primary as @xmath8 and the inclination to be @xmath9 on the assumption that the rotational axis is perpendicular to the orbital plane . based on the fact that the secondary star is unseen at all wavelengths and thus has luminosity at least 100 times smaller than that of the primary , they estimated the secondary to be an m0-m3 red dwarf . ii peg is among the most active rs cvn systems , and it is one of a small number of binaries in which the h@xmath10 line is always seen in emission @xcite . recently , @xcite reported on contemporaneous photometric and spectroscopic observations of ii peg , finding that the h@xmath10 emission and photometric intensity are strongly anticorrelated , suggesting that regions of high chromospheric activity are physically associated with the spots . this conclusion was corroborated by a rotational modulation of the intensity of the he i d@xmath11 line . based on an estimated radius of @xmath12 and @xmath13 , they estimated the inclination between the rotation axis and the line of sight to be @xmath14@xmath15 . @xcite confirmed via long - term monitoring of @xmath1 as well as the @xmath16 and @xmath17 colors that ii peg is redder when it is dimmer , as would be expected if the dimming is caused by cool spots . a number of studies have attempted to determine the spot temperatures . @xcite modeled light and color curves obtained in 1977 with a single circular spot , finding a spot temperature of @xmath18 . @xcite obtained @xmath19 from observations in the fall of 1979 ; @xcite obtained @xmath20 for fall 1980 ; @xcite obtained @xmath21 for fall 1981 ; @xcite obtained @xmath22 for fall 1986 ; and @xcite obtained @xmath23 for 19861987 . by modeling the strengths of tio absorption bands , @xcite found evidence for multiple spot temperatures , finding @xmath24 to vary between @xmath25 and @xmath26 as the star was observed through slightly less than one rotational period . more recently , from a spot model applied to contemporaneous photometry and spectroscopy , @xcite obtained @xmath27 . @xcite used a simple analytic two - spot model to fit photoelectric light curves of four chromospherically active binaries : @xmath28 and , @xmath29 gem , v711 tau , and ii peg . the ii peg data were acquired from 19731992 , and subdivided into 37 individual light curves . they plotted `` migration curves '' for twelve long - lived spots they identified in the data from the times of minimum light obtained via the spot - model curve fits . a migration curve shows the variation in the phase of minimum light with time , where the phase was computed using the orbital ephemeris and represents the fractional part of the number of rotation periods since an arbitrary starting time . assuming tidal locking and no differential rotation , a given spot would always cross the central meridian of the stellar disc as seen from earth once per revolution period and thus always at the same phase . however , if the star exhibits latitude - dependent differential rotation , we would expect to see a given spot progressively advanced or retarded in phase relative to the orbital ephemeris . a plot of the phase of minimum light versus time for a given spot should then be a straight line with slope determined by the difference in the rotation period of the latitude of the spot and the revolution period . this was precisely what henry et al . observed in the data , for ii peg and the other stars . the plots for different spots had different slopes , demonstrating latitude - dependent differential rotation . the degree of differential rotation was specified in terms of the differential rotation coefficient , @xmath2 , defined for the sun by fitting the rotation period as a function of latitude with the relation @xmath30 where @xmath31 is the rotation period for latitude @xmath32 and @xmath33 is the rotation period at the equator . for the sun , @xmath34 . if the differential rotation of other stars has the same functional form as for the sun , and if the rotation periods for spots sampling a range of latitudes are determined for a star , then the coefficient @xmath2 is given by @xmath35 where @xmath36 , @xmath37 , and @xmath38 are the maximum , minimum , and average observed periods , and @xmath39 is a distribution function which relates the the total range in rotational period sampled to the number of spots for which the period has been determined @xcite . the value of @xmath39 ranges from 0.5 for two spots to over 0.9 when the number of spots exceeds six . @xcite used eight of the twelve spots they observed on ii peg ( four spots were observed over intervals too short to allow their periods to be obtained reliably ) to determine @xmath2 using equation ( [ eq : kf ] ) , with the result @xmath40 . @xcite performed an analysis similar to the present study , inverting light curves acquired between 1974 and 1998 to produce images of the stellar surface . in contrast to the smoothing function used here ( see [ sec : li_algorithm ] and in particular equation ( [ eq : smoothingfunction ] ) ) , they used maximum entropy and tikhonov regularization . they concluded that the distribution of spots on ii peg consists of a component distributed uniformly in longitude which does not rotationally modulate the light curve ( but does produce a secular variation in the mean intensity ) , plus an unevenly distributed component responsible for the rotational modulation . their analysis indicated that the uniformly distributed component varied in total area with a period of @xmath41 . they determined the unevenly distributed component to be concentrated around three active longitudes , one of these having an essentially permanent presence but a cycle in spot area with period @xmath42 . they found the activity of the other two active longitudes to switch back and forth , with one active while the other is inactive , with a period of @xmath43 . however , there is an interval of @xmath44 before the switch in which both longitudes are active . there is thus a period of @xmath45 during which only one of the two longitudes engaged in the `` flip - flop '' behavior is active , which agrees with the switching period deduced by @xcite from the times of light minima . from a periodogram analysis , @xcite found periodicities in the mean magnitudes for the spot - model fits of their 37 light curves of @xmath46 and @xmath47 . they interpreted the 4.4-yr period as reflecting the average lifetime of the spots and the 11-yr period as representing a different timescale . rodon et al . interpreted the 4.7-yr periodicity arising in their analysis as corresponding to the 4.4-yr period obtained by henry et al . , while they interpreted the 9.5-yr period they saw in the total area of the spot component which is unevenly distributed in longitude as corresponding to the 11-yr period found by henry et al . henry et al . noted that their two - spot model , which assumed circular spots varying only in radius over time , was not fully adequate to explain the variations with time of the ii peg light curve . in particular , when the amplitude of the rotational modulation due to spots they designated g and h was diminishing , the mean brightness of the star stayed roughly constant . similar behavior was seen for another pair of spots , which they designated j and k. if the decrease in amplitude were due simply to a decrease in the spot radii , the mean brightness of the star should have increased ( assuming no change in the brightness of the photosphere outside the spots ) . on the other hand , if instead the spots were being drawn out in longitude by differential rotation while maintaining nearly constant area , then the amplitude would decrease while the mean brightness stayed constant , as observed . the present study is suited to look for evidence for such drawing out ( or compression ) in longitude of active regions by differential rotation , as we produce images of the active regions and observe changes in them over periods of several months . by simultaneously inverting contemporaneous @xmath0 and @xmath1 light curves , we exploit differences in the limb darkening as seen through the two filters to achieve significantly better latitude resolution than is possible when using light curves obtained through only a single filter @xcite , thereby allowing us to directly detect differential rotation in our images . it should be noted , however , that we do not claim to obtain accurate spot latitudes from our two - filter inversions ; nonetheless , simulations like those detailed in harmon & crews show that _ relative _ spot latitudes can be obtained with good reliability , i.e. , when two spots are present , the one at the lower latitude is rendered as such . in this regard our approach differs from that of rodon et al . , who used their @xmath1-filter surface imagery just to derive quantities that they claim are independent of the regularization criterion , such as the distributions of the spots versus longitude , the changes in the distribution over time , and the variations of the total area covered by spots . in [ sec : li_algorithm ] , we discuss the method used to invert the light curves so as to produce these images . in [ sec : dataanalysis ] , we discuss the division of the over nineteen years worth of @xmath0 and @xmath1 light curves into separate data sets and the procedure used to process them for inversion . in [ sec : results ] , we discuss in detail the results for the six intervals over which we saw good evidence in our images for alteration of the active region configuration by latitude - dependent differential rotation . finally , in [ sec : discussion ] , we show that our results are consistent with the result for the value of the differential rotation coefficient @xmath2 for ii peg inferred by @xcite based on longer - term monitoring of the times of minimum light due to individual spots using their two - spot model rather than surface imaging . light - curve inversion ( li ) is a photometric imaging technique which produces a map of a star s surface based on the brightness variations produced as dark ( or possibly bright ) starspots are carried into and out of view of earth by the star s rotation . it makes no _ a priori _ assumptions regarding the number of spots on the surface or their shapes . the details regarding the implementation of the algorithm are presented in @xcite , along with the results of extensive tests in which artificial stellar surfaces were used to create light curves , which were then inverted . in harmon & crews , the technique is called `` matrix light - curve inversion , '' because it evolved from the original formulation described in @xcite and called by that name . however , because the formulation described in @xcite and as modified in the present work no longer uses matrices , we shortened the name . here we outline the method , and refer the reader to harmon & crews for more details . the stellar surface is subdivided into @xmath48 bands in latitude of equal angular widths @xmath49 . each latitude band is further subdivided into patches which are all `` spherical rectangles '' of equal widths in longitude @xmath50 , where @xmath51 is the number of patches in the @xmath52 latitude band . the @xmath51 are chosen to be proportional to the cosine of the latitude ( to within the constraint that the @xmath51 must be integers ) so that the areas of all the patches are nearly equal . the visible pole is defined to be the north pole , with latitude @xmath53 , while the hidden south pole has latitude @xmath54 . the @xmath55 patch in the @xmath52 latitude band is designated patch @xmath56 . the first patch in each latitude band , patch @xmath57 , straddles the meridian with longitude @xmath58 , defined to be the one which intersects the equator on the approaching limb of the star at an arbitrarily chosen reference time @xmath59 . longitude increases in the direction of the star s rotation , so the sub - observer longitude at @xmath60 is thus @xmath61 . in the absence of interstellar absorption , at the time @xmath62 of observation number @xmath2 through filter @xmath63 , the intensity @xmath64 observed at earth is ( in the limit that the number of patches is large ) @xmath65 where @xmath66 is the number of filters , @xmath67 is the number of observations through filter @xmath63 , @xmath68 is the specific intensity ( w m@xmath69 sr@xmath70 ) along the outward normal of patch @xmath56 integrated over the passband of filter @xmath63 , @xmath71 is the solid angle of patch @xmath56 as seen from earth at time @xmath62 ( we set @xmath72 if the patch is on the far side of the star ) , @xmath73 is the factor by which the specific intensity emitted in the direction of earth is attenuated by limb darkening compared to that emitted along the outward normal ( so that @xmath74 is the specific intensity emitted along the line of sight to earth ) , and @xmath75 is the index of the southernmost latitude band which is visible from earth . the goal of li is to find a set of computed patch intensities @xmath76 that mimics the actual variations of surface brightness across the stellar surface as closely as possible . ( we use a caret over a quantity to indicate that it represents a value as computed by the li algorithm . ) since we generally do not know the actual radius and distance of the star very precisely , we content ourselves with finding only the _ relative _ brightnesses of the patches to one another . to this end we simply define the radius of the star to be 1 and use the area of a patch projected onto the plane of the sky as a proxy for the solid angle it subtends at earth . we use the limb - darkening coefficients published by @xcite to determine the values of the @xmath73 in equation ( [ eq : ink ] ) . the benefit of observing through multiple filters is that we can take advantage of the differences in the degree of limb darkening as seen through different filters in order to significantly increase the latitude resolution of the inversions , as explained in @xcite . in order to take advantage of this information , we must simultaneously invert all of the filter light curves . this in turn requires that we couple together the @xmath76 for different values of the filter index , @xmath63 . to do this , we designate the filter for which the light curve has the lowest noise as the `` primary filter '' and assign it filter index @xmath77 . for simplicity we assume that the actual stellar surface can be described via a two - component model in which all the spots have the same temperature @xmath24 and thus emit the same specific intensity along the outward normal as seen through filter @xmath63 , which we designate at @xmath78 ; similarly , we assume that all points on the surface outside spots are part of a photosphere of uniform temperature @xmath79 and emitting specific intensity @xmath80 along the outward normal . however , it should be noted that the reconstructed surface created by the inversion of the data does _ not _ have the property that the @xmath76 can have only one of two values ; they are continuous variables . we then define the intensities of the patches as viewed through filter @xmath81 via the linear scaling @xmath82 . \label{eq : scaling}\ ] ] here @xmath83 is the average value of the @xmath84 , @xmath85 is the estimated value of @xmath86 on the actual stellar surface , and @xmath87 is the estimated value of @xmath88 for the actual stellar surface . we estimate these ratios by calculating the planck function at the central wavelength of the filter in question at the assumed spot and photosphere temperatures . it would be more accurate to integrate the product of the planck function and the filter sensitivity functions over their passbands , but typically we do not know the spot and photosphere temperatures with sufficient precision to justify the extra effort . the scaling given in equation ( [ eq : scaling ] ) from @xmath89 to @xmath76 has the property that when @xmath90 according to our estimate , then @xmath91 as well . we are using @xmath83 as a proxy for @xmath92 , which should be a reasonable approximation as long as the stellar photosphere comprises most of the surface . since the patch brightnesses through all the filters in the model are entirely determined in terms of their brightnesses @xmath89 as seen through filter 1 , the problem reduces to finding these values , so for notational simplicity we define @xmath93 . as is well known , the problem of determining the @xmath94 is extremely sensitive to the presence of even small amounts of noise in the data . this can be seen by considering the effect on the light curve produced by a myriad of small spots distributed all over the surface . as the star rotated , at any given time nearly equal numbers of spots would be rising over the approaching limb and setting over the receding limb . thus , the total contribution to the star s brightness from the spots would be nearly but not exactly constant , so that the effect of the spots would be to impart a small - amplitude , high - frequency ripple on the light curve , very similar to the effect of random noise in the observations . conversely , if we attempt to fit noisy data , then unless precautions are taken , the resulting model surface will be covered with spurious small spots introduced in order to `` explain '' the presence of the noise in the signal . to avoid this dilemma , rather than simply finding the set of @xmath94 that yields the best fit to the light curve data , we determine the @xmath94 by finding the set of them which minimizes the _ objective function _ @xcite : @xmath95 here @xmath96 represents the set of the @xmath94 , while @xmath97 represents the set of observed intensities @xmath64 , i.e. , the data light curve . the function @xmath98 expresses the goodness - of - fit of the calculated light curve @xmath99 ( with components @xmath100 ) obtained from @xmath96 to the data light curve @xmath97 , such that smaller values of @xmath98 imply a better fit to the data . the _ smoothing function _ @xmath101 is defined such that it takes on smaller values for surfaces that are `` smoother '' in some appropriately defined sense . finally , @xmath28 is an adjustable lagrange multiplier called the _ tradeoff parameter _ , and @xmath0 is an adjustable parameter called the _ bias parameter _ , which is discussed below . note that as @xmath102 , the first term on the right dominates , so that minimizing @xmath103 is equivalent to minimizing @xmath104 , and we obtain the solution which best fits the light curve data but suffers from the spurious noise artifacts discussed above . on the other hand , as @xmath105 , the second term dominates , so that minimizing @xmath103 produces a very `` smooth '' surface lacking in noise artifacts , but also producing a very poor fit to the data . thus , by varying @xmath28 , we adjust the tradeoff between goodness - of - fit and smoothness of the model surface . if we choose @xmath28 such that @xmath98 is equal to a corresponding estimate of the amount of noise in the data , then in a rough sense we can say that by minimizing the objective function , we find the smoothest solution @xmath96 for which the corresponding light curve @xmath99 fits the data light curve @xmath97 to a degree which is as good as but not better than is justified by the noise in the data . in this way , we obtain a model surface which fits the data well , but not so well that it is dominated by noise artifacts . for the goodness - of - fit function in this study , we use @xmath106 here we assume that light curves have been obtained through @xmath66 different photometric filters ( @xmath107 in the present work since we use @xmath0- and @xmath1-filter data ) , and that the magnitudes have been converted to intensities . since our goal is only to find the relative values of the @xmath94 , it suffices to use relative rather than absolute intensities for the light curve data in the calculation of @xmath98 . the number of observations in the light curve obtained through filter @xmath63 is @xmath67 , while @xmath108 is the total number of data points in all the light curves . the estimated noise variance in the light curve data for filter @xmath63 expressed in magnitudes is @xmath109 . in @xcite it is shown that to a good approximation , the true noise variance @xmath110 is given by @xmath111 where @xmath112 is the true noise - free value of the intensity ( which is of course unknown unless one is doing a simulation ) . if we define @xmath113 to be the true error in the measurement @xmath64 , and @xmath114 to be the deviation between the calculated and true intensities , then with this notation @xmath115 while @xmath116 if @xmath99 is a good match to the data @xmath97 , then the @xmath117 and @xmath118 are small quantities , and we can expand the right side of equation ( [ eq : gterm ] ) as @xmath119 then @xmath120 \nonumber \\ & = & \frac{1}{p}\sum_{n=1}^{p_n}\left[p_n\frac{\tilde{\sigma}_n^2}{\sigma_n^2 } + \frac{(2.5\log_{10}e)^2}{\sigma_n^2 } \sum_{k=1}^{p_n } \frac{\delta_{nk}^2 - 2\epsilon_{nk}\delta_{nk}}{\tilde{i}_{nk}^2 } + \ldots \right]\end{aligned}\ ] ] if the reconstructed intensities @xmath100 perfectly matched the true intensities @xmath112 , and in addition the estimated noise variances @xmath121 were equal to the true noise variance @xmath122 , then we would have all @xmath123 , and so we would have @xmath124 to lowest order in the @xmath117 . for this reason , given the estimates @xmath121 and for a given value of the bias parameter @xmath0 , we vary @xmath28 until the stopping criterion @xmath124 is attained to a predetermined precision . the smoothing function used in this study is @xmath125 where @xmath126 if @xmath127 , while @xmath128 if @xmath129 . thus , patches brighter or darker than average incur a penalty in that they increase the value of @xmath101 ( and thus the function @xmath130 to be minimized ) by an increasing amount as the deviation from the average increases . note that @xmath101 satisfies the criterion that it takes on its minimum possible value of zero for a featureless surface which is perfectly `` smooth '' in that all the patch brightnesses @xmath94 are equal , and that surfaces showing greater deviations about the average are judged as `` rougher . '' for @xmath131 , the penalty for a patch being brighter than average by a given amount is @xmath0 times larger than for a patch darker than average to the same degree . thus , @xmath0 biases the solution toward having most patches just slightly brighter than average to represent the stellar photosphere , which is assumed to be almost uniformly bright like the sun s , while a smaller number are much darker than average to represent the dark starspots . this is the reason for the name `` bias parameter . '' the simulations described in @xcite show that as @xmath0 is increased , the ratio @xmath132 decreases , so that the darkest patch becomes darker relative to the average patch brightness . we use this ratio as a proxy for the assumed ratio of the spot and photosphere brightnesses as seen through filter 1 . for a given value of the tradeoff parameter @xmath28 , the bias parameter @xmath0 can be varied until @xmath133 , the estimated spot - to - photosphere brightness ratio seen through filter 1 , to a predetermined precision . the scaling given by equation ( [ eq : scaling ] ) ensures that @xmath134 , the estimated spot - to - photosphere brightness ratio for filter @xmath63 . the procedure for inverting a series of light curves obtained through a set of filter passbands is then as follows . the input parameters are the estimated noise variances @xmath109 of the light curves , the estimated spot and photosphere temperatures @xmath24 and @xmath79 , and the inclination angle @xmath10 of the rotation axis to the line of sight . the @xmath121 in the definition of @xmath98 , @xmath24 , and @xmath79 are used to obtain the values of @xmath87 and @xmath85 ( including @xmath135 ) in equation ( [ eq : scaling ] ) , and the inclination @xmath10 is used in finding @xmath71 and @xmath73 in equation ( [ eq : ink ] ) . as described in @xcite , two copies of a root - finding subroutine are used in concert so as to find the values of @xmath28 and @xmath0 such that @xmath124 _ and _ @xmath134 to the desired precision . the result is a set of solutions , one for each combination of the input parameters . how we select one of these to represent the `` best '' solution is described in [ sec : dataanalysis ] . the raw data consisted of johnson @xmath0 and @xmath1 differential magnitudes paired with corresponding heliocentric modified julian dates , acquired from heliocentric mjd 47115.8086 to 54136.6221 ( 1987 november 16 to 2007 february 5 ) with the tennessee state university t3 0.4-m automated photometric telescope at fairborn observatory in arizona @xcite . the complete @xmath0 and @xmath1 data sets are plotted as the upper and lower panels of figure [ fig : completedataset ] . the first task was to convert the modified julian dates to rotational phases . the rotational phase @xmath136 is defined as @xmath137 where @xmath138 is the time of the observation , @xmath59 is an arbitrary reference time used for all the observations , @xmath139 is the rotational period of the star , and @xmath140 is the greatest integer which is less than or equal to @xmath141 . thus , @xmath142 represents the fraction of a rotation through which the star has turned relative to the orientation it had at time @xmath59 . on the assumption that the star exhibits differential rotation , there is no such thing as _ the _ rotational period , so @xmath139 here represents a suitable average of the rotation period over all latitudes . on the assumption of tidal locking , it is reasonable to use the orbital period for this average . in the present study , @xmath143 d was used based on the orbital period obtained by @xcite , and @xmath59 was chosen as jd 2443033.47 , based on the orbital ephemeris of @xcite , in which @xmath59 represents the time of superior conjunction , when the primary is farthest from the observer . the next task was to subdivide the data into individual data sets comprised of pairs of @xmath0 and @xmath1 light curves suitable for inversion . each light curve needed to contain enough data points so as to provided good phase coverage . ideally this would be achieved using data acquired during a single rotation of the star , since this would minimize the chance that the spot configuration had evolved significantly during the time interval spanned by the data . however , in practice this was not feasible , because the star s rotation period is too long to allow for continuous monitoring during a single rotation , and because the telescope was not dedicated solely to this study . the desire for good phase coverage is thus in opposition to the desire to minimize the number of stellar rotations covered in any one data set , so some compromises were necessary . if data for different revolutions showed a systematic shift in the magnitudes , it was clear that the stellar surface features had evolved by an unacceptable amount during the interval in question ; otherwise , data from additional rotations could be included if needed so as to improve the phase coverage . several groups of observations were discarded because they were temporally isolated by many rotation periods from the observations nearest them in time and contained an insufficient number of observations to produce good phase coverage . in the end , 68 pairs of @xmath0 and @xmath1 light curves were created that were subsequently inverted to produce the results reported in this study . the mean number of observations per @xmath0 light curve was 22.3 , the median was 21.5 and the standard deviation was 6.3 , while for the @xmath1 light curves the mean was 22.1 , the median was 21 and the standard deviation was 5.9 . the final task before inverting the light curves was to convert the differential magnitudes to relative intensities . the formula used for both the @xmath0 and the @xmath1 passbands was simply @xmath144 where @xmath145 was the differential magnitude for the observation in question , and the reference magnitude @xmath146 was the smallest value of @xmath145 in the entire data set for the given filter . an intensity of @xmath147 was thus assigned to this observation . for the @xmath0 filter , @xmath148 , while for the @xmath1 filter , @xmath149 . no attempt was made to calibrate the @xmath1 intensities relative to the @xmath0 intensities in an absolute sense , because the li algorithm neither requires nor would make use of this information , as mentioned in [ sec : li_algorithm ] . @xcite obtained @xmath9 for the inclination of the rotation axis . we performed inversions assuming their nominal value of @xmath14 . we also performed inversions for @xmath150 as well , considering this to be prudent given the considerable uncertainty in the inclination . as will be seen from the discussion of the individual data sets below , the results for both assumed inclinations were generally consistent , increasing our confidence in their validity . while one might argue that we also should have performed inversions for assumed inclinations greater than @xmath151 , say @xmath152 or @xmath153 , simulations like those reported in @xcite show that the method works poorly in such circumstances . we thus chose not to do so . for the spot and photosphere temperatures , we used @xmath154 k since this is compatible with the estimates by other authors mentioned in [ sec : introduction ] , and @xmath155 k based on the work of @xcite . the estimates @xmath85 of @xmath86 and @xmath87 of @xmath88 appearing in the scaling given by equation ( [ eq : scaling ] ) were obtained by evaluating the planck function describing blackbody radiation at the filter effective wavelengths @xmath156 nm and @xmath157 nm . the photosphere temperature @xmath79 and surface acceleration due to gravity @xmath6 are the input parameters used by @xcite to calculate limb - darkening coefficients based on the atlas stellar atmosphere models of @xcite . we used @xmath158 , which is appropriate for a k2 subgiant @xcite , and is the value in van hamme s tables which is closest to the result @xmath5 of @xcite . van hamme gives coefficients in steps of 250 k for temperatures in the range @xmath159 , so there is no entry for our value of @xmath155 k. we simply substituted the values corresponding to @xmath160 k , the nearest listed temperature to ours , since we do not know @xmath79 or @xmath6 accurately enough to justify interpolating . from table 2 in van hamme s paper , we find for the @xmath0 filter that @xmath161 with @xmath162 and @xmath163 , where @xmath164 is the cosine of the angle between the outward normal to the surface and the observer s line of sight . for the @xmath1 filter @xmath165 with @xmath166 and @xmath167 . given the latitude and longitude of patch @xmath56 and the angle of inclination @xmath10 , it is straightforward to compute @xmath164 for the center of the patch at any given observation time @xmath62 and thus to obtain @xmath73 in equation ( [ eq : ink ] ) . the next step was to obtain estimates of the noise variances @xmath168 and @xmath169 for each of the 68 light curves for each filter . it is well known that in practice twomey s criterion of choosing the tradeoff parameter @xmath28 so that the variance between the data and the reconstruction is equal to the noise variance leads to over - smoothing @xcite . in our case this would lead to a loss in resolution of the reconstructed surfaces . however , the technique we use to determine @xmath28 avoids this problem , because we obtain an `` effective noise level '' rather than using an estimate of the noise variance based on the scatter in the comparison star magnitudes . @xcite describes simulations in which an artificial star is used to generate light curves to which random noise of a known variance @xmath110 is added . this allowed the effects of using an underestimate @xmath170 of the noise variance in the goodness - of - fit function @xmath98 to be determined . it was found that for a given light curve , when the ratio @xmath171 falls below a certain value ( typically between 0.90 and 0.98 ) , the solution `` falls apart '' in that it starts to show very obvious noise artifacts . the transition to this behavior is quite sharp in that it takes place over a narrow range of values of this ratio . this gives a practical means to determine how low @xmath109 can be pushed while still yielding acceptable solutions . thus , we can avoid the over - smoothing associated with the twomey criterion by performing inversions for a range of values of @xmath109 and then choosing the lowest value which leads to a solution free of obvious noise artifacts . this value is what we call the `` effective noise level '' for the light curve . because the transition is not perfectly sharp , the precise choice of @xmath109 is to some extent a judgment call , so we were conservative in our choices . the first round of inversions was thus a series of single - filter inversions of all 68 @xmath0 and 68 @xmath1 light curves so as to assign an effective noise to each one . the next round was to take each pair of contemporaneous @xmath0 and @xmath1 light curves and invert the pair in combination so as to produce the finished surface map for that pair . a complication is that when using li to simultaneously invert light curves obtained through several filters , the effective noise levels as determined from single - filter inversions sometimes ( but not always ) lead to under - smoothed surface images corrupted by noise artifacts . thus , we used the effective noise levels obtained from the initial round of single - filter inversions as starting points , and ran a series of inversions for each pair of light curves using nearby values of @xmath172 and @xmath173 . the resulting images were inspected to determine the lowest values of @xmath172 and @xmath173 that did not result in obvious noise artifacts ; again we were conservative in our judgments . in general , when simultaneously inverting multiple light curves , the deviation for a given filter between the data and reconstructed light curves differs somewhat from the effective noise . this arises because the convergence criterion for the li algorithm is based on the _ overall _ deviation between the data and reconstructed light curves through the various filters taken together as a whole , rather than on the deviations for the individual filters . this is necessary because the scaling given by equation ( [ eq : scaling ] ) , which defines the patch intensities assigned to the secondary filter(s ) in terms of their values as seen through the primary filter , makes it impossible to independently tweak how well the individual reconstructed light curves match the data light curves . each of the 68 data sets consisting of paired @xmath0 and @xmath1 light curves was inverted according to the procedure outlined in [ sec : li_algorithm ] . upon careful examination of the resulting images , the data sets were divided into 21 groups . each group covers a span of time during which we can see evolution of a particular set of spots , except for three cases in which a group consists of a single data set because it is temporally isolated from the data sets immediately before and after it by long gaps in the data . in six of these groups we saw good evidence for differential rotation , and as looking for such evidence was our primary goal in this work , we present in detail only the results for these six groups here , arranged in chronological order . the time spans corresponding to these six groups are indicated via horizontal bars in the plots of figure [ fig : completedataset ] . the remainder of our images are presented in figures 1733 in the online version of the journal . table [ tab : datasets ] summarizes the properties of each of the data sets and groups . the data sets are numbered in chronological order in the first column , with the group to which each set was assigned indicated in the second column . boldface entries in the first two columns denote the groups for which we present evidence of differential rotation . the starting and ending heliocentric modified julian dates and corresponding utc calendar dates for the @xmath0 and @xmath1 light curves of each data set are given in the fourth through seventh columns ; a blank entry for a @xmath1 light curve indicates that the value is the same as for the @xmath0 light curve from the same set . the number of data points in each data set is shown in the @xmath174 column . the number of rotation periods covered by each data set is listed in the `` # per . '' the last two columns list the `` effective noise '' in magnitudes for each inversion , as described in section [ sec : dataanalysis ] , for the assumed rotation axis inclinations of @xmath150 and @xmath14 . the values in the table are @xmath175 times larger than the actual values , e.g. , an entry of `` 154 '' means that the effective noise was 0.0154 mag . rrrrrrrrrrr 1 & 1 & @xmath0 & 47115.8086 & 47141.7351 & 1987 nov 16 & 1987 dec 12 & 15 & 3.86 & 154 & 154 + & & @xmath1 & & & & & 17 & & 194 & 194 + 2 & 1 & @xmath0 & 47171.6466 & 47198.5822 & 1988 jan 11 & 1988 feb 07 & 10 & 4.01 & 138 & 132 + & & @xmath1 & & 47199.5824 & & 1988 feb 08 & 12 & 4.15 & 226 & 216 + * 3 * & * 2 * & @xmath0 & 47417.7681 & 47433.8609 & 1988 sep 13 & 1988 sep 29 & 20 & 2.39 & 268 & 258 + & & @xmath1 & 47415.8437 & & 1988 sep 11 & & 23 & 2.68 & 326 & 326 + * 4 * & * 2 * & @xmath0 & 47434.7230 & 47469.8305 & 1988 sep 30 & 1988 nov 04 & 32 & 5.22 & 164 & 158 + & & @xmath1 & & & & & 30 & & 208 & 200 + * 5 * & * 2 * & @xmath0 & 47470.6837 & 47525.6801 & 1988 nov 05 & 1988 dec 30 & 31 & 8.18 & 174 & 174 + & & @xmath1 & & & & & 28 & & 294 & 286 + * 6 * & * 2 * & @xmath0 & 47526.5929 & 47556.5813 & 1988 dec 31 & 1989 jan 30 & 12 & 4.46 & 144 & 144 + & & @xmath1 & & 47549.5824 & & 1989 jan 23 & 13 & 3.42 & 146 & 146 + * 7 * & * 3 * & @xmath0 & 47779.7480 & 47818.8716 & 1989 sep 10 & 1989 oct 19 & 35 & 5.82 & 304 & 296 + & & @xmath1 & & & & & 33 & & 298 & 298 + * 8 * & * 3 * & @xmath0 & 47824.6882 & 47850.7976 & 1989 oct 25 & 1989 nov 20 & 30 & 3.88 & 392 & 390 + & & @xmath1 & & & & & 31 & & 410 & 402 + * 9 * & * 3 * & @xmath0 & 47853.6658 & 47921.5825 & 1989 nov 23 & 1990 jan 30 & 38 & 10.10 & 238 & 214 + & & @xmath1 & & 47926.5752 & & 1990 feb 04 & 33 & 10.84 & 238 & 206 + 10 & 4 & @xmath0 & 48183.8086 & 48236.6947 & 1990 oct 19 & 1990 dec 11 & 25 & 7.86 & 278 & 278 + & & @xmath1 & & 48235.6694 & & 1990 dec 10 & 25 & 7.71 & 182 & 184 + 11 & 5 & @xmath0 & 48397.9614 & 48437.9355 & 1991 may 21 & 1991 jun 30 & 21 & 5.94 & 124 & 118 + & & @xmath1 & 48394.9734 & & 1991 may 18 & & 27 & 6.39 & 94 & 94 + 12 & 6 & @xmath0 & 48872.9343 & 48898.8417 & 1992 sep 07 & 1992 oct 03 & 17 & 3.85 & 82 & 80 + & & @xmath1 & & & & & 17 & & 56 & 64 + 13 & 6 & @xmath0 & 48905.7999 & 48945.7709 & 1992 oct 10 & 1992 nov 19 & 28 & 5.94 & 200 & 196 + & & @xmath1 & & & & & 28 & & 196 & 198 + 14 & 6 & @xmath0 & 48951.7163 & 49022.5799 & 1992 nov 25 & 1993 feb 04 & 20 & 10.54 & 112 & 106 + & & @xmath1 & & & & & 20 & & 106 & 106 + 15 & 7 & @xmath0 & 49135.9488 & 49166.9202 & 1993 may 28 & 1993 jun 28 & 23 & 4.61 & 124 & 128 + & & @xmath1 & & & & & 23 & & 122 & 118 + 16 & 7 & @xmath0 & 49235.9074 & 49253.7684 & 1993 sep 05 & 1993 sep 23 & 15 & 2.66 & 46 & 40 + & & @xmath1 & & & & & 15 & & 44 & 46 + 17 & 7 & @xmath0 & 49275.8128 & 49340.6260 & 1993 oct 15 & 1993 dec 19 & 25 & 9.64 & 144 & 140 + & & @xmath1 & 49282.7438 & & 1993 oct 22 & & 26 & 8.61 & 132 & 134 + 18 & 7 & @xmath0 & 49347.6242 & 49384.5901 & 1993 dec 26 & 1994 feb 01 & 17 & 5.50 & 94 & 106 + & & @xmath1 & 49345.6833 & & 1993 dec 24 & & 18 & 5.79 & 104 & 116 + 19 & 8 & @xmath0 & 49638.7416 & 49671.8291 & 1994 oct 13 & 1994 nov 15 & 22 & 4.92 & 106 & 110 + & & @xmath1 & & & & & 22 & & 104 & 108 + 20 & 8 & @xmath0 & 49674.7171 & 49706.6550 & 1994 nov 18 & 1994 dec 20 & 17 & 4.75 & 56 & 56 + & & @xmath1 & & & & & 18 & & 76 & 76 + 21 & 8 & @xmath0 & 49724.6199 & 49757.5937 & 1995 jan 07 & 1995 feb 09 & 14 & 4.90 & 70 & 76 + & & @xmath1 & & & & & 14 & & 86 & 96 + 22 & 9 & @xmath0 & 49873.9567 & 49909.8632 & 1995 jun 05 & 1995 jul 11 & 26 & 5.34 & 120 & 116 + & & @xmath1 & & & & & 25 & & 110 & 114 + 23 & 10 & @xmath0 & 49982.8845 & 50001.9313 & 1995 sep 22 & 1995 oct 11 & 27 & 2.83 & 88 & 96 + & & @xmath1 & & & & & 28 & & 54 & 58 + 24 & 10 & @xmath0 & 50002.7800 & 50033.7425 & 1995 oct 12 & 1995 nov 12 & 32 & 4.60 & 88 & 86 + & & @xmath1 & & & & & 32 & & 82 & 82 + 25 & 10 & @xmath0 & 50037.8137 & 50062.6899 & 1995 nov 16 & 1995 dec 11 & 27 & 3.70 & 146 & 140 + & & @xmath1 & & & & & 26 & & 120 & 116 + 26 & 10 & @xmath0 & 50066.7615 & 50123.5946 & 1995 dec 15 & 1996 feb 10 & 30 & 8.45 & 162 & 164 + & & @xmath1 & & 50129.5853 & & 1996 feb 16 & 30 & 9.34 & 132 & 132 + * 27 * & * 11 * & @xmath0 & 50391.8348 & 50435.6313 & 1996 nov 04 & 1996 dec 18 & 26 & 6.51 & 120 & 124 + & & @xmath1 & & & & & 25 & & 114 & 104 + * 28 * & * 11 * & @xmath0 & 50436.6379 & 50490.5894 & 1996 dec 19 & 1997 feb 11 & 26 & 8.02 & 208 & 204 + & & @xmath1 & & 50494.5846 & & 1997 feb 15 & 25 & 8.62 & 152 & 140 + * 29 * & * 11 * & @xmath0 & 50590.9563 & 50642.8849 & 1997 may 22 & 1997 jul 13 & 25 & 7.72 & 116 & 104 + & & @xmath1 & & & & & 28 & & 108 & 100 + * 30 * & * 11 * & @xmath0 & 50714.8051 & 50755.7973 & 1997 sep 23 & 1997 nov 03 & 35 & 6.10 & 120 & 120 + & & @xmath1 & & & & & 35 & & 126 & 108 + * 31 * & * 11 * & @xmath0 & 50756.7863 & 50795.6926 & 1997 nov 04 & 1997 dec 13 & 23 & 5.79 & 72 & 80 + & & @xmath1 & & & & & 21 & & 70 & 78 + * 32 * & * 11 * & @xmath0 & 50797.6863 & 50837.6257 & 1997 dec 15 & 1998 jan 24 & 20 & 5.94 & 124 & 114 + & & @xmath1 & & & & & 19 & & 118 & 104 + * 33 * & * 11 * & @xmath0 & 50838.6203 & 50856.5946 & 1998 jan 25 & 1998 feb 12 & 13 & 2.67 & 172 & 180 + & & @xmath1 & & & & & 13 & & 120 & 128 + * 34 * & * 12 * & @xmath0 & 51085.9791 & 51115.7864 & 1998 sep 29 & 1998 oct 29 & 25 & 4.43 & 86 & 86 + & & @xmath1 & 51086.7915 & & 1998 sep 30 & & 25 & 4.31 & 116 & 96 + * 35 * & * 12 * & @xmath0 & 51116.7725 & 51144.7050 & 1998 oct 30 & 1998 nov 27 & 28 & 4.15 & 120 & 122 + & & @xmath1 & & & & & 27 & & 112 & 108 + * 36 * & * 12 * & @xmath0 & 51148.7022 & 51182.6674 & 1998 dec 01 & 1999 jan 04 & 24 & 5.05 & 76 & 84 + & & @xmath1 & & & & & 24 & & 104 & 114 + * 37 * & * 12 * & @xmath0 & 51183.6640 & 51224.5942 & 1999 jan 05 & 1999 feb 15 & 24 & 6.09 & 172 & 174 + & & @xmath1 & & & & & 23 & & 162 & 158 + * 38 * & * 13 * & @xmath0 & 51429.8915 & 51475.8034 & 1999 sep 08 & 1999 oct 24 & 29 & 6.83 & 146 & 146 + & & @xmath1 & & & & & 28 & & 108 & 108 + * 39 * & * 13 * & @xmath0 & 51477.8148 & 51505.7306 & 1999 oct 26 & 1999 nov 23 & 27 & 4.15 & 202 & 208 + & & @xmath1 & & & & & 24 & & 118 & 130 + * 40 * & * 13 * & @xmath0 & 51506.7278 & 51535.6724 & 1999 nov 24 & 1999 dec 23 & 23 & 4.30 & 94 & 100 + & & @xmath1 & & & & & 23 & & 76 & 72 + * 41 * & * 13 * & @xmath0 & 51537.6574 & 51586.5994 & 1999 dec 25 & 2000 feb 12 & 29 & 7.28 & 152 & 138 + & & @xmath1 & & & & & 30 & & 118 & 126 + * 42 * & * 14 * & @xmath0 & 51805.8758 & 51833.8204 & 2000 sep 18 & 2000 oct 16 & 17 & 4.16 & 124 & 126 + & & @xmath1 & & & & & 18 & & 154 & 156 + * 43 * & * 14 * & @xmath0 & 51838.8332 & 51879.8006 & 2000 oct 21 & 2000 dec 01 & 18 & 6.09 & 80 & 74 + & & @xmath1 & & & & & 18 & & 94 & 84 + * 44 * & * 14 * & @xmath0 & 51884.7888 & 51945.6130 & 2000 dec 06 & 2001 feb 05 & 25 & 9.05 & 130 & 136 + & & @xmath1 & 51886.6803 & 51946.6051 & 2000 dec 08 & 2001 feb 06 & 21 & 8.91 & 180 & 176 + 45 & 15 & @xmath0 & 52178.8580 & 52214.7927 & 2001 sep 26 & 2001 nov 01 & 22 & 5.34 & 152 & 152 + & & @xmath1 & & & & & 22 & & 128 & 130 + 46 & 15 & @xmath0 & 52216.7903 & 52261.6769 & 2001 nov 03 & 2001 dec 18 & 17 & 6.68 & 168 & 150 + & & @xmath1 & & & & & 15 & & 120 & 118 + 47 & 15 & @xmath0 & 52266.7173 & 52313.6093 & 2001 dec 23 & 2002 feb 08 & 16 & 6.97 & 124 & 130 + & & @xmath1 & 52267.6603 & & 2001 dec 24 & & 15 & 6.83 & 154 & 130 + 48 & 16 & @xmath0 & 52539.8107 & 52595.7358 & 2002 sep 22 & 2002 nov 17 & 34 & 8.32 & 136 & 140 + & & @xmath1 & & & & & 33 & & 120 & 118 + 49 & 16 & @xmath0 & 52597.7373 & 52645.6685 & 2002 nov 19 & 2003 jan 06 & 29 & 7.13 & 100 & 102 + & & @xmath1 & & & & & 30 & & 88 & 94 + 50 & 16 & @xmath0 & 52649.6412 & 52677.5907 & 2003 jan 10 & 2003 feb 07 & 22 & 4.16 & 134 & 124 + & & @xmath1 & & & & & 21 & & 80 & 78 + 51 & 17 & @xmath0 & 52894.9328 & 52914.8841 & 2003 sep 12 & 2003 oct 02 & 17 & 2.97 & 96 & 94 + & & @xmath1 & & & & & 18 & & 164 & 188 + 52 & 17 & @xmath0 & 52915.8656 & 52941.8171 & 2003 oct 03 & 2003 oct 29 & 19 & 3.86 & 212 & 206 + & & @xmath1 & & 52944.8047 & & 2003 nov 01 & 20 & 4.30 & 134 & 132 + 53 & 17 & @xmath0 & 52947.7834 & 52986.7346 & 2003 nov 04 & 2003 dec 13 & 18 & 5.79 & 78 & 62 + & & @xmath1 & & & & & 19 & & 72 & 70 + 54 & 17 & @xmath0 & 52988.7597 & 53035.6267 & 2003 dec 15 & 2004 jan 31 & 21 & 6.97 & 184 & 190 + & & @xmath1 & & & & & 20 & & 118 & 118 + 55 & 18 & @xmath0 & 53255.8137 & 53284.8259 & 2004 sep 07 & 2004 oct 06 & 18 & 4.31 & 104 & 102 + & & @xmath1 & & & & & 18 & & 54 & 62 + 56 & 18 & @xmath0 & 53285.8297 & 53315.7442 & 2004 oct 07 & 2004 nov 06 & 17 & 4.45 & 70 & 70 + & & @xmath1 & & & & & 17 & & 68 & 62 + 57 & 18 & @xmath0 & 53326.7403 & 53355.6576 & 2004 nov 17 & 2004 dec 16 & 17 & 4.30 & 66 & 66 + & & @xmath1 & & & & & 18 & & 64 & 66 + 58 & 18 & @xmath0 & 53357.6446 & 53405.6231 & 2004 dec 18 & 2005 feb 04 & 20 & 7.14 & 114 & 128 + & & @xmath1 & & & & & 19 & & 146 & 156 + 59 & 19 & @xmath0 & 53521.9609 & 53566.8945 & 2005 may 31 & 2005 jul 15 & 21 & 6.68 & 136 & 134 + & & @xmath1 & & & & & 21 & & 124 & 124 + 60 & 20 & @xmath0 & 53627.6919 & 53646.7436 & 2005 sep 14 & 2005 oct 03 & 20 & 2.83 & 140 & 142 + & & @xmath1 & & & & & 20 & & 86 & 88 + 61 & 20 & @xmath0 & 53648.7212 & 53673.7304 & 2005 oct 05 & 2005 oct 30 & 16 & 3.72 & 74 & 92 + & & @xmath1 & & & & & 16 & & 42 & 52 + 62 & 20 & @xmath0 & 53676.7199 & 53704.6848 & 2005 nov 02 & 2005 nov 30 & 18 & 4.16 & 116 & 104 + & & @xmath1 & 53677.7290 & & 2005 nov 03 & & 14 & 4.01 & 98 & 106 + 63 & 20 & @xmath0 & 53706.6824 & 53734.6434 & 2005 dec 02 & 2005 dec 30 & 13 & 4.16 & 80 & 82 + & & @xmath1 & & & & & 14 & & 70 & 72 + 64 & 20 & @xmath0 & 53742.6353 & 53773.6036 & 2006 jan 07 & 2006 feb 07 & 17 & 4.61 & 62 & 54 + & & @xmath1 & & & & & 17 & & 86 & 84 + 65 & 21 & @xmath0 & 53875.9673 & 53906.8918 & 2006 may 20 & 2006 jun 20 & 18 & 4.60 & 72 & 76 + & & @xmath1 & 53873.9700 & 53907.9678 & 2006 may 18 & 2006 jun 21 & 20 & 5.06 & 90 & 96 + 66 & 21 & @xmath0 & 53995.8552 & 54031.7949 & 2006 sep 17 & 2006 oct 23 & 19 & 5.34 & 102 & 96 + & & @xmath1 & & & & & 19 & & 72 & 60 + 67 & 21 & @xmath0 & 54040.7623 & 54094.6978 & 2006 nov 01 & 2006 dec 25 & 26 & 8.02 & 108 & 106 + & & @xmath1 & & & & & 25 & & 92 & 90 + 68 & 21 & @xmath0 & 54103.6923 & 54136.6221 & 2007 jan 03 & 2007 feb 05 & 13 & 4.90 & 110 & 102 + & & @xmath1 & & 54134.6225 & & 2007 feb 03 & 12 & 4.60 & 104 & 88 @xcite produced doppler images of ii peg for 199298 using what they call the `` occamian '' approach . we comment here on how their images compare qualitatively to our photometric images obtained at nearly the same time . first , it should be noted that there is a consistent tendency for corresponding spots to appear at higher latitudes in their doppler images in comparison to our photometric images . this is likely due to our having only two filter passbands available , @xmath0 and @xmath1 . as shown in @xcite , having data through four passbands available ( e.g. , _ bvri _ ) significantly enhances the latitude resolution of the photometric inversions compared to when only two are used . furthermore , a spectral line profile is effectively a one - dimensional image of the stellar surface @xcite , while the individual data in a light curve are zero - dimensional images in that only the integrated brightness of the surface is indicated by them . thus , doppler imaging has intrinsically higher resolution than does photometric imaging , so it is reasonable to presume that the latitudes obtained from the doppler images are more reliable , and thus we caution the reader that the latitudes of the spots on our photometric images should not be accepted at face value . however , the extensive simulations reported in @xcite show that _ relative _ spot latitudes can be reliably determined from li : when two spots at different latitudes are present , the lower - latitude spot is generally rendered as such , even though the individual spot latitudes may not be accurately rendered , particularly when only two - filter data are available . thus , we should be able to detect differential rotation in our photometric images , though we can not go as far as to reliably determine the rotational period as a function of latitude on the stellar surface . the 1992 august image of @xcite compares favorably to our data set 12 ( 1992 september 7 october 3 ) photometric image , though the spot latitudes on the doppler images are considerably higher . in both images , a large spot has an appendage projecting toward lower longitudes . their 1993 december image shows two distinct high - latitude spots almost @xmath176 apart , with one moderately larger than the other . our set 18 ( 1993 december 24 1994 february 1 ) image shows what appear to be two widely - separated spots with a `` bridge '' connecting them . the bridge may be simply an artifact of the inversion , as such bridges are seen in simulations in which two spots close to one another are present on the artificial stellar surface . their 1994 november and 1995 january images show a pair of spots at higher latitudes than our group 8 ( 1994 october 13 1995 february 9 ) images . our images show a pair of spots connected by a bridge . their 1995 july image and our set 22 ( 1995 june 5 july 11 ) image both show a high - latitude feature which extends across a wide range in longitude . their 1995 october image and our group 10 ( 1995 september 22 1996 february 16 ) images show a pair of isolated spots well - separated in longitude , with one larger than the other . their doppler image for 1996 october shows ii peg shortly before one of the time intervals over which we see good evidence for differential rotation , that of group 11 ( 1996 november 4 1998 february 12 ) , and their images for 1997 june , august , and december show the star during the interval covered by our images . our results for this interval are described in [ sec : group11 ] below . similarly , their doppler images for 1998 october and november represent the star during another interval over which we saw evidence of differential rotation , that of group 12 ( 1998 september 29 1999 february 15 ) , which we discuss in [ sec : group12 ] . we defer the comparison of their doppler and our photometric images for these intervals to those sections . figure [ fig : group2plots ] shows plots of the light curves for the four data sets comprising group 2 , which span the interval from heliocentric mjd 47417.7681 47556.5813 ( 1988 september 11 1989 january 30 ) . also shown in figure [ fig : group2plots ] are the reconstructed light curves , i.e. , the light curves computed from the stellar surface maps , for the cases in which the assumed inclination of the rotation axis to the line of sight is @xmath150 . the intensities have been normalized such that the brightest datum over the entire data set considered in this study for the given filter is set equal to 1 . the quantity @xmath29 on each plot represents the rms deviation expressed in magnitudes between the data light curve and the reconstructed light curve . figure [ fig : group2 ] shows the reconstructed surfaces obtained from these light curves via the li procedure , with the results for an assumed axial inclination of @xmath150 shown in the top row , and for @xmath14 shown in the bottom row . each column of two images consists of the pair of @xmath177 and @xmath14 reconstructions for the date ranges indicated at the top of the column . note that if we let @xmath178 in the specific intensity distribution @xmath179 , where @xmath32 and @xmath180 are stellar latitude and longitude , and also reverse the direction of stellar rotation , the resulting rotational light curve is unchanged . thus , based on photometry alone we can not distinguish between a given brightness distribution and its mirror image . for definiteness , in generating our images we have adopted the convention that the star rotates counterclockwise as viewed from above the visible pole , so that spots would be carried across our view from left to right with the visible pole at the top of the image . we do not show here the other 64 light curves which we inverted . figure [ fig : group2plots ] is included so that the reader may get a sense of the comparison between the data and reconstructed light curves for a typical example . however , information about the characteristics of the source light curves along with the displayed longitude of disc center and assumed axial inclination is provided in the form of six numbers which appear below each of the reconstructed surface images we present . from left to right , these are as follows , using the values for the image at upper left in figure [ fig : group2 ] as examples : * `` 190 '' indicates that the longitude of disc center for the surface as displayed is @xmath181 . this is not to be confused with the actual sub - earth longitude on the stellar surface , which varies from observation to observation . unless otherwise indicated , the latitude of disc center for all surface images is @xmath182 , as opposed to the actual sub - earth latitude of @xmath183 . * `` 45 '' indicates that @xmath150 for the reconstruction . * `` 20 '' and `` 23 '' are the number of data points in the @xmath0 and @xmath1 light curves used for the reconstruction , respectively . * `` 268 '' is the `` effective rms noise '' expressed in magnitudes ( see [ sec : dataanalysis ] ) for the @xmath0-filter light curve , multiplied by @xmath175 , so that the actual effective rms noise is 0.0268 mag in this case . similarly , `` 326 '' indicates that the effective rms noise for the @xmath1-filter light curve is 0.0326 mag . since the @xmath0-filter data were of lower noise in this case , @xmath0 was designated the `` primary filter '' @xmath77 for this inversion in the sense of the discussion in [ sec : li_algorithm ] . note that the effective rms noise used in the inversion differs from actual deviation @xmath184 mag reported in figure [ fig : group2plots ] for the @xmath0 light curve of set 3 . the reason for this is discussed at the end of [ sec : dataanalysis ] . * `` 2.7 '' indicates that light curve data used to create the images span 2.7 revolution periods of the binary , which we assume also to be the rotation period of the star because of tidal locking . ( note that if there is differential rotation , this rotation period can strictly speaking only apply to some particular latitude . ) before discussing the images , it should be noted that simulations @xcite show that when two separate circular spots are close together , they may appear in the reconstructions as a single elongated spot because of limitations in the resolution of the method . indeed , what appears to be a single large spot may in fact be a magnetic active region having complex structure which can not be resolved . we will thus tend to use `` spot '' and `` active region '' somewhat interchangeably , but tending to prefer the latter terminology for a more extensive dark region on an image with a decidedly non - circular appearance . furthermore , as mentioned above , `` bridges '' which appear to connect spots together may be artifacts of the inversion procedure . the set 3 images for both @xmath150 and @xmath14 appear to show a lower - latitude spot at left and a higher - latitude spot at right connected by a bridge . the persistence of spottedness at high latitudes near the upper center of each image suggests that there is indeed a high - latitude magnetic active region in this location , and that the bridge is not simply an inversion artifact in this case . if ii peg exhibits differential rotation in the same sense as the sun in that lower latitudes have higher angular velocities , then we should see the longitude difference between the lower - latitude spot and the other features diminish over time . this does appear to happen as one peruses the images from left to right . in particular , the lower - latitude spot appears to catch up to and move underneath the high - latitude active region . this could represent two separate spots which appear merged in the images due to the limited resolution of the reconstructions , or it could represent a single large active region extended in latitude , which rotates ( roughly speaking ) about an axis through its center due to shearing by differential rotation carrying its southern end around the star faster than its northern end . the interpretation of the spot appearing at far right in the set 3 images is problematic because it clearly appears at a higher latitude than the spot on the left in this pair of images , but in the set 4 images the spot on the right appears to be at a comparable latitude to the one on the left . this may simply represent a limitation in the latitude discrimination of the spots in the inversions , which certainly is nt perfect , particularly given that only @xmath0 and @xmath1 light curves were inverted as opposed to having more filter data available . the spot on the right appears in the @xmath150 image for set 5 , but not in the @xmath14 image . in simulations it is seen that a small spot close to a much larger spot may be `` masked '' by the larger spot , causing the smaller spot not to appear in the reconstruction . that this spot appears in the @xmath14 image suggests that such masking has occurred for the @xmath150 image . it is important to note that sets 3 and 4 were acquired over a total span of 7.9 stellar rotations and the corresponding images differ noticeably , while set 5 by itself spans 8.2 rotations . it is thus likely that there was significant evolution of the spots during the time set 5 was acquired . in general , all the surface images shown here should be regarded as representing a sort of `` average '' appearance of the starspots during the interval over which the data were collected . it should also be noted that only twelve @xmath0 and thirteen @xmath1 observations were available for the set 6 time span . thus , some additional caution should be used in interpreting the corresponding images . figure [ fig : group2diffrot ] shows plots of the total span in longitude @xmath185 of the spotted region versus the heliocentric modified julian date of the midpoint of the time interval over which the data were collected corresponding to each image shown in figure [ fig : group2 ] . the plot on the left is for @xmath150 , while the one on the right is for @xmath14 . to obtain @xmath185 , an _ _ idl _ _ widget which displays the reconstructed image for user - adjusted values of the latitude and longitude of disc center was employed . the stellar equator was placed at disc center , and the longitude of disc center varied so as to determine the longitudes of the easternmost and westernmost dark patches within the spotted region . this is admittedly an imprecise determination , since there is no sharp delineation between what constitutes a `` spot patch '' versus a `` photosphere patch . '' nevertheless , in practice the edges of the spotted regions in the images are reasonably well - defined , as can be seen in figure [ fig : group2 ] and similar figures below . a line was then fitted to the data using a simple unweighted least - squares procedure . the slope of the line represents a measure of the rate at which the longitude span of the spotted region is changing , expressed in degrees per day . there is considerable variation between the results for @xmath150 and @xmath14 , but both plots behave as would be expected if ii peg exhibits differential rotation in the same sense as the sun : @xmath185 decreases with time , because the trailing region of the active region is at lower latitudes than is the leading region . in [ sec : discussion ] , we show that the fits shown in figure [ fig : group2diffrot ] and similar figures for the other data sets showing evidence of differential rotation are consistent with the conclusions of @xcite regarding the value of the differential rotation coefficient @xmath2 defined in equation ( [ eq : diffrotp ] ) . figure [ fig : group3 ] shows inversions of group 3 , consisting of three data sets obtained from mjd 47779.7480 47926.5752 ( 1989 september 10 1990 february 4 ) , with @xmath150 in the top row and @xmath14 in the bottom row . there are no light curve data for the interval 1989 january 31 september 9 . thus , no information is available regarding the evolution of the surface between the end of the time span covered in [ sec : group2 ] and that covered here . the images in figure [ fig : group3 ] give the distinct impression that the low - latitude spot on the right is moving away from higher - latitude active regions to its right due to differential rotation , though it should be said that the spot at far left in the third column appears to be a new feature which has emerged within the interval covered by these light curves . an interesting question is whether or not the low - latitude spot at right in the images in figure [ fig : group3 ] is the same as the low - latitude spot at left in the images in figure [ fig : group2 ] . then over the combined span mjd 47417.7681 47926.5752 ( 1988 september 11 1990 february 4 ) we see the low - latitude spot go from being well behind the high - latitude spot / active region in longitude to being well ahead of it , and thus a distinctive manifestation of differential rotation . the images are consistent with this hypothesis . the midpoint of data set 7 corresponding to the leftmost images in figure [ fig : group3 ] was at mjd 47798.8 , while the midpoint of data set 6 corresponding to the rightmost images in figure [ fig : group2 ] was at mjd 47451.1 , so that 257.7 days elapsed between the two midpoints . taking @xmath186 for @xmath150 from figure [ fig : group2diffrot ] as a rough estimate of rate at which the low - latitude spot should separate in longitude from the high - latitude spot / active region near the central meridian of each figure , we would expect the longitude of the low - latitude spot to have increased relative to the high - latitude spot by approximately @xmath61 in this time span . using @xmath187 for @xmath14 yields a relative increase in longitude of @xmath188 by the low - longitude spot . the actual amount by which the low - latitude spot is ahead of the high - latitude spot in the images for set 7 is about @xmath151 . if the images for set 6 in figure [ fig : group2 ] are taken to represent the two spots having approximately equal longitudes , then the low - latitude spot being @xmath151 ahead in set 7 is within the range of @xmath188 @xmath61 suggested by the plots in figure [ fig : group2diffrot ] . figure [ fig : group3diffrot ] shows how @xmath185 for the group 3 data sets varies in time , obtained via the same procedure as the plots in figure [ fig : group2diffrot ] . in this case we expect @xmath189 for solar - sense differential rotation , since the low - latitude spot now starts out ahead in longitude . the magnitudes of the slopes are comparable to those in the plots in figure [ fig : group2diffrot ] , strengthening the interpretation that we are seeing the effects of differential rotation throughout this data set and that discussed in the preceding section . figure [ fig : group11_45 ] shows the inversions of group 11 , consisting of seven data sets obtained from mjd 50391.8348 50856.5946 ( 1996 november 4 1998 february 12 ) , all for an assumed inclination of @xmath150 . figure [ fig : group11_60 ] shows inversions of the same light curves for @xmath14 . the first 4 images ( sets 27 30 ) in both figures appear to show a trailing low - latitude spot at left catching up with and passing under a high - latitude feature near the central meridian , highly suggestive of differential rotation in the same sense as that of the sun . from this we would expect to see the low - latitude spot begin to lead the high - latitude feature and thus appear to the right of it in the next three images ( sets 31 33 ) . in the @xmath177 image for set 31 , some hint that this is occurring is seen in the way that the dark southern part of the spotted region is tilted along a diagonal such that its southernmost extension is ahead in longitude . this is not seen in the corresponding @xmath14 image . the set 32 images for both inclinations are similar to the set 30 images . one may speculate that when previously separate large active regions are carried into proximity by differential rotation , they may merge or otherwise interact such that magnetic tension resists deformation or separation by the differential rotation of the surrounding plasma . if so , this would argue against our interpretation that the southern spot in the data set described in [ sec : group3 ] is the same as that seen in the [ sec : group2 ] data . however , it may be that whether or not this proposed suppression of differential rotation occurs depends on the details of the magnetic field configuration and the minimum separation achieved . the set 33 images show high - latitude activity leading the southern portion of the spotted region in longitude . this gives the appearance of the sense of the differential rotation having reversed , but it should be noted that both the @xmath0 and the @xmath1 light curves inverted to create these images had only 13 data points , which calls the reliability of these images into question . it is also possible that new high - latitude spot activity had simply emerged . this interpretation is supported by the appearance of the images for set 32 , particularly for the @xmath14 inversions . high - latitude activity does seem to be protruding at the northern end of the spotted region , which may represent the beginning of the emergence of new magnetic flux . @xcite present doppler images for 1996 october and 1997 june , august , and december , the first of which shortly precedes and the latter three of which overlap in time with the photometric images discussed in this section . we have no data for 1996 october , but their image gives some indication of the high - latitude region seen in between the two initial spots in our later images . as is typically the case , the spots seen in their doppler images are at higher latitudes than in our photometric images . their 1997 june image shows a large high - latitude active region with a large @xmath190 extent in longitude , along with a pair of lower - latitude spots . in their 1997 august image , the extent in longitude of the large active region has diminished considerably , and it appears that this is largely due to the disappearance of activity at the leading edge of the region . a smaller spot persists on the opposite side of the star . in their 1997 december image , the width in longitude of the large active region has further diminished , but this time it appears that either the activity along the trailing edge has diminished , or that the trailing edge has caught up to other parts of the region . there is a southward projection at the trailing edge which is not present in the 1997 august image . our set 29 ( 1997 may 22 july 13 ) images differ considerably from their 1997 june image . ours show a spot close to the equator trailing a higher - latitude active region . our set 27 ( 1996 november 4 december 18 ) images show the larger low - latitude spot present in their 1997 june image , which appears to fade in our set 28 ( 1996 december 19 1997 february 15 ) images . our set 30 ( 1997 september 23 november 3 ) , set 31 ( 1997 november 4 december 13 ) , and set 32 ( 1997 december 15 1998 january 24 ) images are comparable to their 1997 december image in showing a single large active region on one side of the star . the southern protrusion in their image could explain the fact that our images show activity down to near the equator , though in their image the protrusion only extends down to @xmath191 latitude . figure [ fig : group11diffrot ] shows plots of the variation in @xmath192 with time for the seven data sets in group 11 . the two plots at top include only the results for the first four data sets for each inclination , while the two at bottom include all seven sets . the fits are better for the former two since as discussed above , the trend toward narrowing of the spotted region in longitude ceases for the latter three data sets . figure [ fig : group12_45 ] shows inversions for @xmath150 of group 12 , consisting of four data sets obtained from mjd 51085.9791 51224.5942 ( 1998 september 29 1999 february 15 ) , while figure [ fig : group12_60 ] shows the same for @xmath14 . in both figures , the images in the top row have latitude @xmath193 at disc center , while for the images in the bottom row the north ( visible ) rotation pole is at disc center . for both assumed inclinations , the images show a leading low - latitude spot or active region at right and a trailing high - latitude spot / active region at left . as time progresses , the separation in longitude between these features increases , just as would be expected if solar - sense differential rotation is carrying the low - latitude feature around the star faster than the high - latitude region . this is most easily seen in the images looking down on the pole . note that the degree of magnetic activity appears to increase on both ends of the spotted region during this time span . @xcite present doppler images for 1998 october and november which overlap the interval under consideration here . their 1998 october image shows an active region spread extensively in longitude that is not resolved into separate spots ; in the november image , this region appears split into three separate spots . the difference is conceivably due to the variation in the noise artifacts between the two images . our data set 35 ( 1998 october 30 november 27 ) image shows a similar elongated feature , but lacks the southward protrusion seen in the middle region of the active region in the doppler images . figure [ fig : group12diffrot ] shows the variation of @xmath192 with time for this time interval . as expected , @xmath192 increases with time , since the low - latitude , faster - moving activity is leading in longitude . figure [ fig : group13 ] shows inversions of group 13 , composed of four light curves obtained between mjd 51429.8915 51586.5994 ( 1999 september 8 2000 february 12 ) , with @xmath150 for the top row and @xmath14 for the bottom row . for both assumed inclinations , the trailing edge of the spotted region appears to be at lower latitudes than the leading edge . this may represent two separate spots which are too close to be resolved , or two spots embedded within an active region . for differential rotation in the same sense as the sun , we expect that over time the separation in longitude between the leading and trailing edges ( or the two spots ) should diminish . this is what is seen . the last pair of images appear to show that the low - latitude trailing spot ( or extension of the spotted region ) has caught and passed underneath higher - latitude activity . @xcite produced surface maps of ii peg via doppler imaging for 1999 july august , 2000 february and 2001 november december , the first two of which are relevant to the present discussion . their 1999 july august image is qualitatively similar to our set 38 ( 1999 september 8 october 24 ) images , showing a broad region of high - latitude activity with southward projections at its leading and trailing ends , with the trailing end projecting further south . their 2000 february image primarily gives the appearance that the active region may have simply shrunk in longitude . however , it could also be argued that the trailing edge has caught up to the leading edge . figure [ fig : group13diffrot ] shows that the the extent in longitude @xmath185 of the spotted region described here diminishes with time , as expected for solar - sense differential rotation . the final time interval over which we saw good evidence for differential rotation was from mjd 51805.8758 51946.6051 ( 2000 september 18 2001 february 6 ) , which we partitioned into three data sets which were assigned to group 14 . figure [ fig : group14 ] shows the corresponding inversions , with assumed inclination @xmath150 for top row and @xmath14 for the bottom row . for both values of @xmath10 , a protrusion toward lower latitudes can be seen at the bottom of the spotted region . this protrusion appears to drift toward greater longitudes over the time interval covered by these observations , suggesting that it is a spot being carried around the equator faster than the rest of the spotted region because of differential rotation . however , we can not quantify this differential rotation by plotting the variation of @xmath185 with time as for the data groups discussed above , because in this case the total extent in longitude of the active region did not change significantly . the easternmost and westernmost fringes of the active region appear to lie at similar middle latitudes , so that we would not expect @xmath185 to change due to differential rotation here . the longitude of the southward protrusion is within the range of longitudes spanned by the activity to its north , so that its drift in longitude relative to the rest of the spotted region is not manifested in the variation of @xmath192 . instead , in figure [ fig : group14diffrot ] we show the variation in longitude of the middle of the southern protrusion with time . this is admittedly a vague conception due to the amorphous nature of the protrusion in these images , which is likely just an artifact of the limitations in the resolution of the inversions . the procedure was to use the same _ idl _ widget used to obtain @xmath192 for other data sets to estimate the longitude of the midpoint of the southern boundary of the protrusion . the same results were obtained for both @xmath150 and @xmath14 . as can be seen in figure [ fig : group14diffrot ] , the longitude of the protrusion ( or spot ) increases approximately linearly with time , as would be expected from differential rotation if the protrusion or spot maintains an essentially fixed latitude . in this study , we obtained images of the surface of ii pegasi by inverting data sets consisting of light curves collected over intervals of several months to look for evidence of differential rotation in the changes in the spot configurations . by contrast , @xcite inferred the presence of differential rotation using the very different approach of determining the rotation periods of individual spots observed over spans of several years . it is of interest to ascertain whether or not the results presented here are consistent with the earlier study . in contrast with the relation used by henry et al . given by equation ( [ eq : kf ] ) , in the present study it is more natural to characterize the differential rotation in terms of the rate of change in the difference in longitude between spots at different latitudes in our images . if for simplicity we express longitudes in a non - rotating frame of reference ( rather than in the co - rotating frame used in the rest of this work ) , the longitude as a function of time @xmath138 for a spot having latitude @xmath32 is just @xmath194 where @xmath195 is the longitude of the spot at time @xmath196 , and @xmath197 is the angular rotation frequency for latitude @xmath32 , with @xmath31 being the corresponding rotational period . from this the difference in longitude @xmath198 of two spots at latitudes @xmath199 and @xmath200 is simply @xmath201 where @xmath202 is the separation at @xmath196 and @xmath203 . thus , a plot of @xmath192 versus @xmath138 will have slope @xmath204 . in terms of the rotational angular frequency , equation ( [ eq : diffrotp ] ) becomes @xmath205 so that @xmath206 in principle , then , if we observe two spots at known latitudes and plot @xmath192 versus @xmath138 to find the slope @xmath204 , we can find @xmath2 from equation ( [ eq : kfromdeltaomega ] ) . in practice , this is not possible because we can not reliably determine precise spot latitudes from our photometric inversions . however , we can use the plots in figures [ fig : group2diffrot ] , [ fig : group3diffrot ] , [ fig : group11diffrot ] , [ fig : group12diffrot ] and [ fig : group13diffrot ] to make reasonable estimates of the value of @xmath2 . in these figures , @xmath192 represents the extent in longitude of an active region rather than the longitude difference between two spots , because the former could be more reliably estimated from our images . using the revolution period @xmath207 of the binary as our estimate of @xmath33 , we find @xmath208 . excluding the plots having seven data points in figure [ fig : group11diffrot ] , the slopes @xmath204 obtained from the plots for @xmath150 range from @xmath209 to @xmath210 , with a mean of @xmath211 . if we assume that @xmath212 and @xmath213 ( which is almost certainly an overestimate of the spread in latitude responsible for the change in @xmath192 , particularly given the results of doppler imaging ) , we obtain a minimum value for @xmath2 of @xmath214 , a maximum value of @xmath215 , and a mean value of @xmath216 . for @xmath14 , @xmath204 from the slopes of the plots ( again excluding the seven - data - point plot in figure [ fig : group11diffrot ] ) ranges from @xmath217 to @xmath218 with a mean of @xmath219 . these yield a minimum value for @xmath2 of @xmath220 , a maximum value of @xmath216 , and a mean of @xmath214 . these results accord well with the value @xmath40 obtained for ii peg by henry at al . , which lends credence to our assertion that our images represent genuine demonstrations of differential rotation , and to the assertion of henry et al . that stretching of active regions by differential rotation accounts for features of the light curve of ii peg not explainable by their simple two - spot model . of course , the values obtained for @xmath2 are increased if the span of latitude is not from pole to equator . the doppler images of @xcite show activity extending from mid - to - high latitudes , but not including any polar spots , so it seems reasonable to use @xmath221 and @xmath222 in our estimates . for @xmath150 , this results in minimum , maximum and mean values for @xmath2 of @xmath223 , @xmath224 , and @xmath225 , respectively . for @xmath14 the corresponding values are @xmath216 , @xmath226 , and @xmath227 . these results do not agree as well with the result of henry et al . , but are certainly of the same order . it should also be noted that henry et al.assumed that the spots they observed spanned a range in latitude from pole to equator , so that they too would have obtained a larger value of @xmath2 had they assumed a smaller range . n.v . acknowledges support from the ohio wesleyan university summer science research program . g.w.h . acknowledges support from nasa , nsf , tennessee state university , and the state of tennessee through its centers of excellence program . henry , g. w. 1995 , in robotic telescopes : current capabilities , present developments , and future prospects for automated astronomy , asp conf . g. w. henry and j. a. eaton ( provo : astronomical society of the pacific ) , p. 37 henry , g. w. 1995 , in robotic telescopes : current capabilities , present developments , and future prospects for automated astronomy , asp conf . vol . 79 , eds . g. w. henry and j. a. eaton ( provo : astronomical society of the pacific ) , p. 44	we present the results of a study of differential rotation on the k2 iv primary of the rs cvn binary ii pegasi ( hd 224085 ) performed by inverting light curves to produce images of the dark starspots on its surface . the data were obtained in the standard johnson @xmath0 and @xmath1 filter passbands via the tennessee state university t3 0.4-m automated photometric telescope from jd 2447115.8086 to 2454136.6221 ( 1987 november 16 to 2007 february 5 ) . the observations were subdivided into 68 data sets consisting of pairs of @xmath0 and @xmath1 light curves , which were then inverted using a constrained non - linear inversion algorithm that makes no _ a priori _ assumptions regarding the number of spots or their shapes . the resulting surface images were then assigned to 21 groups corresponding to time intervals over which we could observe the evolution of a given group of spots ( except for three groups consisting of single data sets ) . of these 21 groups , six showed convincing evidence of differential rotation over time intervals of several months . for the others , the spot configuration was such that differential rotation was neither exhibited nor contraindicated . the differential rotation we infer is in the same sense as that on the sun : lower latitudes have shorter rotation periods . from plots of the range in longitude spanned by the spotted regions vs. time , we obtain estimates of the differential rotation coefficient @xmath2 defined in earlier work by henry et al . , and show that our results for its value are consistent with the value obtained therein .
the emergence of cooperation in evolving populations with exploitative individuals is still a challenging problem in biological and social sciences . most theories that explain cooperation are based on direct reciprocity , as the famous iterated prisoner s dilemma @xcite . cooperation can also arise from indirect reciprocity when agents help others only if these are known as sufficiently altruistic @xcite . in most of these models a finite population of agents is simulated , pairs of agents meet randomly as potential donator and receiver . a donation involves some cost to the donor while it provides a larger benefit to the receiver . agents reproduce depending on their payoffs after a certain number of such meetings . obviously selfish individuals that do not donate would quickly spread in the population if help is not channeled towards more cooperative players . if agents do not meet repeatedly as in a large population direct reciprocity does not work . indirect reciprocity can solve this problems when donations are given only to those individuals that are known as sufficiently helpful . this mechanism effectively protects a cooperative population against exploiters @xcite . riolo et al . @xcite introduced a model in which cooperation is not based on reciprocity , but on similarity . in this model donations are channeled towards individuals that are sufficiently similar to the donator . to distinguish between different groups of individuals every agent @xmath0 has a tag @xmath1 $ ] . school ties , club memberships , tribal costumes or religious creeds are all tags that induce cooperation . in addition agents have a tolerance threshold @xmath2 , which determines the tag interval that the agent classifies as its own group . an agent @xmath0 donates to another agent @xmath3 if their tags are sufficiently similar , @xmath4 . the cost of such a donation for @xmath0 is @xmath5 while the benefit for @xmath3 is @xmath6 . for simplicity , @xmath7 is normalized to 1 , since a multiplication of payoffs with a constant factor does not change the game . initially , the tag and the tolerance threshold are uniformly distributed random numbers . in each generation every agent acts as a potential donor for @xmath8 other agents chosen at random . hence it is on average also chosen @xmath8 times as a recipient . after each generation each agent @xmath0 compares its payoff with the payoff of another randomly chosen agent @xmath3 and adopts @xmath9 and @xmath10 if @xmath3 has a higher payoff . in addition every agent is subject to mutation . with probability @xmath11 an agent receives a new @xmath12 drawn from a uniform distribution and also with probability @xmath11 a new @xmath13 which is gaussian distributed with standard deviation @xmath14 around the old @xmath13 . if this new @xmath13 becomes smaller than zero it is set to @xmath15 . obviously , it seems to be the best strategy for an individual to donate as little as possible , i.e. to have a very small @xmath13 . however , the whole population would be better off if everybody would cooperate . this `` tragedy of the commons '' can be solved in different ways , e.g. by volunteering @xcite . riolo et al . solve this problem by channeling help towards others that are sufficiently similar to the donator . instead of a cooperative population the formation and decay of cooperative clusters is observed for certain parameter ranges ( high @xmath8 and low @xmath16 , see fig . [ riolo ] ) . the average tolerance of a cooperative cluster grows slowly over time . occasionally it declines sharply . this decline occurs when the cluster is exploited by agents that are sufficiently similar to the cluster s agents to get support , but do not help themselves . however , the mechanism that generates these tides of tolerance remained unclear @xcite . [ ] [ ] [ 1][180]tolerance [ ] [ ] [ 1][180]donationrate , @xmath17 , @xmath18).,title="fig : " ] , @xmath17 , @xmath18).,title="fig : " ] here we develop a minimal model for tag - based cooperation that displays these `` tides of tolerance '' if there is a net average drift towards more cooperation . we find that these fluctuations vanish if such a drift is not included in the model . the importance of this observation stems from the fact that if we have species that can distinguish between themselves and others and donate only to others with the same tag , then this would in the long run lead to a single group of cooperating species having a single tag . but if we introduce a small rate of biased conversions from intolerant to tolerant species we observe a waxing and waning in time of species with different tags . in other words , the small conversion rate leads to a coexistence of different species where different species appear cyclically at different times . this consitutes a new mechanism that generates biodiversity in a group of competing species . this paper is organized as follows . first the model of riolo et al . is simplified in order to allow an analytical treatment . then the system without the effects of mutations is analyzed . thereafter we introduce a drift that increases the tolerance and leads to oscillations of tolerance . we show that the truncated mutations in the model of riolo et al . also lead to such a drift . here we simplify the model of riolo et al . @xcite in order to allow for an analytical treatment . in a first step we restrict the game to only two tags , red and blue . similarly we allow only two tolerances . the agents can either only donate to others bearing the same tag if they have zero tolerance @xmath19 or to every other agent ( @xmath20 ) . this leads to four possible strategies . then we allow partners to donate and to receive in an single interaction instead of defining different roles for donators and receivers . we end up with the payoff matrix [ cols="<,^,^,^,^",options="header " , ] which describes the prisoners dilemma @xcite . the stability of the fixed points with only one tag can be calculated as follows . for @xmath21 and @xmath22 we find the jacobian matrix @xmath23 with the eigenvalues @xmath24 , @xmath25 and @xmath26 . the fixed point is marginal stable as long as @xmath27 , for @xmath28 it becomes unstable . the reasoning can be adopted for the fixed line @xmath29 . a fixed point that is conserved for @xmath30 can be found if all players are tolerant . for @xmath31 the jacobian matrix is given by @xmath32 where @xmath33 and @xmath34 . the eigenvalues of this matrix are @xmath24 , @xmath35 and @xmath36 . @xmath37 ( @xmath38 ) is not possible for @xmath22 . hence the fixed line is instable for @xmath22 . for @xmath30 there is an interval of stability given by @xmath39 . if this inequation and @xmath40 are both fulfilled by @xmath41 , the biased conversions ensure stability of the fixed point although the replicator dynamics alone would make this point instable . the first inequation can only be fulfilled for @xmath42 . for @xmath43 it is always fulfilled and the whole fixed line @xmath44 is stable . the fixed point given by @xmath45 reduces to the mixed nash equilibrium for @xmath22 . the jacobi matrix at this fixed point is @xmath46 the eigenvalues of this matrix are @xmath47 where @xmath48 . for @xmath22 we have @xmath49 and @xmath50 . the third eigenvalue corresponds to an instable direction . the corresponding eigenvector is @xmath51 , which is the normal of the separatrix for @xmath22 . in the case of @xmath30 we have @xmath37 for @xmath38 only if @xmath52 . hence @xmath53 becomes stable where it coincides with the fixed points @xmath54 described in appendix [ addfpstab ] . in all other cases , at least one eigenvalue of @xmath55 is outside the unit circle . numerical simulations show that the additional fixed points for @xmath30 can always be found in the plane spanned by ( @xmath56 ) , ( @xmath57 ) and ( @xmath58 ) . together with @xmath59 and @xmath60 we have three equations that describe these points . two of the solutions are fixed points not described above . the first fixed points @xmath54 can be written as @xmath61 where @xmath62 and @xmath63 . @xmath64 can be calculated by exchanging @xmath65 with @xmath66 and @xmath67 with @xmath68 . these fixed points have only real coordinates for @xmath69 . for @xmath70 we have @xmath71 . the eigenvalues of the jacobi matrix at the fixed points @xmath72 can be calculated numerically . for @xmath73 the fixed points are only stable if @xmath74 . at @xmath75 they collapse with @xmath76 in a supercritical pitchfork bifurcation and form a single stable fixed point . for @xmath74 the fixed points @xmath72 are the only stable attractors and the order measures described in section [ cooperationcost ] can be calculated analytically . we find for @xmath77 @xmath78 @xmath79 for @xmath80 the fixed point @xmath53 becomes stable and we find @xmath81 , @xmath82 and @xmath83 . 99 r. axelrod , _ the evolution of cooperation _ ( basic books , new york , 1984 ) m. a. nowak and k. sigmund , nature * 393 * , 573 ( 1998 ) r. l. riolo , m. d. cohen , and r. axelrod , nature * 414 * , 441 ( 2001 ) c. hauert , s. de monte , j. hofbauer and k. sigmund , nature * 296 * , 1131 ( 2002 ) g. szab and c. hauert , phys . lett . * 89 * 118101 ( 2002 ) g. szab and c. hauert , phys . e. * 66 * , 062903 ( 2002 ) k. sigmund and m. a. nowak , nature * 414 * , 403 ( 2001 ) h. g. schuster , _ complex adaptive systems _ ( scator verlag , saarbrcken , 2002 ) j. hofbauer and k. sigmund , _ evolutionary games and population dynamics _ ( cambridge univ . press , cambridge , 1998 ) j. guckenheimer and p. holmers , _ nonlinear oscillations , dynamical systems and bifurcation of vector fields _ ( springer , new - york , 1983 ) h. g. schuster , _ deterministic chaos . an introduction . _ ( wiley - vch , weinheim , 1995 ) m. frean and e. r. abraham , proc . b * 268 * , 1323 ( 2001 ) b. kerr , m. a. riley , m. w. feldman , and b. j. m. bohannan , nature * 418 * , 171 ( 2002 ) a. traulsen and h. g. schuster , in preparation g. roberts and t. n. sherratt , nature * 418 * , 499 ( 2002 )	recently , riolo et al . [ r. l. riolo et al . , nature * 414 * , 441 ( 2001 ) ] showed by computer simulations that cooperation can arise without reciprocity when agents donate only to partners who are sufficiently similar to themselves . one striking outcome of their simulations was the observation that the number of tolerant agents that support a wide range of players was not constant in time , but showed characteristic fluctuations . the cause and robustness of these tides of tolerance remained to be explored . here we clarify the situation by solving a minimal version of the model of riolo et al . it allows us to identify a net surplus of random changes from intolerant to tolerant agents as a necessary mechanism that produces these oscillations of tolerance which segregate different agents in time . this provides a new mechanism for maintaining different agents , i.e. for creating biodiversity . in our model the transition to the oscillating state is caused by a saddle node bifurcation . the frequency of the oscillations increases linearly with the transition rate from tolerant to intolerant agents .
superconductivity was found in the sufficiently water - intercalated cobaltate compound , na@xmath0(h@xmath1o)@xmath2coo@xmath3h@xmath4o with the bilayer - hydrate ( blh ) structure@xcite . it is amazing that superconductivity is observed only in the hydrous phase with the blh structure . superconductivity has not been observed neither in the monolayer - hydrate nor in the anhydrous compounds . the superconductivity is realized on the two - dimensional coo@xmath4 layers , which consist of a triangular lattice . the superconducting pairing state in the blh compounds has been considered to be unconventional on the basis of the results of the power - law temperature dependence of the specific heat @xcite and the nuclear spin - lattice relaxation rate @xcite in the superconducting state . despite the intriguing physical properties both in the superconducting and normal state , the details still remain to be solved due to the difficulty in the reproducible sample preparation . in the blh compounds , the ground state strongly depends on the chemical compositions , na ion ( @xmath6 ) , oxonium ion h@xmath1o@xmath7 ( @xmath8 ) and water molecule ( @xmath9 ) contents . it is difficult to control these parameters precisely during the water intercalation . in addition , the water molecules easily evaporate into the air , when the samples were preserved in an ambient condition after the water intercalation . the unstable nature of the blh compounds causes the sample dependence of various physical quantities . the sample properties have to be clarified in detail both from the microscopic and macroscopic measurements before the investigation of the physical properties on the blh compounds . in order to summarize the sample dependence of superconducting transition temperatures , we have constructed a phase diagram of the blh compounds on the basis of the nuclear quadrupole resonance ( nqr ) measurements ( fig . [ phasediagram ] ( a ) ) . in the phase diagram , the co - nqr frequency @xmath10 was used as a promising reference for the ground state of the blh compounds @xcite , because the nqr frequency sensitively reflects the crystalline distortions around the co site , which are induced by the water intercalation . theoretically , the distortions have been predicted to have a relation with the formation of superconductivity@xcite . this phase diagram has been reproduced by the subsequent experiments @xcite , and extended to the higher frequency region . it is empirically revealed that the @xmath10 detects the sample dependence of parameters which are closely related to the formation of superconductivity . however , for the full understanding of the relationship between the ground state of the blh compounds and the nqr frequency , the effect of the hole doping to the coo@xmath4 layers has to be investigated in detail , because the nqr frequency depends also on the concentration of the on - site @xmath11 electrons in addition to the dominant lattice contributions . in this study , we performed the co - nqr and nuclear - magnetic - resonance ( nmr ) measurements on three different samples , which are located at three typical positions in our phase diagram . some results have been already published in refs . [ @xcite ] and [ @xcite ] . one of the samples , which possesses the longest @xmath5-axis length and the highest @xmath10 , demonstrates a magnetic transition at approximately 6 k. the effects of the small internal fields were observed on the co - nqr spectra below the temperature@xcite . the sample that has the shortest @xmath5-axis length and the lowest @xmath10 among the three samples shows the superconducting transition at the highest temperature @xmath12 k. in this sample , the nuclear spin - lattice relaxation rate divided by temperature @xmath13 was found to keep increasing down to 1.5 k without showing any anomalies , when the superconductivity is suppressed by the strong magnetic fields @xcite . the third sample possesses the intermediate values of @xmath5-axis length and @xmath10 . in zero fields , the superconducting transition at @xmath14 k and no trace of magnetic anomaly were observed from the nqr measurement on this sample . while the physical properties of this sample resemble in those of the superconducting sample in zero fields , a weak magnetic anomaly appears on this sample in the magnetic fields greater than 5 t. it is considered that the magnetic anomaly appears when the superconductivity is suppressed by the magnetic fields . here , we refer the magnetic ordering sample , the superconducting sample , and the intermediate sample as the mo sample , sc sample , and i m sample , respectively . the results of the magnetization measurements on the sc and i m samples have been already published in ref . [ @xcite ] . the sc and i m samples originate from the same batches as the samples referred as naoh @xmath15 ml and naoh @xmath16 ml in the report@xcite , respectively . the field - induced magnetic anomaly on the i m sample reminds us of the field - induced magnetism in ce - based heavy fermion superconductors cerhin@xmath17 @xcite and cecoin@xmath17 @xcite . it is considered that the superconductivity in cobaltate is also intimately related to the magnetism . and magnetic transition temperature @xmath18 of various samples reported in the literature . these values are plotted against ( a ) the nqr frequency and ( b ) the @xmath5-axis length . the circles and the triangles indicate @xmath19 and @xmath18 , respectively . the down arrow indicates that the superconducting transition was not observed down to 1.5 k. the superconducting transition temperature becomes maximum around @xmath20 . the magnetic phase appears in the colored region , where the @xmath5 axis is approximately @xmath21 . the samples plotted by the filled marks were used in this study . , width=264 ] co - nqr and nmr experiments were carried out with the conventional spin - echo technique . the typical values of the pulse widths and the separation between two pulses were @xmath22 @xmath23s and @xmath24 @xmath23s , respectively . the nqr spectra were obtained by recording the spin - echo intensity with varying the frequency . for the nmr measurements , the powder samples were aligned in a strong magnetic field and fixed by an unreactive material , which did not degrade the sample quality . due to the anisotropy of susceptibility , crystalline @xmath25-plane tends to be parallel to the field direction . as a result , magnetic fields were applied within the @xmath25 plane . the nuclear spin - lattice relaxation time @xmath26 was measured with saturation recovery method . in order to determine @xmath26 , the relaxation curves were fitted by the theoretical formula with single @xmath26 component . the experimental results were sufficiently fitted except for those obtained in the i m sample at low temperatures and strong magnetic fields . the details are explained in [ sec : nqr ] . it is required to investigate the sample properties carefully and confirm the reproducibility on the cobaltate compounds , especially on the hydrous phase . in order to compare the physical properties of our samples to those of others , the superconducting transition temperatures reported in the literature are plotted against the @xmath5-axis length of each sample in fig . [ phasediagram ] ( b ) . a relationship was observed between @xmath19 and the @xmath5-axis length . the scattered data points indicate that the @xmath5-axis length is not the only parameter that determines the ground state of the blh compound . the @xmath5-axis length , however , is considered to be the dominant parameter , and can be a useful reference to compare the sample properties of various reports . in the figure , superconductivity seems to be suppressed in some samples with @xmath27 , probably due to the appearance of the magnetism . the magnetism was reported to be observed on samples in the colored region @xcite . we also found some samples , in which both the magnetic and superconducting transitions are observed . it has not been revealed yet how these two transitions coexist in one sample . .the chemical and physical parameters of the superconducting ( sc ) sample , magnetic ordering ( mo ) sample , and intermediate ( i m ) sample . [ cols="<,<,<,<,<,<",options="header " , ] [ parameters ] the powder samples we measured were carefully characterized by using the inductively coupled plasma - atomic emission spectroscopy , redox titration , and x - ray diffraction measurements . the superconducting transition temperatures were determined by the onset of the meissner signal measured with the _ in situ _ nqr coil ( @xmath28 mhz ) . superconductivity was observed also from the dc magnetization measurements in @xmath29 t with a squid magnetometer . the superconducting transition temperature @xmath30 , na content @xmath6 , and co valence @xmath31 determined by these experiments are listed in table . [ parameters ] . the oxonium ion content @xmath8 is estimated from the relationship @xmath32 . the @xmath5-axis parameter , the nqr frequency , and the superconducting transition temperature of the i m sample are in - between the sc sample and the mo sample . this result suggests that the i m sample would be located in the vicinity of the phase boundary between the superconducting and the magnetic ordering phases . the filled marks plotted in fig . [ phasediagram ] ( b ) represent the positions for the samples studied here . k. , width=283 ] the co - nqr measurements were performed on three samples , and the significant difference was observed in the peak position of the nqr spectra . the nqr spectra of three samples are displayed in fig . [ nqrsp ] , where the peak positions of each spectrum are pointed by arrows . the co - nqr frequency @xmath10 was determined as the resonant frequency of the @xmath33 transitions . the values of @xmath10 on three samples are shown in table . [ parameters ] . on the basis of the nqr frequency obtained here , the positions of present samples are pointed by filled marks in our phase diagram ( fig . [ phasediagram ] ( a ) ) , where the i m sample was found to be located exactly at the phase boundary between the superconducting and magnetic ordering phases . the ground state of the i m sample in zero magnetic field is superconducting , because a clear superconducting transition was observed from the magnetization measurements , which is displayed in the inset of fig . [ nqrt1 ] . it should be noted that the meissner fraction of the i m sample is approximately @xmath34 % of that of the sc sample at @xmath35 k. since these two samples were synthesized from the identical starting material , the comparable diamagnetic signals indicate that the i m and sc samples possess a similar amount of superconducting volume fraction . the nqr spectra of the i m sample shown in fig . [ nqrsp ] do not depend on the temperature from 4.2 k ( circle ) to 1.5 k ( triangle ) . this result indicates that the static internal fields at the co site are absent even at @xmath35 k on the i m sample . we also noticed that the nqr spectral width of the i m sample is comparable to that of the sc sample ( square ) . the sample - independent spectral width suggests that the sample inhomogeneity does not cause any difference in the physical properties of three samples . the apparent spectral broadening was observed only on the mo sample below @xmath36 k , although the nqr spectrum does not show any clear splitting even at the lowest temperature . the spectral broadening is due to the distribution of the weak internal fields at the co site . the distribution was investigated from the nqr spectrum of the mo sample at 1.5 k. the details of the analyses are found in a separate paper@xcite , and explained briefly in [ sec : mo ] . on three samples , which are measured with the co - nqr technique . the superconducting and magnetic ordering temperatures are pointed by arrows . a diverging behavior was observed only on the mo sample . on the other two samples , power - law behavior was observed below @xmath19 . the inset shows the temperature dependence of magnetization measured in @xmath29 t with the squid magnetometer . , width=264 ] the magnetic ordering in the mo sample was also confirmed by the other quantity ; the nuclear spin - lattice relaxation rate @xmath37 , which was measured at the peak positions of the nqr spectra . the temperature dependence of @xmath37 is shown in fig . [ nqrt1 ] . the values of @xmath37 on the mo sample diverge at @xmath38 k where the nqr spectra starts to broaden . this behavior indicates that the electronic spins slow down toward the ordering temperature , because the low - energy excitations of the electronic spins near the magnetic instability give rise to the divergence of @xmath37 . on the other two samples , the divergence was not observed at all , and @xmath37 starts to decrease just below @xmath39 k and @xmath40 k for the i m and sc samples , respectively . the absence of the hebel - slichter peak @xcite and the power - law temperature dependence at low temperatures suggest that the blh cobaltate is an unconventional superconductor with line - nodes on the superconducting gap @xcite . the relaxation curves of the nuclear magnetization were sufficiently fitted by the theoretical function with single @xmath26 even in the superconducting state . this result confirms that both the sc and i m samples demonstrate superconductivity on the whole part of the samples . concerning with the results of the nqr measurements at zero magnetic field , the physical properties of the i m sample resemble in those of the sc sample . the difference between the i m and sc samples is observed only in the strong magnetic fields . the solid line and dotted line are the experimental result and the simulated result , respectively . the spectral shape is understood as the typical two - dimensional powder pattern . the sharp peaks arising from @xmath41cu , @xmath42cu in the coil and @xmath43na in the sample are observed at the positions pointed by arrows . the peak marked by asterisk is assigned to be an impurity phase . the inset is the close view of the center line ( @xmath44 transition ) . two - peak structure was observed . , width=264 ] the co - nmr measurements were carried out in the various external magnetic fields on the i m and sc samples . the external fields were randomly applied within the @xmath25-plane , which is perpendicular to the principal axis of the electric field gradient ( @xmath5 axis ) . the nmr spectrum was consistently explained by a typical two - dimensional powder pattern @xcite . the full nmr spectrum of the sc sample is displayed in fig . [ nmrsp ] . the dotted line displayed in fig . [ nmrsp ] is the simulated nmr spectrum on the above condition . most of the spectral peaks , except for the sharp peak marked with an asterisk , were accounted for by the calculation . we assigned the peak marked with the asterisk as the signal from an impurity phase , because the signal has no angle dependence and a long @xmath26 . the candidate for the impurity phase is the paramagnetic metal coooh , which would be formed by the excess of the water intercalation@xcite . it is noteworthy that no secondary phase was observed in our samples from the x - ray diffraction measurements . the signal intensity from the impurity phase was too small to affect the measurements on the center line . the resonance peaks arising from @xmath43na in the sample and from @xmath41cu , @xmath42cu of an nmr coil were also observed , but they were well separated from the center line . here , we concentrate on the center line arising from the @xmath45 transition , which is shown in the inset of fig . [ nmrsp ] . we found that the center line consists of two peaks pointed by arrows @xcite . the double - peak structure is ascribed to the in - plane anisotropy of the electric field gradient and knight shift . the nmr spectra of the sc and i m samples measured at above and below @xmath46 in various fields are shown in figs . [ scsp ] and [ imsp ] , respectively . the peak positions at @xmath47 are shown by arrows in fig . [ scsp ] . within the second order perturbation of the electric quadrupole interaction in the nuclear - spin hamiltonian , the difference of the magnetic fields ( @xmath48 ) between the double central peaks at the resonant frequency of @xmath49 is expressed as @xmath50 where @xmath51 and @xmath52 ( @xmath53 ) are knight shift and electric field gradient along @xmath54 directions , respectively , and @xmath55 is the nuclear gyromagnetic ratio . the knight shift is defined as the difference between the reference field ( @xmath56 ) and the resonance field divided by the resonance field , @xmath57 . it should be noted that the resonant frequency ( @xmath49 ) dependence of the first term , which originates from knight shift anisotropy , is opposite to that of the second term which represents the second - order effect of the electric field gradient . as increasing fields , the difference of the magnetic fields once becomes minimum around 4 t , and it becomes larger again above 7 t. the minimum of the splitting width is observed around 4 t , because the energy of the quadrupole interaction is comparable to that of the zeeman interaction around this field . , respectively . peak positions at high temperatures are indicated by down arrows . , width=264 ] the appearance of superconductivity is detected from the shift of the nmr spectra , because the in - plane spin susceptibility decreases below @xmath19@xcite . as shown in figs . [ scsp ] ( a ) , ( b ) and ( c ) , the nmr spectra shift to the higher fields below @xmath19 in the fields smaller than 7 t , due to the decrease in spin susceptibility in the superconducting state . the spectral shift was observed on all the spectral weight . this result clearly indicates that the superconductivity is realized in the whole part of the sc sample . in 10 t , the shift of a spectral maximum is hardly detectable , but a small change in the spectral shape , which is related to the decrease of spin susceptibility , was observed . the small change is indicated by the horizontal arrows in fig . [ scsp ] ( d ) at the slope of the spectrum . the spectral change becomes smaller than the experimental accuracy in high fields , because the change in spin susceptibility becomes subtle near the upper critical field , while the spectral width becomes broad in high fields due to the distribution of knight shift . no obvious difference between the nmr spectra at 3.8 k and at 1.9 k in fig . [ scsp ] ( e ) shows that superconductivity does not occur down to 1.9 k in 14 t. it should be noted that the spectral shape does not change at low temperatures and high fields in the sc sample . this result indicates that the temperature dependence of knight shift and electric field gradient is negligibly small in this temperature and field ranges . k ) and the low temperatures ( @xmath58 k ) are shown in each figure . peak positions of the right peak at high and low temperatures are pointed by down arrows . , width=264 ] in the i m sample , the spectral shift was also observed below @xmath19 at a low field ( @xmath59 t ) , indicating that the knight shift decreases in the superconducting state due to the decrease in spin susceptibility . it should be noted that the spectral shape and width does not change passing through the superconducting transition at this field . the invariance of the spectral shape indicates the absence of magnetic transitions in the fields smaller than 2 t. the shift of the whole spectral weight and the large meissner fraction , which is comparable to that of the sc sample , confirm that the i m sample demonstrates superconductivity on the whole sample at @xmath60 t. when the external fields are greater than 5 t , the obvious change in the spectral shape was observed at low temperatures as shown in figs . [ imsp ] ( b ) , ( c ) and ( d ) . the spectral shape varies due to the change in the distribution of the hyperfine fields at the co site . the variation in the spectral shape in the i m sample is apparent at low temperatures , when the temperature variation in the spectral shape is compared with those obtained in the sc sample at high fields , which are shown in figs . [ scsp ] ( d ) and ( e ) . the field distribution originating from the vortices is estimated to be the order of @xmath61 t at this field range@xcite , when the typical penetration depth ( @xmath62 nm ) and coherence length ( @xmath63 nm ) for the blh compounds@xcite are taken into account . the spectral broadening in the i m sample is considered to be due to the appearance of static magnetic fields at the co site . the broadening can not be explained by the magnetism on the small portion of the sample , which does not show superconductivity at all . in 13 t , the nmr spectral intensity of the i m sample starts to decrease below @xmath64 k due to the short spin - spin relaxation time near the magnetic instability . the spectral intensity gradually recovers at low temperatures . the smallest spectral intensity is @xmath65 % of that obtained at high temperatures . this result suggests that the majority of the sample is in the magnetic state below @xmath64 k. while the peak position shifts to the higher fields below 4 k in 5.6 t , the peak position shifts to the lower fields , when @xmath66 t is applied . it is considered that the knight shift increases at low temperatures . the details of the temperature dependence of knight shift are presented in [ sec : mo ] in the sc sample at various magnetic fields . the arrows indicate the superconducting transition temperatures at each field , which was determined by the deviation of @xmath13 from the normal - state temperature dependence . the dashed line is the fitting function described in the text . , width=264 ] the temperature dependence of @xmath13 was measured on the sc and i m samples in various fields up to 14 t. in the sc sample , the relaxation curves were consistently fitted by the function for the relaxation of nuclear spins with single @xmath26 component . this function is written as @xmath67 where @xmath68 is the time after the saturation pulse@xcite . the coefficient @xmath69 is a fitting parameter , which possesses the value close to @xmath70 . the relaxation - rate measurements were performed on the left - side peak of the center spectrum for the sc sample , while @xmath26 was measured on the right - side peak for the i m sample . the relaxation rate could not be measured on the left - side peak in the i m sample , because the double peak structure becomes ambiguous at low temperatures in high fields . the similar temperature dependence of @xmath26 was observed for both peaks in the sc sample , although the absolute values were approximately twice larger on the left - side peak than on the right - side peak . the temperature dependence of @xmath13 in the sc sample is displayed in fig . [ sct1 ] , together with that obtained in zero magnetic field with the nqr measurement . the arrows indicate the superconducting transition temperatures in each field , which were determined as the temperature where @xmath13 starts to deviate from the normal - state temperature dependence . when the superconductivity is suppressed by the strong magnetic fields , the values of @xmath13 continue to increase extending the temperature dependence in the normal state . in the sc sample , no anomaly other than the superconducting transition was observed from the relaxation measurements even in the field of 14 t. this result is consistent with that of the spectrum measurements , which does not represent any spectral broadening at low temperatures . . the experimental data can not be fitted by the single @xmath26 at low temperatures . inset shows the temperature dependence of @xmath13 at 5.6 t , 10.1 t , and 0 t(nqr ) . the absolute value of @xmath13 obtained with nqr measurements was normalized to those with nmr measurement . , width=264 ] the physical properties of the i m sample are quite different from those of the sc sample in the magnetic fields . at the temperatures above approximately @xmath64 k , the relaxation curves were consistently explained by the theoretical function written in equation ( [ fn : relax ] ) . however , the experimental results could not be fitted by the same function , when the i m sample was cooled down below @xmath64 k. the results of the least square fitting to the whole relaxation curves obtained at 2.5 k and 5.5 k are exhibited in fig . the insufficient fit at 2.5 k is considered to be due to the appearance of short-@xmath26 components , which originate from the ordered magnetic moments . in fig . [ imt1 ] , the relaxation curves are plotted against the product of the time and the temperature , in order to represent the existence of the short-@xmath26 components at low temperatures . if a short-@xmath26 component contributes to the relaxation curve , the slope would become steep . obviously , the relaxation curve obtained at 2.5 k indicates the presence of short-@xmath26 components in the short time regime . in addition , the existence of the long-@xmath26 components is also observed in the long time regime . the relaxation rate in the magnetically ordered state is distributed in space probably due to the distribution of the internal fields at the co site . the relaxation curve in the magnetic state can not be explained even with two distinct components ( long and short @xmath26 ) . it is considered that the @xmath26 values are continuously distributed . the spatial distribution could not be specified with the present nmr experiments on powder samples . we fitted all the relaxation curves to the same theoretical function with the single @xmath26 component even in the magnetically ordered state to determine the typical @xmath26 values . the least square fits with using full time range are considered to give the relaxation rates on the major fraction of the sample , although the results include the large errors below @xmath64 k. the temperature dependence of @xmath13 is displayed in the inset of fig . the absolute values of the nqr results , which are shown together , were normalized to those of the nmr results . the temperature dependence of @xmath13 above @xmath71 k is the same among the measurements in three different fields . in the normal state , where the relaxation curves were consistently fitted by the theoretical curve , @xmath13 has a tendency to diverge toward @xmath64 k , and an anomaly is observed at @xmath64 k , when the superconductivity is suppressed by the strong magnetic fields . the temperature dependence in the normal state suggests that a magnetic transition occurs at @xmath64 k. the appearance of the anomaly in the i m sample can not be explained by the sample inhomogeneity , because the anomaly is absent in any fields in the sc sample , whose sample inhomogeneity is comparable to that of the i m sample . in the previous report@xcite , we investigated the temperature dependence of @xmath72 in the normal state from the co - nqr measurements . we found that the sample - dependent behavior of @xmath13 can be consistently fitted by a function with two fitting parameters @xmath73 and @xmath74 , which is expressed as @xmath75 the first term @xmath76 is expressed as @xmath77 which exhibits a pseudo - gap behavior above 100 k. here , @xmath73 is a proportionality constant related to the band structure at the fermi level and to the hyperfine coupling constant . the other parameter @xmath74 is the ordering temperature for the magnetically ordering samples and the measure of the closeness to the magnetic instability for the samples without magnetic ordering . the pseudo - gap behavior is commonly observed in mono - layered hydrate and non - hydrate cobaltates@xcite as in blh compounds above 100 k , while the additional second term expressed as @xmath78 is observed only in blh compounds at low temperatures . based on the framework of the self - consistent renormalization ( scr ) theory , the functional form of the second term is expected when the magnetic fluctuations possess three - dimensional antiferromagnetic correlations@xcite . the temperature dependence of @xmath13 in the sc sample under the magnetic fields along @xmath25 plane can be explained by the parameters @xmath79 and @xmath80 k , respectively . the parameters for the i m sample are @xmath81 and @xmath82 k. the different @xmath73 values are due to the angle dependence of the hyperfine coupling constants . as shown in fig . [ sct1 ] , @xmath13 in the sc sample increases with decreasing temperature in the normal state , and the increase is interrupted by the onset of superconductivity . it is noteworthy that @xmath13 continues to increase in the field - induced normal state , and follows the fitting curve defined above 5 k. this is a great contrast to the korringa behavior ( @xmath83 = const . ) , which would be observed when the electron - electron interactions are weak enough to form the fermi - liquid state . the continuous increase in @xmath72 down to @xmath84 indicates the existence of the strong electronic correlations in the sc sample . the temperature dependence of @xmath13 , which is explained by the equation ( [ fn : scr ] ) with @xmath85 k , strongly suggests that the electronic state of the sc sample is close to the quantum criticality , which is characterized by @xmath86 k. we expect that the quantum critical point exists in a sample that has the similar chemical composition as the sc sample . the precise sample control is required to explore the quantum critical point . the temperature dependence of @xmath13 reminds us of the specific - heat results on cecu@xmath87si@xmath87@xcite and cecoin@xmath88@xcite in magnetic fields , both of which are situated at or near the quantum critical point . k. the clear spectral broadening was observed at @xmath35 k. in the inset , the estimated internal field distribution is exhibited . , width=264 ] the broadening was observed in co - nqr spectra below @xmath89 k in the mo sample . the distribution of the internal fields was estimated from the spectra . in general , the nuclear spin hamiltonian is written as @xmath90 where @xmath91 and @xmath92 are the electric field gradient and the nuclear quadrupole moment , respectively . the first term of the right hand side expresses the electric quadrupole interaction , and the second term indicates the zeeman interaction originating from the internal fields . below a magnetic ordering temperature , the zeeman interaction emerges and increases due to the appearance of the internal fields at the nuclear site . the relationship between the resonant frequency and the internal fields for the mo sample is derived from this hamiltonian . first , we determined the direction of the internal field to @xmath93 , where the @xmath5 axis is the principal axis of the electric field gradient , because the resonance peak arising from @xmath94 transitions become broader than that arising from @xmath95 transitions . if the internal fields were along the @xmath5 axis , @xmath96 transition line would become the broadest within the three nqr lines . obviously , this is not a case . the anisotropic broadening of @xmath96 transitions below @xmath89 k can also be consistently explained by assuming @xmath93 . these two results suggest that the internal fields direct to the @xmath25 plane . next , the intensity of the internal fields is estimated using the relationship between the resonant frequency and the internal field strength , which is derived from the equation ( [ fn : mo ] ) . the details of the analyses are described in the separate paper@xcite . for the estimation of the fraction , we used the nqr spectra in the frequency range of @xmath97 mhz , because almost linear relationship was observed between the frequency and the internal field strength in this frequency range . the result is exhibited in the inset of fig . [ monqrsp ] . it was found that the maximum fraction is at zero field , and that rather large fraction is in the small field region . we also point out that a weak hump was observed on the fraction around @xmath98 t , which corresponds to 0.1 @xmath99 when we adopt 1.47 t@xmath100 as the coupling constant @xcite . it should be noted that nqr measurements were performed on the co site , where the magnetic moments are located . the distribution of the hyperfine fields at the co site suggests that the size of the ordered moments has the spatial distribution . the magnetism in the blh compounds is considered to be a spin - density - wave type with the modulation on the size of the ordered moments . in order to investigate the magnetic structure in detail , neutron diffraction measurements are required . t on the sc sample . ( b ) temperature dependence of the spectral width , which was characterized by the full width at three quarters maximum ( fw3/4 m ) . the symbols are the same as those in ( a ) . a magnetic anomaly was observed at @xmath64 k in the i m sample . ( b)-inset . field dependence of fw3/4 m at 1.5 k on the i m sample . , width=264 ] the internal fields are observed also in the i m sample , when the magnetic fields greater than @xmath101 t are applied . the knight shift of the i m sample scaled at @xmath89 k @xmath102 is displayed in fig . [ shift ] ( a ) , where @xmath103 measured on the sc sample at @xmath104 t is shown together . compared with the result of the sc sample , the knight shift on the i m sample slightly increases below @xmath64 k. it seems that the magnetic structure has a ferromagnetic component , because the magnetization increases below @xmath64 k. the temperature dependence of the magnetization measured on the i m sample was presented in ref . [ @xcite ] . the increase in the magnetization was also observed below approximately @xmath64 k , when the external fields greater than @xmath101 t are applied . the resemblance of the temperature dependence between the bulk magnetization and the knight shift suggests that the increase in knight shift is an intrinsic behavior . in addition to the temperature dependence of @xmath105 , we investigated the temperature dependence of the spectral broadening in several fields by evaluating the spectral width at the three - quarter intensity of the highest peak . we used the full width at the three - quarter maximum ( fw3/4 m ) in order to minimize the spectral width arising from the in - plane anisotropy of the knight shift and the electric - field gradient . these contributions are small near the peak intensity . the line width estimated near the peak intensity originates mainly from the distribution of the hyperfine fields at the co site . we defined the value @xmath106 as the difference of fw3/4 m from that at @xmath101 k @xmath107 which are plotted as a function of temperature in fig . [ shift ] ( b ) . the spectral broadening due to the electric field gradient is completely canceled by subtracting the fw3/4 m at @xmath101 k , because the electric field gradient does not depend on temperatures below @xmath108 k. the temperature dependence of @xmath106 possesses the magnetic origin . at @xmath60 t , the spectral width slightly decreases below @xmath19 , because the spin susceptibility decreases in the superconducting state . the decrease in the spectral width in the superconducting state is observed when the spectral width is determined mainly by the inhomogeneity of the knight shift . this result indicates that the magnetic - field distribution generated by the vortices in the superconducting state is smaller than those arising from the distribution of knight shift even in @xmath60 t. in the fields greater than @xmath101 t , the fw3/4 m starts to increase below @xmath64 k , where the anomaly was also observed in the temperature dependence of @xmath13 . it is considered that the static internal fields cause the broadening on the nmr spectra of the i m sample . the field dependence of the onset temperature , where the spectra start to broaden , is considerably weak in the field range @xmath109 t. the fw3/4 m of the sc sample measured in @xmath104 t does not represent any anomaly at @xmath64 k , but shows a gradual increase , which is roughly proportional to the bulk susceptibility . the spectral width at @xmath35 k @xmath110 is plotted as a function of external fields in the inset of fig . [ shift ] ( b ) . a clear increase is observed at 4 t , and the values are saturated above 6 t. the saturated value is approximately 0.02 t , which corresponds to the ordered moment of 0.01 @xmath99 by assuming the hyperfine coupling constant being 1.47 t/@xmath111@xcite . it can be considered that the weak magnetic ordering , which possesses ordering temperature @xmath112 k , is masked by the superconductivity with @xmath71 k. the magnetism appears when the superconductivity is suppressed below @xmath18 by external magnetic fields . the emergence of the internal fields in the external magnetic fields and the clear superconducting transition in zero field indicate the intimate relationship between superconductivity and magnetism . the similar field - induced magnetism has been observed in cuprate and heavy - fermion superconductors , in which superconductivity occurs near the magnetic instability @xcite . intriguingly , the spectral broadening was observed at the same time as the decrease in the knight shift due to superconductivity at 5.6 t. in this field , superconductivity coexists with magnetism within the i m sample . one might consider that the i m sample includes two different phases , which are the superconducting phase and the magnetic phase . this possibility , however , can be ruled out by the large superconducting volume fraction of the i m sample at low fields , which is comparable to the sc sample , and the obvious change in the nmr spectra at high fields . the i m sample varies the ground state depending on the external fields . at the intermediate fields @xmath113 t , superconductivity coexists with magnetism there are two candidates to comprehend the relationship between magnetism and superconductivity in the i m sample . one is the uniform coexistence , where the ordered moments appear in the superconducting state . the other possibility is that the magnetism occurs in the region , where the superconductivity is suppressed , such as the vortex cores , and is apart from the superconducting region . in the present study , the spectral broadening with magnetic origin was not observed in the superconducting state at low fields , and was suddenly observed when the superconductivity was suppressed below approximately @xmath64 k. our experimental result might suggest that the magnetism can not coexist with superconductivity uniformly , but occurs in the vortex core . however , it is difficult to observe spatial distribution of the electronic state from the broad nmr spectra of the powder samples . nmr measurements on single crystals are required to obtain the sharp nmr spectra and to investigate the spatial distributions of @xmath26 in high fields . we performed co - nqr and nmr measurements on three blh na@xmath0(h@xmath1o)@xmath2coo@xmath3h@xmath4o samples , which demonstrate various ground states . the experimental results presented here suggest that the superconductivity in blh compound is intimately related to the magnetism . since a subtle change in the chemical compositions strongly affects the physical properties , it is required to characterize the samples from both macroscopic and microscopic measurements . a magnetic transition was observed in zero magnetic field in the mo sample . in the mo sample , it was found that the spatially distributed internal fields appear below @xmath89 k. in the sc sample , which shows the highest @xmath12 k , the magnetic fluctuations were found to keep increasing down to @xmath35 k , when the superconductivity was suppressed by the strong magnetic fields . the temperature dependence of @xmath13 suggests that the sc sample is located in the vicinity of the quantum critical point . in the i m sample , which possesses the intermediate parameters , a magnetic anomaly was observed only in the magnetic fields greater than @xmath101 t , while this sample shows the clear superconducting transition at @xmath114 k and no trace of the magnetism at zero field . we point out that the physical properties of the i m sample in the magnetic fields are quite different from those of the sc sample , although the properties in zero magnetic field are similar . it seems that the magnetism in the i m sample appears in the field - induced normal state , such as at the vortex core , but our experimental resolution on the powder sample was not enough to determine how these two phases coexist . co - nmr measurements using single - crystalline sample are needed to explore the details of this magnetic anomaly . we thank h. ohta , c. michioka , y. itoh , h. yaguchi and y. maeno for experimental support and valuable discussions . we also thank h. ikeda , k. yamada , y. yanase , m. mochizuki and m. ogata for valuable discussions . this work was partially supported by crest of the japan science and technology agency ( jst ) and the grant - in - aid for the global coe program `` the next generation of physics , spun from university and emergence '' from the ministry of education , culture , sports , science and technology ( mext ) of japan , and by grants - in - aid for scientific research from the japan society for the promotion of science ( jsps)(no.16340111 and 18340102 ) and mext(no.16076209 ) .	in order to investigate the relationship between superconductivity and magnetism in bilayer - hydrate cobaltate na@xmath0(h@xmath1o)@xmath2coo@xmath3h@xmath4o co nuclear magnetic resonance ( nmr ) and nuclear quadrupole resonance ( nqr ) measurements were performed on three different samples , which demonstrate various ground states at low temperatures . the appearance of small internal fields is observed in the nqr spectra below approximately 6 k on one of the samples that possesses the largest @xmath5-axis length and the highest nqr frequency . the other two samples exhibit superconducting transition in zero magnetic field , while these two samples show different ground states in the magnetic fields greater than 5 t. the comparison of the nmr spectra of these two samples obtained in high magnetic fields reveals the appearance of static internal magnetic fields at the co site below 4 k in the sample that possesses the intermediate @xmath5-axis length and the nqr frequency .
a most general description of a planet is that it is a self - gravitating object composed of a heavy element core and an envelope of gas . terrestrial like planets are dominated by solid cores whereas giant gas planets are mainly hydrogen gas . given these observational facts , it should come as no surprise that there are two competing scenarios for planet formation that take opposite views on what comes first ( for a recent review see * ? ? ? * ) . in the top - down scenario hydrogen gas cloud comes first and the solid element core is assembled later . in the bottom - up picture the solid core must be made first before a bound gas envelope appears . in the former scenario planets mass , especially gas , as they mature , whereas in the latter planets gain mass with time . the top - down hypothesis takes roots in the @xcite proposition that planets begin their existence as self - gravitating condensations of @xmath1 jupiter masses of gas and dust formed in the solar nebula by gravitational instability ( gi ) of the disc . @xcite showed that microscopic grains grow and sediment to the centre of such gas clumps within a few thousand years , presumably forming a massive solid core there ( cf . also * ? ? ? * ; * ? ? ? these cores are the seeds of terrestrial - like planets in the model . @xcite and @xcite proposed that these cores could be all that is left of the original gas protoplanets if the dominant gas component is disrupted by tidal forces from the sun ( this process was rediscovered by * ? ? ? it is natural in this picture that giant planets in the solar system are located further away from the sun than terrestrial - like planets . on the other hand , @xcite instead posited that microscopic dust grains in the protoplanetary disc grow larger and then somehow collect into huge rocks of at least @xmath3 1 km size , called planetesimals . these coalesce into even larger solid cores . low mass cores become terrestrial planets . largest cores , of mass @xmath4 ( 10 earth masses ) , attract gaseous atmospheres from the protoplanetary disc and end up as giant gas planets ( e.g. , * ? ? ? this bottom - up scenario is now called core accretion ( e.g. , * ? ? ? * ; * ? ? ? * ) and is by far the most popular planet formation theory . differentiation between these two theories was thought to be straight forward based on the solar system data . gi unstable discs were argued not to produce self - gravitating clumps at all at the location of the solar system planets due to an inefficient disc cooling @xcite , so the sequence of events envisioned by the top - down picture could not be even started . ca picture , on the other hand , proved quite successful in explaining the solar system @xcite . however , the above criticism of the top - down scenario neglects the possibility of planet migration ( that is , shrinking of the planet s orbit due to gravitational torques from the protoplanetary disc , see * ? ? ? * ; * ? ? ? ca planets were equipped with migration since @xcite , who showed convincingly that hot jupiters could not have formed at their observed planet - star separations , @xmath5 au , and are more likely to have migrated from their assumed birth location of @xmath6 au . in contrast , the role of migration for gi planet formation model was not appreciated until @xcite . these authors pointed out that gravitationally unstable discs do not actually need to form gas clumps at @xmath3 a few au to explain the observed planets there : in simulations , most gi fragments are found to migrate rapidly from their birth locations at @xmath7 au into the inner disc ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? it is hence plausible that all of the observed giant planets were hatched by gi in the outer disc and were then pulled much closer to the host star by the gravitational disc torques . furthermore , some of the fragments could give birth to earth - mass or more massive cores by grain growth and sedimentation , and be tidally disrupted , potentially providing a `` new '' pathway au to arbitrarily close to the star . ] to forming all kinds of planets at all separations in a single framework that was called `` tidal downsizing '' ( td ) . we note in passing that @xcite recently presented the results of the palms survey which shows that the frequency of giant gas planets at large separations ( @xmath8 au ) is very low , e.g. , less than @xmath9% , implying that the `` poster child '' gi - planet system hr 8799 @xcite is very rare . @xcite conclude that `` disc instability is not a common mode of giant planet formation '' . in the context of td hypothesis , the observations of @xcite , unfortunately , do not tell us directly about disc fragmentation properties at these separations ; they rather show that gi - planets rarely survive at their initial large separations to the present day . in fact , given that the inward migration times of gi planets are as short as @xmath10 million years @xcite , it has been argued that it is not clear how any of the observed gi planets ( such as the multi - planetary system hr 8799 ) survive . for example , @xcite found that all of their 13 simulated gas clumps were either tidally destroyed , migrated into the inner few au , or became brown dwarfs due to gas accretion . observations of @xcite are therefore in excellent agreement with expectations of td . any gi model that does not include migration of planets should be treated with suspicion at this day and age when an extremely rapid migration of giant planets in self - gravitating accretion discs has been confirmed by around a dozen independent research teams ( to add to the references above , see also * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the potential of the top - down scenario remains poorly explored to this day , mainly because understanding of different aspects of the model is not yet complete , and connecting them together in order to make solid observational predictions is not trivial . the first population synthesis model for td hypothesis was presented by @xcite , who used a semi - analytical approach to disc evolution and an analytical description for fragment contraction , grain growth and the resultant core formation . the results of their study are not particularly encouraging for td . many fragments were found to run away in mass by accreting gas and became brown dwarfs at large separations from the parent star . others were disrupted by tides before they could assemble a massive core . while solid cores with masses up to @xmath11 were assembled inside the fragments by grain sedimentation , most of them were locked inside brown dwarf mass fragments . out of a million gas fragments simulated , only one yielded an earth mass core _ without _ an overlaying massive gas envelope . more recently , population synthesis was also performed by @xcite , who used different prescriptions for fragment cooling and did not address core formation . in contrast to @xcite , the latter study found that most of the gas fragments yield hot jupiters rather than brown dwarfs at large separations . this large difference in the results of the two studies motivates the present paper , which improves on both by using 1d numerical rather than analytical descriptions of the coupled disc and planet evolution . the eventual goal is to have a physically constrained and therefore predictive model of the td hypothesis , put in a computationally expedient framework , and enabling statistical comparisons to the observed planetary data . while a detailed comparison to the data is to be reported in follow up publications , as a way of illustrating the methods and the information obtained at the end of the runs , the question of gi fragments surviving the initial rapid inward migration from their birth place into the inner disc regions is considered . more specifically , the focus is on how and when the fragments can contract and collapse more rapidly than they migrate in . this topic has direct observational connections . giant planets are observed to be much more frequent around metal - rich stars than around metal - poor ones @xcite . this correlation has been argued to provide a direct support to ca theory since metal - rich environments assemble massive cores much more readily ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . at the same time , gravitational instability planets ( and thus td planets also ) are expected to survive the inward migration phase most readily in metal - poor environments because their radiative cooling is the fastest at low dust opacities @xcite . this therefore appears to contradict the observed planet - metallicity correlation and is a very clear challenge to the td hypothesis . accretion of `` pebbles '' , @xmath12 cm sized grains , from the surrounding protoplanetary disc onto solid bodies , e.g. , large planetesimals up to massive cores , has been studied by a number of authors ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the process is interesting for many reasons , including explaining a potentially more rapid growth of massive cores at @xmath3 tens of au than by accretion of planetesimals @xcite and may contribute to formation of uranus and neptune @xcite . @xcite argued that while pre - collapse giant planets may be inefficient gas accretors ( see [ sec : broader ] further on this ) , they nevertheless should accrete pebbles for same reasons as solid cores in the ca picture do , except at higher rates because td fragments have much higher ( @xmath13 jupiter ) masses . adding a pebble injection term to the equations of an otherwise isolated td fragment evolution , a surprising result was obtained : the fragments actually accelerated their contraction and collapsed sooner , not later , despite the increase in the dust opacity associated with the higher metallicity of the fragments . @xcite explained the effect in terms of the extra weight and ( negative , of course ) potential energy that pebbles bring with them into the gas fragments . pre - collapse fragments turn out to be very sensitive to pebble accretion because they are nearly polytropes with low effective adiabatic indexes , @xmath14 . it was found that addition of @xmath9% by mass in grains is sufficient to drive a typical pre - collapse molecular h fragment into collapse . even more recently , @xcite presented a population synthesis like study that included the `` metal overload '' collapse due to pebble accretion on td planets for the first time , and found a strong positive rather than negative correlation of giant planet survival with disc metallicity . his model did not include the processes of grain growth , sedimentation and core formation inside the fragment . this paper expands the work of @xcite by adding core growth processes . further , two ways of enhancing the survived giant planet fraction are considered : pebble accretion and lower dust opacity due to grain growth ( this was not considered in * ? ? ? * at all ) . it is argued that low opacity models require unrealistically small dust opacities ( @xmath1 orders of magnitude smaller than the interstellar grain opacity at same metallicity ) to save the gas fragments from disruptions . even if this were plausible , low opacity models suffer from a host of other issues : ( i ) predict a negative correlation of planet frequency with host metallicity , ( ii ) can not produce solid cores more massive than @xmath15 , ( iii ) fail to explain why the observed giant planets are strongly over - abundant in metals . it is thus clear that low opacity models are a dead end as far as an all inclusive td model for planet formation is concerned . pebble accretion models , on the other hand , offer a more attractive alternative . they do not require unrealistically small opacities , naturally explain why giant planets can be much more metal rich than their host stars , and produce a positive giant planet - metallicity correlation . even the `` discarded '' gas fragments , e.g. , those destroyed by tides or thermal over - heating are important since the masses of solid cores surviving such disruptions can be in the terrestrial and above mass range . the paper is structured as following . a brief introduction into the physical motivation behind the td hypothesis , and a short overview of the literature on td , including potential applications to observations of the solar system and beyond , is given in 2 below . numerical algorithms for simulating the planet - disc interaction and the disc evolution are introduced and some example ( point mass planet ) migration tracks are calculated in [ sec : disc ] . [ sec : planet ] deals with the methods used to calculate the fragment evolution , including pebble accretion , grain growth , settling and core formation , whereas [ sec : together ] explains how the two modules are put in one code and what the initial conditions are . [ sec : opacity ] and [ sec : pebbles ] show examples of fragment evolution tracks for low opacity and pebble accretion models , respectively . in [ sec : discussion ] , the implications of the results obtained for td hypothesis is given . figure [ fig:1 ] shows the evolution of an isolated jupiter mass planet contracting from an initial state in which its central temperature is 100 k. the planet s metallicity is set to solar , and opacity is set to that of the interstellar gas / dust mix @xcite . the calculation is done with the part of the code described in section [ sec : planet ] , although grain growth and grain sedimentation within the planet are neglected . the top panel of figure [ fig:1 ] shows the evolution of planet s radius , @xmath16 , in units of jupiter s radius , @xmath17 , whereas the bottom panel shows the planet s central temperature . the planet s evolution is conveniently described in two stages @xcite . in the first `` pre - collapse '' stage , the planet s central temperature is below @xmath18 k , and hydrogen is in the molecular form . as temperature exceeds @xmath18 k , h@xmath19 molecules begin to dissociate . the process of breaking the molecules up requires 4.5 ev per hydrogen molecule , which is very large : it is equivalent to the thermal kinetic energy of two h atoms at temperature @xmath20 k. the dissociation process is therefore a huge energy sink , and leads to the planet loosing its hydrostatic equilibrium and collapsing dynamically . the new , post - collapse , equilibrium is such that h is atomic and partially ionised , central planet s temperature is well above @xmath21 k , and the planet radius is more than ten times smaller than in the pre - collapse state . in fig . [ fig:1 ] , the dynamical collapse corresponds to the nearly vertical part of the radius and temperature evolution curves . after the collapse , the planet arrives on the post - collapse stage , which is one and the same as the `` hottest start '' track for giant gas planets ( e.g. , * ? ? ? * ) . in fig . [ fig:1 ] , this post - collapse part of the evolution begins at the blue `` o '' symbol in the top panel of the figure . during this phase , the planet contracts from @xmath22 to @xmath23 relatively rapidly , e.g. , in some @xmath24 years , and then takes much longer to eventually contract to @xmath25 . the key point to observe from the top panel of figure [ fig:1 ] is just how extended the planet is at birth ( @xmath26 1 au , or about @xmath27 ) , and that it takes over a million years for the planet to contract to radii comparable to that of jupiter . the red solid horizontal lines show the hill s radius , @xmath28 , of the planet at three distances from a @xmath29 host star . when @xmath16 is larger may be sufficient for disruption , see [ sec : tides ] below ] when @xmath28 , tidal forces from the star exceed self - gravity of the planet , and it can be disrupted , as in the original @xcite scenario . figure [ fig:1 ] thus shows that the planet can be tidally disrupted at the planet - star separations @xmath30 au in the pre - collapse stage . in addition , the planet can be tidally disrupted in the post - collapse stage in the innermost `` hot '' region of the disc , @xmath31 au , provided the collapse happened very recently ( no more than @xmath32 years earlier ) . @xcite proposed that such hot disruptions may be the origin of `` hot '' sub - giant planets such as hot super - earths or hot neptunes . another point worth noting from figure [ fig:1 ] is that the astrophysical metal components of the protoplanet ( that is , all elements heavier than h and he ) can settle into the centre of it provided that condensation temperature of the components is higher than the central temperature of the planet . in this paper we limit our attention to three dominant grain species : water ice , organics called chon , and a mix of fe and silicates ( cf . [ sec : gr_growth ] below ) . the condensation temperatures for these species are marked approximately by the position of the respective text in the bottom panel of figure [ fig:1 ] . note that the planet spends little time in the cold configuration , since it cools relatively rapidly initially . from this one can expect water to be the least able to condense down in td planets , whereas fe and silicates be the most able to do so ( this agrees with results of * ? ? ? * that td cores are mainly composed of rocks ) . grain sedimentation may lead to core formation inside of the protoplanet ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? if the gaseous component of the protoplanet is then disrupted by the tides , the nearly `` naked '' core survives , since its density is much higher . such disruptions may be an alternative to ca origin for terrestrial - like planets @xcite . it is clear that the fate of a gas fragment formed in the outer disc by gi is sealed @xcite by the ratios of the various time scales : @xmath33 , the planet s contraction and collapse time scale ; @xmath34 , the planet s migration time scale ; and @xmath35 , the grain growth and sedimentation time scale . qualitatively , there are several possibilities : 1 . if the migration time is the shortest of the three , the fragment is disrupted with essentially nothing remaining of it ; 2 . if @xmath36 , then grains sediment down before the planet is disrupted . there is therefore a remnant , a solid core @xcite , possibly surrounded by a post - disruption atmosphere remaining bound to the core if the latter is massive enough ; 3 . if @xmath37 , then the planet collapses before it is disrupted , and survives as as a gas giant planet , provided it is not disrupted in the `` hot '' region or pushed all the way into the star . 4 . if grain sedimentation time is the shortest of the three , and the core mass is significant , @xmath38 a few to ten or more earth masses , then the grain component can affect the whole fragment by either triggering its collapse @xcite or destruction @xcite . clearly , it is essential to be able to calculate not only these three time scales but also the detailed planetary structure as well as the protoplanetary disc evolution . the planet formation / destruction road map outlined above is only a very rough guess as to what may actually happen in a realistic disc and planet setting . @xcite found formation and a rapid inward migration of massive self - gravitating gas clumps in their 2d simulations of protostellar discs . they argued that clump destruction episodes could be related to the fu ori outbursts of young stars . @xcite were first to point out that gas clump migration , core formation and then envelope disruption may result in formation of terrestrial - like planets . @xcite noted that dynamics of solids can be important enough to change the fragmentation properties of the gravitationally unstable discs , and to affect the gas clump compositions . these in general do not therefore have to be same as that of their parent discs @xcite . @xcite argued that the high - temperature environment inside the clumps ( before they are disrupted ) may provide a natural nursery for high temperature inclusions found in chondrules of the solar system . @xcite proposed that prograde rotation of the pre - collapse gas fragments may have been imprinted on the spin directions of earth and mars , and yield enough angular momentum for proto - earth to revive the fission hypothesis for the moon formation . @xcite observed recurrent episodes of clump formation , destruction and protostellar outflow injections in 3d simulations of self - gravitating discs . @xcite made detailed studies of planet migration in gravitationally unstable discs , finding migration times of order @xmath9 thousand years . @xcite simulated in 3d a gravitationally unstable gas disc together with the grains that were allowed to grow , and found that grains grew most rapidly inside the clumps and indeed sedimented into the dense cores . @xcite made first 3d simulations of collapse of the rotating molecular clumps , noting the importance of the angular momentum and formation of the post collapse circum - planetary discs . it is important to note that planetesimals are not a pre - requisite for td model , unlike for ca . in the context of td , minor solids , such as asteroids , are remnants of larger solid bodies that were fragmented in collisions . thus , @xcite argued that kuiper and asteroid belts in the solar system could be made in disruptions of molecular gas fragments in which some of the solids sedimented into the centre and got locked into a number of ceres mass bodies . they found that after the gas clump disruption , the distribution of solid bodies naturally forms rings with sharp edges , as observed in the solar system , and that this model does not have the `` mass budget deficit '' , resolving the two well known problem of the classical models of the kuiper belt ( e.g. , * ? ? ? * ) . @xcite argued that gas accretion on relatively low mass gas td fragments ( @xmath39 a few @xmath40 ) is inefficient due to preheating effects from the planet on surrounding gas . these authors found that there is a dichotomy in the fate of the clumps . low mass clumps were found to migrate in rapidly at more or less constant mass , and eventually were destroyed by the host star s tides . clumps more massive than @xmath41 were found to run away in the `` opposite '' sense : accreting gas rapidly , they become as massive as @xmath42 , open deep gaps in their protoplanetary discs , and stall at about the initial star - clump separation . this divergent evolution may explain the divergent results on the fate of the gas clumps in the literature ( giant planets or brown dwarfs see * ? ? ? * for references ) , since small changes to the disc parameters and the treatment of gas thermodynamics may shift the clumps from one of the other regime . @xcite presented 3d radiative hydrodynamics simulations of massive self - gravitating discs . their approach is clearly preferable to all of the previous 2d studies ( e.g. , * ? ? ? * ) and also the earlier 3d studies which used radiative cooling prescriptions ( e.g. , * ? ? ? the rhd simulations could in principle directly address many of the issues needed for a quantitative td population synthesis models , such as the initial clump mass ( * ? ? ? * ; * ? ? ? * e.g. , ) , and the migration and clump accretion history . @xcite find that most of their clumps are as massive as @xmath43 at birth and grow rapidly in mass while migrating in . what is not clear is how general these results are with respect to changes in initial conditions , and how they depend on the dust opacity model assumed by @xcite . here , and everywhere in the paper , only the simplest version of events is considered , in which the planets start out as self - gravitating gas fragments and do _ not accrete _ gas or planetesimals from the disc . both gas accretion and especially planetesimals are virtues of the core accretion model ( e.g. , * ? ? ? it appears to be most prudent to first consider the simplest `` orthodox '' td hypothesis . more complicated versions of td may be considered in the future . this section presents numerical methods with which planet migration and disc evolution are modelled in the paper . as far as the planet - disc interaction is concerned , the planet is treated as a point mass . the evolution of the internal planet s variables is considered in [ sec : planet ] . using a 1d viscous disc evolution approach , @xcite studied planet migration and planet mass loss for very massive ( @xmath44 ) gas planets in the `` hot '' disc region , e.g. , at @xmath5 au . such massive gas fragments are always in the type ii migration regime , when they open a deep gap in the disc @xcite . @xcite were interested not only in the fate of the planets but also in the potential connection between giant planet disruptions and the fu ori outbursts of young stars @xcite . therefore , the planet - disc mass exchange was followed in detail and on short time scales ( much shorter than a year ) to compare them with the observed light curves and spectra of some observed fu ori sources ( e.g. , * ? ? ? * ; * ? ? ? the approach here follows that of @xcite with a few changes . since the fate of less massive gas planets and also that of even less massive remnants ( solid cores , if the planets are disrupted ) is of interest to us here , type i migration regime ( no gap in the disc ) is also included . accretion luminosities of young stars , although calculated here automatically by the viscous disc evolution code , are of no direct interest . therefore , the treatment of the planet s mass loss is simplified from that of @xcite : it is assumed that the planet is disrupted instantaneously if the disruption criteria are satisfied , and the mass deposition back in the disc is neglected for simplicity . the protoplanetary disc is described by a viscous azimuthally symmetric one dimensional time - dependent model that encapsulates the standard accretion disc solution of @xcite with addition of the tidal torque of the planet on the disc . the disc surface density , @xmath45 , is evolved in this approach according to @xmath46 -\frac{1}{r}\frac{\partial}{\partial r}\left(2\omega r^2\lambda\sigma\right ) \label{eq : diffplanet}\ ] ] where @xmath47 is the keplerian angular frequency at radius @xmath48 , viscosity @xmath49 , where @xmath50 and @xmath51 are the midplane sound speed and the disc height scale , @xmath52 , and @xmath53 is the specific tidal torque from the planet . the latter is a weighted sum of the type i and the type ii contributions , @xmath54 where @xmath55 is a switch controlling whether the planet migrates in type ii ( @xmath56 ) or type i ( @xmath57 ) regimes . two - dimensional simulations @xcite show that a deep gap in the disc is opened when parameter @xmath58 where @xmath51 is the disc vertical scaleheight at planet s location , @xmath59 , and @xmath60 is the shakura - sunyaev viscosity parameter . we therefore set @xmath61\right\}\;. \label{kill2}\ ] ] the normalised specific torque for type ii migration is given by the widely used expression , @xmath62 @xcite , where @xmath63 . we smooth the torque term for @xmath64 , where it would have a singularity ( see equation [ eq : torque ] ) . we use the same smoothing prescription as @xcite and @xcite , i.e. for @xmath65 $ ] , where @xmath51 is the disc thickness and @xmath66 is the size of the hill sphere ( roche lobe ) of the planet . as far as we are aware , there is no corresponding expression for @xmath67 , perhaps partly because one usually studies type i migration in the limit where the planet s mass is small compared to the disc mass , so that the back reaction of the planet on the disc can be neglected . our planets can be massive and yet still be in type i regime , especially in the outer self - gravitating disc @xcite , and also in the inner disc if viscosity parameter @xmath68 is `` large '' ( e.g. , greater than @xmath3 0.01 - 0.03 ) . furthermore , gas giant planets studied here are bright and can easily dominate the disc luminosity around their location , e.g , within the region @xmath69 $ ] ( see * ? ? ? * ) . this may affect the thermodynamic disc structure near the location of the planet , yet none of the migration disc studies in the literature currently take this effect into account . faced with this uncertainty , we chose to use the simplest approach that appears reasonable and is to be improved in the future as details of type i migration are understood further . the total torque between the disc and the planet , @xmath70 , in a self - gravitating outer disc @xcite is similar to the standard isothermal disc result @xcite , which yields type i planet migration time scale ( @xmath71 ) @xmath72 where @xmath73 is approximately the local disc mass , @xmath74 is the local keplerian angular velocity , and the @xmath75 factor is introduced to account for a possibility that @xmath76 for our planets , which is not the typical case for type i migrating planets ( e.g. , see 2.1 in * ? ? ? * ) . while the form of the factor @xmath77 is debatable , we note that the main conclusions of this paper do not depend on this factor at all since the migrating planets are always much less massive than the disc while migrating most of their way in . we assume that only the material close to the planet s location exchanges angular momentum with it , and that the strength of the interaction falls off away from the planet as @xmath78\ ; , \label{lambda1}\ ] ] where @xmath79 , and @xmath80 . @xmath81 is found by requiring the total type i torque from the planet on the disc to equal that ( with the minus sign ) from the disc on the planet , that is , @xmath82 . in equation [ time1 ] , @xmath83 is the local keplerian angular velocity . the migration rate of the planet is calculated from the conservation of the total ( disc plus planet ) angular momentum as in @xcite . also as in @xcite , see their 3.2 , the vertical structure of the disc is calculated as in the standard @xcite model , taking into account the irradiation from the star , which can dominate the disc heating at larger radii . the disc thermal balance equation takes into account the finite thermal disc time scale , @xmath84 , which determines how quickly the disc achieves thermal equilibrium if accretion rate varies suddenly . this is important in the inner disc , if the disc accretion rate is high and the thermal disc instability is triggered @xcite . the disc opacity is from @xcite , multiplied by the ratio of the disc metallicity to the solar metallicity , @xmath85 . finally , it should be noted that in this paper only the outcome of the planet rapid inward migration phase from the outer into the inner disc is studied , and a longer time scale evolution of the survived fragments is not addressed . this would require a better treatment of type i migration , which currently remains controversial and model dependent especially for the more massive planets @xcite , and a disc photo - evaporation term to be added to equation [ eq : diffplanet ] to model late disc evolution . while this is relatively straight forward as disc dispersal by photo - evaporation is believed to be reasonably well understood @xcite , we wish to avoid introducing further parameters and complications to our model at this initial stage . figure [ fig : two_discs ] shows the disc evolution in two example calculations for a planet of a fixed @xmath86 mass migrating within a protoplanetary disc , starting from @xmath87 au . we are focused here on the migration part of the model , so we assume that the planet has already contracted to very high density and thus is not a subject to a tidal disruption . the left hand panels of the figure shows the case of a relatively low initial mass gas disc , @xmath88 , and a low viscosity parameter , @xmath89 . the initial conditions for the disc are as described in [ sec : ic ] , except the disc inner boundary is set to @xmath90 au in this section . the right hand panels of figure [ fig : two_discs ] contrast the low mass disc case to a much more massive disc @xmath91 and a viscosity parameter higher by a factor of 8 , @xmath92 . since the steady state accretion rate in the disc scales as @xmath93 @xcite , the right hand panels disc produces approximately 40 times larger accretion rate onto the star than the left hand s disc . the sub - panels of the figure show the disc surface density @xmath94 , the central ( midplane ) disc temperature , @xmath95 , and the disc geometrical aspect ratio , @xmath96 , at three different times as indicated in the legend . the abrupt changes in @xmath94 and @xmath95 of the disc in locations far away from the planet , e.g. , at @xmath97 au , visible in the black solid curves in the right hand panels are due to rapid dust and/or gas opacity changes at these locations . the first disc is much cooler , so that @xmath96 is smaller . due to this , and since the viscosity parameter is low , the disc already develops a deep gap when the planet migrates to @xmath98 au . the gap remains deep but partially opened until the planet migrates to @xmath99 au , at which point the inner disc viscous time is shorter than the planet s migration time , so that the inner disc drains onto the star and there is a complete gap between the star and the planet . the planet continues to be `` pushed '' inward by the angular momentum exchange with the outer disc , and eventually perishes by going through the inner boundary of our computational domain for these tests , @xmath100 au . in contrast , the planet is unable to open a gap of any denomination in the second test because the disc is much hotter , so @xmath96 is higher , and because the disc viscosity is high . this planet is always embedded in its parent disc and migrates in type i regime only . the high disc viscosity in the second disc implies that this disc evolves much faster . this can be surmised from the @xmath45 curves in figure [ fig : two_discs ] : there is a notable decrease in the disc surface density curves st late times even at large @xmath48 in the middle right hand panel , whereas the corresponding curves on the left show that the disc evolves only inside @xmath101 au region for the low viscosity test . finally for this section , the time evolution of the disc environment around the planet in the two calculations presented in figure [ fig : two_discs ] is plotted in figure [ fig : two_discs2 ] . panels ( a ) through ( c ) show the planet - star separation , @xmath59 , the irradiation temperature incident on the planet from the surrounding disc , @xmath102 , and the disc surface density at @xmath103 , respectively . the initial evolution of these quantities , from @xmath104 au to @xmath98 au is similar for the two planets , modulo the fact that the disc is five times less massive , and the planet migrates slower , for the blue dotted curve than for the black solid one . at smaller distances to the star , there is however a profound difference . while the hot disc planet ( solid curve ) continues to migrate in the type i regime , initially quickly and then slower and slower as it nears the inner disc , the other planet opens a deep gap and starts to migrate in type ii regime , as we already saw in figure [ fig : two_discs ] . contrary to the hot disc case , the planet in the low mass disc test accelerates as it enters the inner disc region . therefore , somewhat contrary to intuition , the less massive and cooler disc pushes the planet through the inner boundary much faster ( the hotter disc case planet arrives at the inner boundary not shown in the figure about a million years later ) . focusing now on the irradiation temperature evolution , two significant factors should be noted . first of all , the planet in the hotter disc is bound to find itself in hotter environment , generally . however , a secondary effect amplifies this statement manyfold : since the planet in the hot disc does not open a gap , it finds itself sampling the disc _ midplane , e.g. , central _ temperature @xmath95 . the planet in the colder disc is in the gap , and that region is much cooler : the appropriate irradiation temperature is comparable to the disc effective temperature . there is a factor of at least a few difference in these two temperatures _ at the same radius @xmath48 _ because the disc optical depth is usually considerable and hence the midplane temperature is higher than the effective one ( @xmath105 in the shakura and sunyaev 1973 disc theory , where @xmath106 is the disc optical depth ) . in terms of irradiating flux incident on the planet , there is thus a difference by many orders of magnitude . the `` thermal bath '' effect that may puff up and eventually unbind a planet ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * and also section [ sec : bath ] below ) is therefore much stronger in the hot disc case than it is in the cold disc one . these two tests illustrate why a detailed time - dependent calculation of disc properties , and how these properties change due to the disc interaction with the planet , is a necessity for an accurate statement on what happens with a gas fragment migrating inward inside those discs . @xcite , @xcite , employed a 1d spherically symmetric lagrangian radiative hydrodynamics ( rhd ) algorithm , with grains modelled as a second fluid , to follow evolution of isolated self - gravitating gas fragments . such a method is unfortunately too costly numerically , and lends itself to studies of only a few test cases , while we are interested in being able to investigate , eventually , thousands of models . a more expedient `` follow the adiabats '' method ( e.g. , * ? ? ? * ; * ? ? ? * ) is used in the present paper . namely , since the energy transfer inside the planet is strongly dominated by convection at high opacities / metallicities @xcite , we assume that the fragment is isentropic and is in hydrostatic balance . the initial planet is assumed to have a constant grain to total mass ratio , equal to the initial planet metallicity , @xmath107 . the planet is divided into @xmath108 concentric equal mass gas shells with mass @xmath109 , where @xmath110 is the initial planet s mass . given the central fragment s temperature , @xmath95 , at some time @xmath111 , the structure of the isentropic planet can be found by iterations on the central gas density , @xmath112 , subject to the condition that pressure vanishes at the outer boundary . this procedure yields planetary radius , @xmath113 , the total energy of the planet , @xmath114 , and all the internal properties of the planet , e.g. , the radial coordinates for all of the radial mass zones within the planet , @xmath115 , as a function of mass enclosed within the shell , @xmath116 . in addition , the radiative luminosity , @xmath117 , is calculated . @xmath117 is small deep inside the planet , increases outward and reaches a maximum value close to the planet s surface . we take this maximum value to be the radiative luminosity of the planet , @xmath118 . this is done with the understanding that the very outer regions of the planet will in reality be radiative rather than convective ; however the fact that the pressure scale height is usually very small compared with the planet s radius means that the error in the value of planet s radius , @xmath16 , is also small . the inner radius of the isentropic planet is given by @xmath119 , the solid core radius , as described in [ sec : core ] . we do not model the internal structure of the core . the core releases @xmath120 , the core accretion luminosity , into the rest of the planet . the total energy of the planet is evolved in time according to @xmath121 where @xmath122 is the radiative luminosity is the luminosity that the planet would have in isolation , @xmath118 , _ minus _ that incident on the planet from the surroundings : @xmath123\;. \label{lrad1}\ ] ] here @xmath124 is the irradiating luminosity . @xmath102 is equal to the surrounding disc temperature if the planet is embedded in the disc , or the local disc effective temperature if the planet opens a gap , as explained in [ sec:2discs ] . the last term on the right hand side of equation [ etot1 ] is the change in the energy of the planet if grains accrete on it at the rate @xmath125 . note that since grains accrete on the planet slowly , that is at velocities not exceeding a few m s@xmath126 or else collisions would destroy the grains , the grains kinetic energy input into the planet can be safely neglected ( see [ sec : gr_dyn ] and [ sec : gr_growth ] , and also * ? ? ? equation [ lrad1 ] does not permit negative radiative luminosities for the planet for the following reasons . the structure of strongly irradiated planets , that is , when @xmath127 , where @xmath128 is the temperature of the planet in isolation , is not isentropic since the outer layers become radiative ( e.g. , * ? ? ? experimenting with strongly irradiated planets using the stellar evolution code mesa @xcite , it was found that the outer layers of the planet become approximately isothermal , @xmath129 ( see also the flat regions in @xmath130 vs pressure curves in figure 1 in * ? ? ? the radiative luminosity of the planet becomes much smaller than it is in isolation . it does not become negative , however , except for a short time required for the outer layer thermal balance re - adjustment . furthermore , as @xmath102 is increased , the depth of the outer radiative layer increases . importantly , when the layer contains a good fraction of the planet s total mass and @xmath102 is comparable to the planet s virial temperature , the outer layers start to expand without bounds until mesa stalls by having to use tiny timesteps . physically , the planet becomes unbound and is destroyed very quickly . this is the `` thermal bath '' effect by which gaseous protoplanets are known to be destroyed when the surrounding environment is too hot @xcite . the criterion that we use to capture the thermal bath destruction of the planet is a qualitative one : we require the irradiation temperature to exceed @xmath131 where the virial temperature of the planet is defined as @xmath132 , where @xmath133 is the mean molecular weight . gravitational potential of post - collapse planets is high enough to allow gas accretion onto their surface without a need for radiation since h@xmath19 dissociation energy can be such a sink . hence gas accretion onto the planets is much more likely in the post - collapse case than it is in the pre - collapse case ( see also * ? ? ? grains , including pebbles , would accrete onto the post - collapse planet together with the gas , since such accretion would inevitably generate very hot shock fronts where pebbles would be easily vaporised and thus well mixed and coupled with the gas . however , gas accretion onto planets is not included in this paper ( see [ sec : big_idea ] ) in an attempt to reduce the parameter space of the models . therefore our post - collapse planets do not accrete either gas or grains and hence evolve at a constant mass unless disrupted . to model the post - collapse contraction of the planets , publically available code mesa @xcite is employed to create tables of radii , total energy and luminosity for a given mass planet for _ non - irradiated _ coreless giant planets at gas metallicity of @xmath134 . this metallicity is typical of the metallicity of the planets born in our discs due to metal loading @xcite . the luminosity of the core is neglected in the post - collapse phase because the core s heat content is always much smaller than that of the post - collapse planets ( unlike the pre - collapse planet case ) . having @xmath113 and @xmath135 for isolated planets would be completely sufficient for our purposes of deducing whether the planet contracts more rapidly or gets tidally disrupted in the `` hot '' region ( see the @xmath136 au line in the top panel of figure [ fig:1 ] , and also * ? ? ? * ) , were it not for irradiation of the planet by the surrounding disc or the central star . due to that irradiation , the planet contracts slower than an isolated planet would do . to take this into account , we use an approach similar to that of the pre - collapse phase but employing the mesa tables created and with core luminosity and pebble accretion turned off , as explained above . in this approach the energy equation becomes @xmath137 where @xmath122 is calculated as in equation [ lrad1 ] . now , having found an updated value of the planet s total potential energy , interpolations in the tables are employed to find the corresponding age of the planet and the new values for @xmath113 and @xmath135 . the contraction step can now be repeated , and hence @xmath113 for a migrated irradiated planet is found . the post - collapse planets can in principle be disrupted due to either tides ( [ sec : tides ] ) or the thermal bath effects ( [ sec : bath ] ) , but in the present paper none of our planets go through such disruptions . grain sedimentation is of course important only in the pre - collapse planets since post collapse planets are too hot to permit existence of grains . in this paper , grains are treated as a perturbation to gas dynamics . having found @xmath115 and @xmath138 as described in [ sec : pre - collapse ] , we find the gas velocity for every mass shell in the planet , @xmath139 if grains are sufficiently small , e.g. , @xmath140 cm in practice , where @xmath141 is a spherical grain s radius , then they follow the gas motion closely due to a strong aerodynamic coupling . however , in general grain dynamics can be different from that of the gas , therefore grain sedimentation is modelled as in 3.4 of @xcite . grains are assumed to reach their terminal sedimentation velocity ( with respect to gas ) quickly , @xmath142 where @xmath143 is the grain velocity in the frame of the centre of the planet ( which is motionless in an isolated planet case but is moving with the planet if the latter is embedded in the disc ) , @xmath48 is the radius of the given mass shell inside the planet , @xmath144 is material density of the grain , @xmath141 is the grain size , @xmath50 is the gas sound speed , and @xmath145 is the mean free path of @xmath146 molecules . equation [ vsed1 ] joins smoothly the epstein and the stokes drag regimes . grain sedimentation is opposed by turbulent grain mixing ( e.g. , * ? ? ? * ) and convection ( e.g. , * ? ? ? these two processes can be combined to give a diffusion coefficient , @xmath147 , and be modelled as diffusion of grain concentration , @xmath148 @xcite . the equation describing this process is @xmath149\;. \label{diff_eqn}\ ] ] this defines mean grain diffusion velocity , @xmath150 , @xmath151\;. \label{vdiff1}\ ] ] since the grain concentration , @xmath148 , is highest in the planet s centre due to grain sedimentation , the mean diffusion velocity is positive , confirming that this process opposes grain settling . the diffusion velocity is combined with the sedimentation velocity to give the total grain velocity in the planet s centre of mass frame , @xmath152 this defines the grain mass flux in or out of the lagrangian gas mass shells in the planet , which then defines the rate of change of the grain density @xmath153 at every shell . updating @xmath153 through these mass fluxes is equivalent to solving the grain mass continuity equation , @xmath154 = - \frac{d}{d r}\left[r^2 \rho_g u_a \right]\;. \label{drhodt_full}\ ] ] the diffusion coefficient @xmath147 takes into account two effects : turbulence and convection . for the former , a turbulence parameter @xmath155 is introduced , @xmath156 where @xmath16 is planet s radius and @xmath50 is the local sound speed . convective contribution to @xmath147 is @xcite given by @xmath157 where @xmath158 is the local gas scaleheight , and @xmath159 is the speed with which convective eddies rise . on the basis of the mixing length theory of convection @xcite , @xmath160^{1/3}\ ; , \label{dconv2}\ ] ] where @xmath161 is the local gravitational acceleration and @xmath162 is the local convective flux . the two contributions are added together linearly , so @xmath163 it should be emphasised that turbulence and convection are very important in opposing grain sedimentation and therefore core s growth @xcite . if @xmath150 is sufficiently large then grains can not sediment and in this case @xmath164 const everywhere inside the planet to a good approximation , as we shall see later on . the procedure just described is in place for all of the three grain species , independently of one another . this is necessary because the species have different sedimentation velocities , grain size @xmath141 , and are distributed differently inside the planet . grains are expected to have a distribution of sizes , from interstellar values , @xmath165 m @xcite , to perhaps as large as a few cm by radius due to grain growth . a collision - fragmentation model for dust grains is necessary to follow the grain size distribution ( e.g. , * ? ? ? * ; * ? ? ? the grain distribution function is also a function of position inside the planet . such a complicated treatment is well beyond our numerical resources , given that grain growth is only one of the important processes that is modelled here . in addition , physical uncertainties in grain growth and fragmentation processes would require introducing new poorly constrained parameters , greatly increasing the parameter space for the population synthesis models . following @xcite , grain growth is captured in a simpler framework . the grain distribution function over grain sizes is approximated by just two components : the `` small '' and the `` large '' grains . the small grains are always present due to fragmentation of larger grains in collisions , as suggested by observations and modelling of grain growth in proto - planetary discs @xcite . we assume that the small grains are tightly bound to gas , do not sediment , and are , of course , the dominant source of dust opacity . the population of large grains is represented by a single size @xmath141 and is allowed to move relatively to the gas as described above . the size @xmath141 should be thought as a typical radius of the large grain population in a given gas mass shell . @xmath141 varies with time due to grain growth , fragmentation , vaporisation , and motion of the grains from one radial cell to another . these processes are treated almost exactly as in @xcite , and hence we only give a brief summary of what is done and point out the differences . large grains grow due to brownian motion of small grains , and by sticking collisions as they sediment ( cf . * ? ? ? * ; * ? ? ? * ) : @xmath166\ ; , \label{da_grow}\ ] ] where @xmath167 is brownian velocity here set to 10 cm s@xmath126 ( see 2.4 in * ? ? ? * ) , which is typically much smaller than the sedimentation velocity . the `` stick - or - break function '' @xmath168 , where @xmath169 , is given by @xmath170 this simplifies the treatment of grain collisions used in 2.6 of @xcite , but achieves the same physical goal : grain - grain collisions are perfectly sticking for @xmath171 , but become fragmenting when sedimentation velocity exceeds @xmath172 ( note that @xmath173 ) . the possible grain cross - fragmentation , e.g. , fe grains fragmenting silicate grains @xcite , is neglected in this paper for simplicity . the maximum velocity , @xmath172 , under which grain sticking is still possible , depends on properties of the material , and is best inferred from experiment . there is however a large uncertainty here . @xmath172 was measured to be a few m / s for both silicate @xcite and icy materials @xcite , but @xcite report first laboratory experiments on decimetre - sized dust agglomerates that are bound by surface forces only ( rather than by chemical bonds important for materials such as gypsum ) , and find @xmath172 as small as @xmath174 m / s . pure metallic fe grains may have @xmath172 as large as @xmath175 m / s , although interstellar amorphous fe grains are probably considerably weaker ( see discussion in 2.6.1 in * ? ? ? * ) . by performing numerical experiments , it was found that @xmath172 is an important parameter of the model as it controls the maximum speed with which grains can sediment , and hence the core can be assembled . it is thus left as a free parameter of the models . grain vaporisation rate , @xmath176 , is calculated exactly as in 2.5 of @xcite . grains vaporise rapidly when the surrounding gas temperature exceeds the vaporisation temperature for given species ; the vaporisation temperature is a function of gas pressure . finally , due to diffusion and sedimentation grains may enter or leave a given mass shell , which leads to a change in the grain size if @xmath141 is different in different regions of the planet . utilising the grain mass conservation equation [ drhodt_full ] , we write @xmath177\;. \label{ag_advect}\ ] ] this is an `` advective '' change in the grain size . combining all these processes , we write the full equation for grain growth as @xmath178 grain abundances are as in 2.8 of @xcite , except silicates and fe are combined in one refractory species . the solar abundance of metals is defined as @xmath85 @xcite , which is divided amongst the grain species such that the fractional abundances , @xmath179 of water , chon and silicates are 0.5 , 0.25 , 0.25 times metallicity @xmath107 , respectively . for material density of grains we use @xmath180 g @xmath181 , 1.5 and 1.0 , for silicates , chon and water ice , respectively . the processes of grain growth and sedimentation impose their own constraints on the maximum timestep @xmath182 , which may be more stringent than those from [ sec : rad_con ] . the timestep is required to be small enough that none of the grain properties change by more than 10% between any two successful models of planetary structure . the inner boundary of our computational domain is at the core s radius , @xmath183 , where @xmath184 g/@xmath185 . this radius is many orders of magnitude smaller than the outer radius of the planet , @xmath16 . the structure of the planet at such small radii is unresolved in our simulations . the boundary conditions applied at @xmath186 is @xmath187 and @xmath188 . as in @xcite , the core accretion rate depends on velocity with which grains sediment , @xmath189 , at the first radial zone above the inner boundary , @xmath190 here @xmath191 is the mass of the dust in the first gas zone , @xmath192 is the outer radius of the first gas zone , and @xmath193 is the velocity with which grains arrive into the first zone from the second gas zone . the function @xmath194 is introduced to quench growth of the core in the cases where @xmath195 cm , when grains should be coupled to gas tightly . while the code reproduces the expectation that @xmath196 is very small for small @xmath141 ( since @xmath197 is very small ) , some spurious core growth at small levels does occur . to turn this unphysical growth off , the functional form of @xmath194 is chosen to be @xmath198\;.\ ] ] this treatment is applied to all of the three grain species independently , to determine the accretion rate of these onto the core . these accretion rates are then added up to give the total core accretion rate . the core s mass is set to a negligibly small value in the beginning of the simulations . energy release by the core may have important effects onto the rest of the planet . the simplest approach would be to follow the ca prescription according to which the core luminosity is @xmath199 where @xmath200 is the instantaneous accretion rate of the solids onto the core . however , this approach is likely to strongly over - estimate @xmath201 during its assembly _ in td framework _ since this assumes that radiation diffuses out of the core rapidly . this is probably wrong as opacity of solid cores is very significant . the modelling of cooling of initially hot rocky planets above earth mass is still very uncertain due to insufficient data on opacities , convection , conduction and other material properties at the appropriate temperature and pressure ranges ( e.g. , * ? ? ? the atmospheres of such planets ( the gas layers immediately adjacent to the core ) could hinder rapid cooling . for example , @xcite derive cooling time scales for the atmospheres of earth - like planets to be as long as @xmath24 to @xmath202 years , whereas core assembly times in our models may be as short as @xmath203 years . since a self - consistent treatment of the internal structure of the cores is well beyond the scope of our work , a parameterisation of core accretion luminosity is made following @xcite , in which the core emits its accretional energy on a finite time scale , @xmath204 , where @xmath205 is the kelvin - helmholtz contraction time of the solid core . specifically , @xmath206 where @xmath207 is the residual potential energy of the core , which is integrated in time according to @xmath208 @xmath209 and @xmath119 are the running ( current ) core s mass and radius . evidently , if @xmath210 , then @xmath207 is exactly equal to the potential energy of the core s assembly at mass @xmath211 . the core s luminosity is zero in this case . in more general case , if @xmath212 , then @xmath213 , but the total energy emitted by the core in the limit @xmath214 is the same ( @xmath215 , where the final value of the core s mass and radius are used ) . in the opposite limit , when @xmath216 , @xmath217 . in this paper , @xmath218 years for all of the runs presented . preliminary results indicate that dependence of the results on @xmath204 is moderately weak unless @xmath204 is longer than a few million years , in which case most of the core s energy is saved `` for later '' , that is accumulated for release after the protoplanetary disc is removed . we assume that the gas envelope of the gaseous giant planet is _ completely _ tidally disrupted when the planet fills a large fraction of its roche lobe radius , e.g. , when @xmath219 the factor in front of @xmath220 in this expression depends on the rotation state of the planet . if the planet is in a synchronous rotation with the star , as is usually the case for stellar binary systems ( e.g. , * ? ? ? * ) , then this factor is nearly unity . for hydrodynamical simulations of planets embedded in discs , @xcite suggest the factor is @xmath221 whereas @xcite find it closer to 0.5 . these simulations however sample the earliest phase in the evolution of the planets , some @xmath222 to perhaps @xmath21 years into their existence . during this early phase the planets are most rapidly rotating due to a significant angular momentum at formation ( e.g. , * ? ? ? * ) , hence they may be quite aspherical and hence easier to disrupt by tides . the factor of @xmath223 in equation [ disruption1 ] is probably more relevant for the typically older planets that are studied here . our main results do not depend on this factor sensitively . @xcite show that massive bodies accrete pebbles in the `` hills regime '' , that is when all of the pebbles streaming past the planet within its gravitational reach the hills s radius @xmath224 are accreted . therefore , @xmath225 where @xmath226 , @xmath227 , and @xmath228 is the grain surface density at radius @xmath103 . the efficiency of accretion of large grains by the planet is a function of their size ( e.g. , * ? ? ? * ; * ? ? ? * ) , and only a fraction @xmath229 is in the pebble regime . @xmath230 is a free parameter of the model . note that @xmath231 must be found self - consistently by the disc - planet interaction module ( [ sec : disc ] ) ; it can vary by orders of magnitude at the same distance from the star depending on the disc time evolution , and especially on whether a gap around the planet s location is opened or not . having found the pebble mass accretion rate , pebbles are added to the existing population of grains in the outermost few radial zones of the planet and are then evolved in the manner described earlier . the disc migration and planet evolution modules are combined together into a time - dependent code . this is relatively straight forward . the two modules are called by the main program that stores the values of all important variables at a given time step and passes the information between the two modules . in particular , the disc migration module needs to be provided with the current planetary mass , @xmath110 , which changes relatively slowly due to pebble accretion but can change abruptly due to a tidal disruption of the fragment . the planet evolution module requires the following inputs from the disc migration routine : the pebble accretion rate , @xmath232 , and the irradiation temperature , @xmath233 . the main routine determines a time step , @xmath182 , such that none of the planet or the disc variables change by more than a few percent during that step . if @xmath182 exceeds @xmath234 years ( a free parameter ) , it is set to @xmath235 . the migration and the planet evolution modules are then executed independently from one another for the duration @xmath182 . the exchange variables defined above are held fixed during this step . after the execution they return the information back to the main routine , which first makes a series of checks , such as a comparison of the hill s radius with the planet s radius to assess the fragment s integrity against tidal disruption , and then repeats the procedure by providing the modules with updated input parameters . the disc is initialised with the following surface density profile : @xmath236\ ; , \label{sigma_init}\ ] ] where @xmath237 is the disc length - scale , set to @xmath238 au . the constant @xmath239 is calculated so that the disc contains a given initial mass @xmath240 . the initial surface density profile given by equation [ sigma_init ] is frequently used in studies of protostellar disc evolution ( e.g. , * ? ? ? * ; * ? ? ? the inner boundary of the disc is a free parameter , set to @xmath241 au for tests in [ sec : opacity ] . this is done for simplicity , since a detailed study of planet evolution in the hot region @xmath242 au is beyond the scope of this paper . @xmath243 is the outer radius of the disc radial grid , and is set to a value significantly larger than @xmath237 . the planet is inserted at an initial location , @xmath244 au . the disc structure around the planet initially does not take the planet s presence into account , but this is not important in practice . the planet s mass is much smaller than the surrounding disc mass at @xmath245 ( unless an extremely low disc mass @xmath246 is considered ) , and the planet is usually unable to open a gap in the disc anyway until it migrates inward substantially , as found in 2d simulations @xcite . the planet s initial structure is a polytropic gas sphere of a given total mass and central temperature , @xmath95 , which is a free parameter of the model . the grains are initially mixed with gas uniformly , i.e. , at time @xmath247 , @xmath248 , where @xmath249 is the total grain mass density for all of the grain species , @xmath250 is the gas mass density at radius @xmath48 inside the gas clump , @xmath251 is the initial fragment s metallicity . the volumetric density of a grain species @xmath252 is given by @xmath253 . for simplicity , grains have a uniform initial size @xmath254 m everywhere in the cloud . for a giant planet to survive the formation and disc migration phase , it must stop migrating inward at some point . the may happen on the very inner edge of the disc if the disc inner boundary is set by magnetospheric torques sufficiently far from the stellar surface . alternatively , the disc may dissipate away due to photo - evaporation ( see * ? ? ? * for a review ) before the planet migrates all the way to the inner boundary . clearly , the balance between the rate of the disc photo - evaporation and the planet s inward migration determines whether the planet stalls or migrates to the disc inner edge . our type i migration model ( [ sec : disc ] ) is too simplistic for the inner non self - gravitating part of the disc , and therefore we do not attempt to end the simulations `` properly '' , that is by photo - evaporating the disc ( e.g. , * ? ? ? the runs presented below are performed for just long enough for the planet to migrate from the outer disc into the inner disc , and to thus answer the more limited question of whether the giant planet - to - be survives this migration or not . while a more self - consistent ending of the simulations is needed to ascertain the eventual survival of the planets that collapsed before they migrated into the inner disc , the present study remains useful at limiting the spectrum of models that are able to overcome the tidal disruption barrier and deliver giant planets into the inner disc . this section begins presentation of different tests of the model . we begin with the case of no pebble accretion , @xmath255 . the input parameters and main outcomes of selected simulations discussed in the sections to follow are summarised in table 1 . the runs are labelled as `` m1lowopx '' etc , where `` x '' is an integer referring to a parameter varied within a given series of runs ( see below ) , and the preceeding letters refer to one of the three series of runs . the number after `` m '' is the initial mass of the fragment in jupiter masses . all of the simulations are performed for the same disc parameters in this paper , the disc viscosity parameter @xmath256 and the initial disc mass @xmath257 . the conclusions of the paper are qualitatively independent of the disc parameters . the main outcome of a simulation fragment disruption or collapse is best inferred from the row labelled @xmath258 , which shows the planet s mass at the end of the simulation in jupiter masses . for runs in which the planet is disrupted , @xmath258 is the mass of the solid core assembled inside the fragment before it was disrupted , and it is of course much less massive than the starting mass of the planet . as we have seen in [ sec : big_idea ] , the main challenge to forming giant planets via td is a long contraction time of giant planets , @xmath13 million years for @xmath259 , which implies that planets migrate inward more rapidly than they contract and hence they are disrupted ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? the migration time is found to be @xmath260 years , depending on disc parameters and the planet s mass ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? in order to get more giant planets to survive the fast inward migration , the radiative contraction time of the planets is to be shortened somehow . one way of achieving this is to invoke grain growth and therefore dust opacity reduction ( e.g. , * ? ? ? * ; * ? ? ? * ) . to test this possibility , a series of tests in which the opacity of @xcite is multiplied by arbitrary constant factors @xmath261 , where @xmath262 , is ran jumping ahead , for @xmath263 , it is found that opacity reduction of around 100 is needed for the planet to survive . figure [ fig : m1_op0.03 ] shows the simulation m1lowop4 with @xmath264 ( @xmath265 ) . the upper panel ( a ) presents the evolution of the planet s location , @xmath59 , planet s radius , @xmath16 , and the hill s radius , @xmath28 , with the solid , dotted and dashed lines , respectively . the planet starts off at @xmath266 au and migrates inward on the time scale of @xmath267 years . initially the planet is much more compact than the hill s radius , but @xmath28 shrinks as the planet s orbit shrinks whereas the planet s radius does not . this eventually causes planet s disruption at @xmath268 years . the middle panel ( b ) of the figure shows several temperature characteristics of the planet . in particular , the blue dash - dot line shows the effective temperature of the planet in isolation , @xmath269 , that is defined by @xmath270 , where @xmath118 is the isolated planet luminosity ( cf . equation [ lrad1 ] ) . the dotted red curve shows the irradiation temperature . since @xmath271 already at birth of the planet , the planet can not contract radiatively , since @xmath272 according to equation [ bath ] . for this reason , the planet does not contract ; @xmath273 const and @xmath274 const . since the planet continues to migrate inward rapidly , it is eventually destroyed . as is seen in panel ( b ) of figure [ fig : m1_op0.03 ] , @xmath233 increases rapidly as the planet migrates in , and at @xmath275 years the thermal bath planet destruction ( [ sec : bath ] ) takes place since @xmath102 exceeds @xmath276 . in passing , it should be noted that if the thermal bath destruction did not occur , this particular planet would have been destroyed very soon nevertheless by tidal disruption since the tidal disruption criterion ( equation [ disruption1 ] ) is almost satisfied at @xmath275 years . panel ( c ) of fig . [ fig : m1_op0.03 ] shows the mass of the solid core assembled in the centre of the protoplanet . the remnant of the disruption is a low mass solid core , @xmath277 . it is composed almost equally of chon and silicates . the presence of chon materials in the core is explained by the arrested evolution of the planet : since it is unable to cool and contract , it stays relatively cold for longer , and this enables chon grains to sediment down into the core . water ice can not condense out even in these conditions , which is consistent with earlier results of @xcite who found that td cores contain little water . the remnant core is very low mass , so that it does not migrate appreciably during the simulation , stalling at @xmath278 au . the simulations m1lowopx were all stopped at @xmath279 million years , but the remnant s migration time is @xmath280 years at the end of the run , so not much migration would occur in the typical 3 million years disc lifetime even if the simulations were continued for longer . figure [ fig : m1_op0.01 ] shows an identical run but with opacity further reduced , @xmath281 , labelled m1lowop7 in table 1 . in this case , the planet s intrinsic luminosity exceeds the irradiating one at early times ( cf . the dash - dot blue and the dotted red curves in panel b ) . the planet thus contracts rapidly , and its central temperature increases rapidly ( the solid black curve in panel b ) . the planet undergoes the second collapse at @xmath282 years , well before it is challenged thermally ( due to external irradiation ) or tidally ( due to tidal forces from the star ) . the core of the planet is slightly less massive than in lowop1 , and is completely dominated by silicates because the planet heats up quickly and the chon grains have little chance to sediment into the core . although we do not present the run m1lowop6 , for which @xmath283 , we note that it ended with the planet disrupted . the conclusion from this series of tests is that opacity reduction by a factor of @xmath284 , compared to the interstellar grain opacity at solar metallicity , is needed to counter the short inward migration times and the thermal bath effect , at least for the parameters of the disc and the planet mass chosen in this section . figure [ fig : mcores ] shows the mass of the cores formed in the series of runs m1lowopx as a function of the opacity reduction factor @xmath285 , and section [ sec : discussion ] considers broader implications of these runs for td theory of planet formation . finally for this section , figure [ fig : m5_op0.1 ] shows simulation m5lowop2 , where the fragment is more massive , @xmath286 . the practical interest in considering more massive fragments is that such fragments cool much more rapidly than @xmath3 jupiter mass ones ( see figure 1 in * ? ? ? therefore , one may reason that dust opacity may not need to be reduced by orders of magnitude for such fragments . indeed , in simulation m5lowop2 ( see table 1 ) , @xmath287 , that is , the dust opacity is reduced by only a factor of 2 , yet the planet is able to contract and collapse before being tidally disrupted . the collapse occurs at @xmath288 years , when the planet is at separation @xmath289 au . one reason for the fast contraction of the planet is that the planet s effective isolated temperature , @xmath269 , is always above the irradiation temperature ( cf . panel b ) , and the second is that the planet s migration is dramatically slowed down when the planet opens a gap in the disc at @xmath290 au . however , the planet eventually migrates all the way to the inner boundary condition at @xmath291 au after about 0.5 million years ( see the inset in panel a of the figure ) . it would become either a hot jupiter , if its migration stalls in the regions closer to the star that are not simulated here , or be swallowed by the star completely . note also that despite having 5 times more mass in metals than the @xmath263 runs , the fragment only manages to assemble a tiny silicate core , @xmath292 ( cf . panel c of the figure and the table ) . this is because the fragment spends too little time in the molecular h regime ( the pre - collapse stage ) , so that even the silicate grains have too short a time window in which to grow and sediment . this finding is consistent with results of @xcite who found that the higher the mass of the fragment , the less efficient is core formation . to save space , the simulation m5lowop1 is not shown in figures but its results are listed in table 1 . in this higher opacity , @xmath293 case , the planet does not manage to contract into the giant planet and is disrupted at @xmath294 au . the core formed in this simulation is only slightly more massive , @xmath295 . from the m5lowopx series of runs we learn that ( i ) the planet must be more massive than about @xmath296 to contract faster than migrate in at solar metallicity , at least for the disc parameters chosen in this series of runs , and that ( ii ) massive planets do not make much more massive cores than less massive planets do . this last finding is not new at all @xcite . in this section , models with pebble accretion onto the gas fragments and with interstellar dust opacity , scaled to the instantaneous metallicity of the fragment , are considered . the initial conditions for the disc and the planet are exactly the same as for the runs m1lowopx , but here the opacity reduction factor is @xmath293 . additionally , since the metallicity of the fragments now varies with time as pebbles accrete onto the fragments , the opacity coefficient is chosen to be proportional to the fragment s metallicity : @xmath297 , where @xmath298 is the interpolated table dust and gas opacity at solar metallicity from @xcite . pebble accretion rate is calculated as described in [ sec : dot_mz ] . table 1 lists input parameters and key results from several runs of this series , which are labelled m1pebx , where x is an integer between 1 and 9 which sets the pebble mass fraction in the disc , @xmath299 . presumably , higher metallicity discs would have higher @xmath300 as grain growth is faster at higher metallicities , so high ( low ) values of @xmath285 may be taken as a proxy for high ( low ) values of disc metallicity . figure [ fig : m1_terr1 ] shows the results of the run m1peb4 , @xmath301 in the format similar to figure [ fig : m1_op0.01 ] to [ fig : m5_op0.1 ] , except that the bottom panel ( c ) also shows the evolution of the planet s metallicity , @xmath107 , defined as the ratio of the mass of the metals to the total mass of the planet . @xmath107 is shown with the cyan dash - triple - dot curve , and the relevant scale is given on the right hand side edge of panel ( c ) . due to pebble accretion , the planet s metallicity increases with time sharply until time @xmath268 years . at that point accretion of pebbles slows down since the planet starts opening a gap in the disc around its location , so that @xmath231 plummets . as a result , the planet contraction slows down ( note that the dashed blue curve in panel ( a ) flattens at that point ) . due to a continuing inward migration , the hills radius of the planet ( red dotted curve in panel a ) continues to shrink , and the planet is tidally disrupted at time @xmath302 years , when the tidal disruption criterion ( equation [ disruption1 ] ) is met . the core s mass at that point is @xmath303 and the planet s location is @xmath304 au . since the core s possible atmosphere is neglected in this paper , the mass of the planet after the disruption is set to that of the core , and the metallicity is set to @xmath305 by definition . also , the temperature of the planet is arbitrarily set to 0 to make the cases of envelope disruption visibly distinct from the cases when the envelope collapses and heats up ( e.g. , fig . [ fig : m5_op0.1 ] ) . in reality the core s temperature may be quite high but that is not modelled here ( [ sec : core ] ) . the core s composition in run m1peb4 is dominated by silicates ( see panel c of fig . [ fig : m1_terr1 ] ) , yet over a third of the core s mass comes from organics ( chon ) . next figure , [ fig : m1_hjup1 ] , shows simulation m1peb3 , which has a slightly higher fraction of pebbles in the disc , @xmath306 . this turns out to be sufficiently high to enforce collapse of the fragment before it is tidally or thermally disrupted . metallicity of this planet at collapse is @xmath307 . this higher metallicity translates into larger grain sizes in the planet ( since @xmath308 term in equation [ dadt_full ] is larger ) and a higher grain growth rate of the core . the mass of the core at collapse is @xmath309 . it is interesting to note that the chon mass of the core is actually slightly lower than in the run m1peb4 , and hence the core is over 90% silicates . this result is driven by a faster contraction rate of the planet in m1peb3 . the time window for chon grains settling into the core is now shorter , so despite more chon mass inside the planet , a smaller mass of that reservoir arrives at and is locked into the core . the planet in simulation m1peb3 continues to migrate in a regime intermediate between type i and ii regimes and eventually arrives at the disc s inner edge . the planet may survive as a hot jupiter if the disc is removed while the planet is still migrating in or if the inner boundary of the disc is cut off by magnetospheric interactions sufficiently far from the star s surface . finally , table 1 lists the results of the most pebble - rich run of the series , m1peb1 , @xmath310 . interestingly , the higher abundance of pebbles did not increase the core s mass or the planet s metallicity at collapse . the latter is due to the fact that the planet collapses when it accretes enough pebbles @xcite , which is independent of the rate at which the metals are added to the planet . the core s mass appears to be lower due to a shorter time span available for grain sedimentation . in summary , runs m1pebx with x=1,2,3 produced a giant planet that collapses before it was tidally disrupted , whereas runs with x@xmath311 resulted in tidal or thermal disruptions of the giant planets - to - be . since a higher pebble accretion rate may be expected at higher metallicity environments , one expects a positive correlation of the giant planet detection frequency with metallicity of the host , as shown by @xcite . figure [ fig : structure ] shows the planet s internal variables as a function of the total ( gas plus metals ) mass enclosed in concentric shells for a planet from run m1peb3 at time @xmath312 years , before it is tidally disrupted . the planet s structure in the pre - collapse ( or pre - disruption ) stages are similar to one another , so fig . [ fig : structure ] is representative of the pre - collapse td planets structure in general . the temperature in the planet , as expected , is maximum in the centre and falls off towards the planet s outer edge ; see the solid curve in panel ( a ) . in the same panel , the dotted red curve shows radius in units of au as a function of the enclosed mass . the planet s outer radius is @xmath313 au . the blue dashed curve shows the local metallicity , @xmath314 , defined as the ratio of the metal s mass in the given mass shell to the total mass of that shell . note that the highest metallicity gas is at the centre of the planet , as may be expected if metals ( grains ) are able to sediment to the planet s centre . in addition , there are two somewhat sharp features in the @xmath314 function , one at @xmath315 , and the other near the outer edge of the planet . these two features mark two important transitions within the planet . the nature of these transitions is best inferred from the bottom panel of fig . [ fig : structure ] , ( c ) , which shows the grain size for the three grain species that are considered in this paper : water , chon and silicates . the vaporisation temperatures for these species are very different ( see figure 1 ) . the silicate grains are the most refractory of the three , so they are able to sediment down all the way to the centre of the cloud in fig . [ fig : structure ] . despite this , note that the grain size for silicates is limited to a few cm . this is a direct consequence of the grain breaking velocity set to @xmath316 cm / s for this simulations . larger grains would sediment faster than this and would fragment due to collisions . in the outer regions of the planet , the grains are not limited by fragmenting collisions . instead , the limitation here comes from the balance between the rate at which the grains grow and the rate at which they sediment down : larger grains are constantly removed from the outer regions of the planet by sedimentation , so there exist a quasi equilibrium between small growth injection at the outer edge of the cloud , grain growth and sedimentation of the grains . grain diffusion due to convection in particular is a very important part of this quasi - equilibrium , since convection opposes grain sedimentation ( e.g. , * ? ? ? effects of convective grain mixing are best revealed by looking at the grain volume density distributions ( the middle panel , b ) . it is notable that there are transitions in @xmath317 , where @xmath318 is the grain volume density of species @xmath252 , at locations where grain size @xmath319 changes abruptly . for example , chon grains are `` small '' in the outer region of the cloud , then `` large '' between @xmath320 and @xmath321 , and very small again for @xmath322 . just as with silicate grains , this is because grain growth is sedimentation - limited in the outer region , then fragmentation - limited in the middle , and finally grain vaporisation dissolves the grains and puts them into the gas phase . convection controls chon grains volume distribution in the inner part of the cloud , where @xmath323 const to a good degree . in the fragmentation - limited range , @xmath324 is strongly decreasing outward , until the grains are small and convection again takes over . water ice grains are able to settle down only through the outermost @xmath325 of the planet , and therefore there is a very large gradient in the water ice concentration there . in the rest of the planet water is very well mixed with the h / he phase . these grain growth , fragmentation and convective mixing equilibria drive the chemical abundances of the species with the planet . in the outer @xmath326 of the planet , water is more abundant than chon and silicates by a factor of @xmath327 rather than the expected ( solar ) abundance ratio ( two ) . on the other hand , silicates are over - abundant over water and chon compared to their initial relative abundances by a factor of @xmath9 in the centre of the planet . td fragments are hence natural thermo - mechanical element differentiation `` factories '' while they are in the molecular h ( pre - collapse ) phase ( see * ? ? ? * for more on this point ) . figure [ fig : mcores ] presents a bird s view of the results of the three series of runs presented in table 1 by showing only the dependence of the core mass , @xmath209 , on the pebble mass fraction , @xmath230 , for the pebble accretion runs m1pebx ( triangles ) , and on the opacity reduction factor , @xmath285 , for the `` low opacity '' runs m1lowopx or m5lowopx ( diamonds and squares , respectively ) . in addition , the runs that resulted in the collapse of the giant planet rather than its disruption are connected with one another by solid segments of the same colour as the symbols . for example , the three black triangles in the upper right of fig . [ fig : mcores ] show the runs m1peb1 to m1peb3 , for which the pebble mass fraction is the highest . the rest of the black triangles refer to runs m1peb4 to m1peb9 , for which the fragments were disrupted before they could collapse . it is of course expected that @xmath230 and @xmath285 are proportional to disc metallicity , since both are expected to increase as @xmath107 increases , with other parameters being equal . we see that opacity reduction by a factor of @xmath284 compared to the interstellar opacity at solar metallicity is needed for planets of 1 jupier mass ( red diamonds ) to collapse sooner than they are disrupted . it could be argued that planet migration inside lower mass discs than @xmath328 used universally for all the tests in table 1 could allow for a smaller reduction in opacity ( that is , higher @xmath285 ) , but much lower mass discs are unlikely to become self - gravitating and give birth to a gi fragment in the first place . furthermore , observed giant planets in the solar system and beyond ( e.g. , * ? ? ? * ) are known to be over - abundant in metals by a factor of @xmath9 or more than @xmath329 . thus the needed opacity reduction is a factor of @xmath222 for the observed metal - rich giant planets of about one jupiter mass . while grain growth can reduce opacity @xcite , grain fragmentation should be included in such models as well @xcite . we have seen that collisions with grains as large as a few cm in size are frequent enough inside the pre - collapse fragments ( see [ sec : internal ] ) to limit grain size by fragmentation . this was found for the breaking velocity set to @xmath330 m s@xmath126 , a relatively high value , but perhaps reasonable for the refractory grains strengthened by sintering in the high temperature environment inside the fragment ( e.g. , * ? ? ? amorphous grain aggregates are found to fragment at velocities as little as @xmath331 m s@xmath126 . despite these uncertainties in the input grain physics , it is hard to imagine that fragmentation of smaller grains would not be efficient inside the pre - collapse planets . it appears unreasonable to us to require the dust opacity to be reduced by such a huge factor as @xmath222 . even if nature does manage to reduce dust opacity to such tiny values somehow , the trends of the low opacity td models for giant planet formation directly contradict the observations . it is obvious from fig . [ fig : mcores ] that low metallicity environments would be more hospitable to planet formation via gi / td models _ if _ radiative cooling was the bottle neck for giant planet contraction , in agreement with the results of @xcite . this trend is exactly opposite to what is observed @xcite . furthermore , as seen from fig . [ fig : mcores ] , the core masses assembled by such rapidly collapsing gas fragments would be rather small , @xmath332 . when the `` low opacity '' run fragments ( diamonds in fig . [ fig : mcores ] ) are tidally or thermally disrupted , the mass of their cores is far too small to explain rocky terrestrial planets , save for the more massive super - earth planets . considering more massive gas fragments relaxes the requirement for the low opacity , as is seen from the sequence of squares in fig . [ fig : mcores ] . for @xmath286 , the required reduction in opacity is only a factor of @xmath333 , which is quite reasonable . however , the mass of the cores assembled inside such high mass planets is still very low . this is because more massive planets are hotter to begin with and contract more rapidly , leaving too little time for grains to settle down ( as was found previously by * ? ? ? * ; * ? ? ? disruptions of high mass fragments would therefore fail to explain the abundant massive rocky planets ( e.g. , * ? ? ? the high mass gas fragments would also not be able to explain formation of lower mass giant planets , @xmath334 , which are much more abundant in the data than planets of mass @xmath335 . there is also no clear reason why the high mass radiatively cooling fragments would follow a positive metallicity correlation . finally , whatever the mass of the gas fragment , the low opacity models do not provide an explanation for why giant planets are found to be much more abundant in metals than their host stars . therefore , it is our opinion that in terms of development of td planet formation theory that could account for all types of observed planets , low dust opacity models are not only physically unlikely , they are a dead end . in contrast , pebble accretion models ( i ) yield a positive planet - metallicity corelation @xcite ; ( ii ) are very metal rich , e.g. , @xmath336 ( cf . entry @xmath337 in table 1 ) ; ( iii ) can assemble sufficiently massive solid cores ( cf . [ fig : mcores ] ) to potentially explain rocky planets as remnants of disrupted gas fragments @xcite . while there is a number of issues not yet addressed by td , it appears to us that td with pebble accretion has a significant potential , and it must be studied further . [ cols="<,^,^,^,^,^,^,^,^,^",options="header " , ] in this article , steps are taken towards a numerically efficient way of computing the planet - disc coevolution in the framework of the tidal downsizing hypothesis for planet formation . two ways of avoiding tidal disruption during planet migration from the outer @xmath284 au disc into the inner one are considered : low dust opacity in the fragments and pebble accretion on to the fragments . the former pathway requires extreme dust opacity reduction ( by @xmath1 orders of magnitude ) , predicts a negative correlation of giant planet frequency of occurrence with metallicity , and leaves low mass @xmath338 solid cores as disruption remnants . the significant over - abundance of observed giant planets in metals @xcite is also not addressed by this class of models . we conclude that fragments cooling and contracting radiatively rather than due to pebble accretion are not likely to explain observed planets . the most probable physical reason for this irrelevance of low opacity models is that collisional grain fragmentation keeps the supply levels of small grains sufficiently high ( as argued for protoplanetary discs by * ? ? ? * ) , so that dust opacities never drop by orders of magnitude below the interstellar values . in contrast , high opacity ( @xmath339 ) pebble accreting fragments produce a positive metallicity correlation , yield massive solid cores as remnants and also explain why giant gas planets are strongly over - abundant in metals . on the basis of this paper , pebble accretion appears a key ingredient to a successful td model for planet formation . we hope this paper will serve as impetus for the broader planet formation community to start investigating td in greater depth than has been achieved so far . theoretical astrophysics research in leicester is supported by an stfc grant . the author thanks richard alexander , discussions with whom were very influential in shaping the disc evolution model of the paper . this paper used the dirac complexity system , operated by the university of leicester , which forms part of the stfc dirac hpc facility ( www.dirac.ac.uk ) . this equipment is funded by a bis national e - infrastructure capital grant st / k000373/1 and dirac operations grant st / k0003259/1 . dirac is part of the uk national e - infrastructure .	tidal downsizing ( td ) is a recently developed planet formation theory that supplements the classical gravitational disc instability ( gi ) model with planet migration inward and tidal disruptions of gi fragments in the inner regions of the disc . numerical methods for a detailed population synthesis of td planets are presented here . as an example application , the conditions under which gi fragments collapse faster than they migrate into the inner @xmath0 a few au are considered . it is found that most gas fragments are tidally or thermally disrupted unless ( a ) their opacity is @xmath1 orders of magnitude less than the interstellar dust opacity at metallicities typical of the observed giant planets , or ( b ) the opacity is high but the fragments accrete large dust grains ( pebbles ) from the disc . case ( a ) models produce very low mass solid cores ( @xmath2 ) and follow a negative correlation of giant planet frequency with host star metallicity . in contrast , case ( b ) models produce massive solid cores , correlate positively with host metallicity and explain naturally while giant gas planets are over - abundant in metals . [ firstpage ] accretion , accretion discs quasars : general black hole physics galaxies : evolution stars : formation
experimental @xcite and numerical @xcite studies of the dynamics of driven disordered media , such as vortex arrays in type - ii superconductors and charge density waves in metals , have clearly indicated that when the disorder is strong these systems exhibit a spatially inhomogeneous plastic response upon depinning , without long wavelength elastic restoring forces @xcite . in this plastic regime dislocations proliferate due to the interplay of drive , disorder and interactions , and the system is broken up in fluid - like regions flowing around solid regions @xcite . it was proposed recently @xcite that a description of shear deformations in this plastic regime may be obtained by focusing on the dynamics of coarse - grained degrees of freedom ( the solid - like regions ) and replacing the elastic couplings of local displacements with viscoelastic couplings of local velocities . it was found that the simplest , maxwell , model of viscoelasticity yields interesting features , such hysteretic depinning and switching , which have also been observed in experiments . the local viscoelastic couplings were introduced as an ad hoc way of mimicking the presence in the system at any given time of unbound dislocations , whose motion is in turn responsible for plastic slip . in this paper we provide some justification to the model studied in ref . by showing that the dynamics of a two - dimensional crystal with a finite concentration of _ annealed _ free dislocations , as well as vacancy and interstitial defects , is governed by the hydrodynamic equations of a viscoelastic medium ( with hexatic order ) . starting with the hydrodynamic equations for a two - dimensional solid with finite concentrations of dislocations , vacancies and interstitials obtained some time ago by zippelius , halperin and nelson @xcite ( zhn ) , we show that such equations can be recast in the form of hydrodynamic equations for a viscoelastic medium , with a maxwell form for the nonlocal transport coefficients @xcite . it has of course long been recognized that the creation and motion of dislocations are the main mechanism by which a crystal undergoes plastic deformations @xcite . a number of efforts @xcite have been made to describe the motion of dislocations through crystalline solids and relate the plastic strain rate to the dislocation dynamics . there has also been work specifically on the effect of dislocation motion on linear and nonlinear stress relaxation in a two - dimensional crystal @xcite , where it was shown _ for a particular geometry _ that the crystal responds viscoelastically when free dislocations are present . our work generalizes this by obtaining equations that describe the viscoelastic response of a two - dimensional solid to stresses in an _ arbitrary geometry_. the equations incorporate viscoelastic effects in the response to both shear and compressional deformations , and have precisely the structure of the viscoelastic equations for a simple viscous fluid @xcite . furthermore , our derivation yields expressions for the ( maxwell ) relaxation times for shear and compression in terms of microscopic parameters . of course , in the presence of _ quenched _ disorder and drive , dislocations are continuously generated and healed in the system . furthermore , they are not always free to move and relax a local stress as they may be pinned by disorder . for these reasons one may question the assumption behind the model introduced in ref . , namely that the plastic response of extended media driven over strong quenched disorder at zero temperature may be described at large scale by the same equation that govern the stress relaxation due to the motion of annealed dislocations . at best , the effective density of `` free '' dislocations that can relax local stresses will be a strong function of the applied driving force , disorder , and possibly time . as shown in ref . , the depinning of a driven viscoelastic medium does , however , exhibit a number of features seen in experiments , namely memory effects and coexistence of weakly and strongly pinned degrees of freedom . it does , therefore provide a useful starting model for the description of the complex dynamics of these systems . finally , it should be pointed out that crystals with annealed dislocations and simple viscous fluids are not the only examples of viscoelastic media . linear viscoelasticity is of course one of the distinctive properties of complex and polymeric fluids , that can fill containers of any shape and yet may shrink like rubber when stretched and released quickly . molecular theories of viscoelasticity for complex fluids have been developed by various methods and there is a vast literature on the subject @xcite . it is well known that a dense fluid has insufficient time to flow in response to a high frequency strain rate , but instead reacts elastically , as a solid would @xcite . this leads to the appearance in the fluid of propagating shear waves , with an associated peak at nonzero frequency in the spectrum of transverse current fluctuations . such viscoelastic effects are easily incorporated phenomenologically into the hydrodynamic description of fluids @xcite . if a shearing force is applied to a fluid yielding a stress @xmath0 , the local strain at each point can be expressed in terms of derivatives of the displacement field @xmath1 at that point . in a steady - state situation the flow is purely viscous and the stress is proportional to the local rate of strain , @xmath2 where @xmath3 is the shear viscosity . by contrast , if the force is applied suddenly , the instantaneous displacement is related to the stress via a typical elastic stress - strain relation , @xmath4 where @xmath5 is the instantaneous ( high - frequency ) shear modulus . an interpolation between these two forms yields the maxwell model of viscoelasticity , where the stress - strain rate relation takes the form of a differential equation , @xmath6 the local flow velocity , @xmath7 , of the constituent particles is identified as the rate of change of the displacement , @xmath8 , and @xmath9 is called the maxwell relaxation time . the maxwell model of viscoelasticity given by eq . ( [ maxwell ] ) yields elastic response to high frequency shearing forces ( @xmath10 ) and viscous response at low frequency ( @xmath11 ) . in general , viscoelastic effects are present in the response to both shear and compression and can be described by modifying the hydrodynamic equations of a fluid by introducing nonlocal ( in space and time ) thermodynamic quantities and transport coefficients . this can be done phenomenologically through the use of frequency sum rules as constraints , or can be formulated more systematically by deriving such equations via formal projection operator methods or kinetic theory techniques @xcite . in its simplest realization this approach yields the maxwell model of viscoelasticity . in this paper we show that the hydrodynamic equations for a two - dimensional solid with a finite concentration of free dislocations proposed by zhn are equivalent to a set of hydrodynamic equations for a simple fluid ( with hexatic order ) , with a maxwell - type stress - rate of strain relation . for convenience , we define a symmetrized strain rate , @xmath12 , as @xmath13 where we have separated the traceless part @xmath14 and the trace @xmath15 . it is also useful to separate the stress tensor @xmath16 into its traceless part and its trace , i.e. , @xmath17 we find that , at long wavelengths , the traceless part of the stress tensor is related to the strain rate according to @xmath18 where @xmath19 is the shear relaxation time , which is obtained in terms of microscopic parameters ( see eq . ( [ tau_s ] ) below ) . equation ( [ sigmatilde_ij ] ) is precisely of the maxwell form . if the product @xmath20 is identified with the shear viscosity @xmath3 of the fluid , eq . ( [ sigmatilde_ij ] ) yields viscous flow in response to low frequency shearing forces . it also describes the fact that a liquid always flows in response to a steady shear . in contrast , even a liquid has a finite low frequency compressional modulus , that is , it resists elastically low frequency compressional deformations . the trace of the stress tensor , @xmath21 , contains both reversible contributions proportional to pressure changes arising ( assuming temperature fluctuations are negligible ) from the fractional change in volume and irreversible and dissipative contributions proportional to velocity gradients . it can be written as @xmath22 with @xmath23 the deviation of the density from its equilibrium value @xmath24 , and @xmath25 the low frequency bulk modulus of the solid ( which generally depends on the density of vacancies and interstitials defects ) . to a good approximation @xmath25 is also the compressibility of the liquid phase @xcite . we find that the dissipative part @xmath26 of the stress tensor is related to the local rate of strain via the equation @xmath27 where @xmath28 is the high frequency compressional modulus of the solid in the absence of defects . at high frequency defects do not have time to respond to an applied force and therefore do not contribute to the elastic constants . finally , @xmath29 is the compressional relaxation time , which again is given below in terms of microscopic quantities . both relaxation times , @xmath19 and @xmath29 are inversely proportional to the density of free dislocations , @xmath30 . as a result , they both diverge as the melting temperature @xmath31 of the two - dimensional solid is approached from above as @xcite @xmath32 with @xmath33 . at the longest length scales , the bulk viscoelasticity described by eq . ( [ sigma_kk^ve ] ) is due solely to dislocation _ climb _ , while the shear viscoelasticity is due to both dislocation climb and _ glide_. dislocation climb which requires the creation of vacancies and interstitials , proceeds more slowly than dislocation glide which does not involve a change in particle number . for this reason , @xmath34 , in general . of course , the 2d solid does not melt directly into a isotropic liquid at @xmath31 , rather it enters the hexatic phase , where it has only short - ranged translational order , but retains quasi - long - ranged bond orientational order ( boo ) . at the longest length scales , the presence of boo does not affect the stress - rate of strain relations . when terms of higher order in the gradients of the hydrodynamic fields are included , the existence of boo becomes apparent and modifies the stress - strain rate relations at finite length scales , as discussed in section [ grad2 ] below . in particular , this gives wavevector - dependent corrections to @xmath19 and @xmath29 , corresponding to a distribution of relaxation times . the rest of the paper is organized as follows . in section [ review ] we review the hydrodynamic equations for a two - dimensional solid with vacancy and interstitial defects @xcite . in section [ grad1 ] the viscoelastic stress - rate of strain relations are derived in the longest length scale regime , and expressions for the long - wavelength viscoelastic memory times are presented . in section [ grad2 ] the effects of incorporating higher order gradients are investigated , including the wavevector dependent corrections to @xmath19 . we begin by reviewing the hydrodynamics of the solid phase , following refs . , and . we consider a two - dimensional hexagonal crystal with a finite density of vacancies and interstitials . as pointed out by martin @xcite , incorporating a finite density of point defects is crucial to properly distinguish between local lattice deformations and local changes in the number density . for simplicity , temperature fluctuations are neglected in all of the following . in the absence of dislocations , the equations for conservation of density , @xmath35 , momentum , @xmath36 , and net defect density , @xmath37 , i.e. , density of interstitials minus density of vacancies , and the equation of motion for the displacement field , @xmath38 , can be written as @xmath39 [ xtalhdyro ] where @xmath40 is the mass of the constituent particles and @xmath24 is the average number density . these conservation laws must be completed with constitutive equations for the two fluxes , the momentum flux or stress tensor , @xmath41 , and the defect current , @xmath42 . assuming purely relaxational dynamics for the defects , @xmath43 is given by @xmath44 where @xmath45 is a constant kinetic coefficient and @xmath46 is the defect pressure , @xmath47 with @xmath48 a susceptibility proportional to the sum of the equilibrium densities of vacancies and interstitials , and @xmath49 a constant associated with the coupling between changes in defect density and bulk distortions in the lattice . the constitutive relation for the stress tensor , @xmath16 , is @xmath50 where @xmath51 is the symmetric part of the strain tensor @xmath52 , given by @xmath53 the coefficients @xmath5 and @xmath54 are the lam constants . we note that the equilibrium bulk modulus , @xmath25 , of the solid is defined in terms of the response to a static isotropic compression as @xmath55^{-1}$ ] , with @xmath56 the area of the crystal . in the presence of defects it is then given by @xmath57 on the other hand , if one were to measure the bulk modulus on time scales long compared to the elastic relaxation time , but short compared to the defect diffusion time , the effective bulk modulus , denoted by @xmath28 , would be @xmath58 corresponding to the purely elastic contribution obtained in the absence of defects . in a finite - frequency measurements , the value @xmath59 is reached at sufficiently high frequency . finally , changes in the defect density can be related to changes in the overall density and bulk strains in the lattice , as @xmath60 where @xmath61 and @xmath62 are the deviations of the density and defect density from their equilibrium values , respectively . in the presence of dislocations , it is no longer possible to define a continuous single - valued displacement field , @xmath1 . one can define a single - valued , but discontinuous displacement field by introducing cuts at the location of the dislocations . the strain field @xmath63 is still single valued and continuous , except at the location of the dislocations , and it satisfies @xmath64 where @xmath65 is the lattice spacing , @xmath66 the antisymmetric unit tensor , and @xmath67 is the burgers - vector charge density . for a set of @xmath68 free dislocations at discrete points @xmath69 , @xmath70 , with burgers vector @xmath71 , it is given by @xmath72 dislocations are assumed to be annealed and free to equilibrate . the burgers vector charge density is an extra hydrodynamic variable with a conservation law , @xmath73 where @xmath74 is the burgers current of the @xmath75th component of dislocation charge in the @xmath76th direction . for length scales longer than the average distance between free dislocations , @xmath77 , and time scales longer than the mean collision time between free dislocations , one can make some simplifying approximations . specifically , one can ignore the effect of dynamical polarization of bound dislocation pairs and also processes by which neutral dislocations pairs ( i.e. , dislocation pairs of opposite burgers vectors ) are created and annihilated . it was then shown in ref . , that the constitutive relation for the burgers current can be written as @xmath78 the three terms on the right hand side of eq . ( [ dislocationcurrent ] ) correspond to the motion of dislocations in response to stresses in the lattice ( the usual peach - koehler force ) , gradients in defect pressure and gradients in the dislocation density , respectively . the tensors @xmath79 and @xmath80 are given by @xmath81\;,\nonumber\\ \label{c}\\ d_{jl}^{ik } & = & \frac{d_g - d_c}{4}(\delta_{ij}\delta_{kl } + \delta_{jk}\delta_{il } ) + \frac{d_g+3d_c}{4}\delta_{jl}\delta_{ik}\ ; , \nonumber\\ \label{d}\end{aligned}\ ] ] [ cdtensors ] where @xmath82 and @xmath83 are the diffusions constants for dislocation glide and climb , respectively . the conservation laws for density and momentum , eqs . ( [ n ] ) and ( [ g ] ) , remain unchanged in the presence of dislocations . a new term must , however , be added to the equation for the conservation of defect density , reflecting the fact that a climbing dislocation acts as a source or sink of vacancies and interstitials . equation ( [ n_delta ] ) is then replaced by @xmath84 since @xmath1 is multi - valued , it is more convenient to consider an equation of motion for the strain rather than eq . ( [ u ] ) for the displacement field . this is given by @xmath85 where , to linear order , @xmath86 . the first two terms are simply obtained by differentiating the right hand side of eq . ( [ u ] ) . the third term arises from the presence of dislocations . the conservation laws of density , eq . ( [ n ] ) , momentum density , eq . ( [ g ] ) , defect density , eq . ( [ adjustedn_delta ] ) , and burgers vector charge , eq . ( [ bconservation ] ) , together with the constitutive equations for the currents , eqs . ( [ defectcurrent]),([elasticsigma_ij ] ) and ( [ dislocationcurrent ] ) , and the equation of motion for the local strains , eq . ( [ conservationofw_ij ] ) , constitute a closed set of equations to describe the dynamics of a 2d solid with an equilibrium concentration of free annealed dislocations and of vacancy and interstitial defects . the work by zhn focused on the transverse and longitudinal hydrodynamic modes of these equations . these authors showed that the low frequency modes of a solid with dislocations are the same as those of the hexatic phase . in contrast , our goal here is to eliminate the dislocations and defects degrees of freedom and obtain a closed set of hydrodynamic equations for the density , momentum density and bond angle . by this procedure we show that the hydrodynamic equations of a solid with free dislocations are equivalent , at long wavelengths , to the hydrodynamic equations of a viscoelastic medium with bond angle order . the latter consists of conservation laws for the density and momentum density , an equation of motion for the bond angle , with a constitutive equation for the stress tensor that takes the form of a differential equation . our strategy is to eliminate the defect density , @xmath87 , and the burgers - vector charge density , @xmath88 , from the hydrodynamic equations of a solid with dislocations discussed in the previous section . the outcome will be a closed set of equations for density , momentum density , bond angle field and stress tensor . it is convenient to write the strain tensor @xmath52 as the sum of its symmetric and antisymmetric parts , @xmath89 only the symmetric part enters in the expression for the stress tensor , according to eq . ( [ elasticsigma_ij ] ) . one can therefore invert eq . ( [ elasticsigma_ij ] ) to express @xmath90 as a function of the stress tensor and of the density , as @xmath91 where eq . ( [ densitiesrelation ] ) was used to eliminate the defect density in favor of the density and the strain . the antisymmetric part , @xmath92 , has only one independent component which is simply the bond angle field , @xmath93 @xcite , @xmath94 the equation of motion for the strain , eq . ( [ conservationofw_ij ] ) , can then be split into an equation of motion for the symmetrized strain , given by @xmath95 and an equation for the bond angle field , @xmath96 we note that the trace of eq . ( [ conservationofw_ij^s ] ) simply gives the continuity equation for the density . we can now use eq . ( [ w_ij^s ] ) to rewrite the equation of motion for the symmetric part of the strain tensor as a differential equation for the time evolution of the stress tensor , with the result @xmath97 in order to eliminate the defects and dislocation degrees of freedom , we now insert the constitutive equations for the defect and dislocation fluxes on the right hand side of eqs . ( [ bondangle ] ) and ( [ stress1 ] ) and use @xmath98 and @xmath99 to eliminate the defect density and dislocation density in favor of density and strain field . finally , the strain field is related to the stress tensor and to the bond angle field using eqs . ( [ w_ij^s ] ) and ( [ theta_dis ] ) . this completes the transformation of the equation for the strain tensor into two closed coupled equations for the stress tensor and the bond angle field . for clarity , we present in this section the equations obtained when only terms of lowest order in the hydrodynamic fields are retained . in the next section we will include higher order terms and show that these yield wavevector dependent relaxation times for shear and compression . at long wavelengths , the gradients of the defect current density can be neglected both in the equation for the bond angle , eq . ( [ bondangle ] ) , and in the equation for the stress tensor , eq . ( [ stress1 ] ) . similarly , one can drop the contribution to the dislocation current due to diffusion of dislocations and approximate @xmath100 or @xmath101\;.\end{aligned}\ ] ] carrying out some tedious algebra , we obtain the desired set of hydrodynamic equations , consisting of exact conservation laws for density and momentum , the equation for the bond angle field @xcite , @xmath102 and the constitutive equation for the stress tensor , @xmath103 where we recall that @xmath104 . the shear and compressional relaxation times ( or `` maxwell '' times ) are given by @xmath105 where @xmath106 is the `` effective '' bulk modulus that one would obtain by integrating out the defect density fluctuations from the free energy . ( 25,160)(0,-32 ) ( -20,-25 ) ( 20,20)(0,0 ) the shear relaxation time has a simple physical interpretation . considering , for simplicity , only dislocation glide , the motion of a dislocation of burgers vector @xmath107 a distance @xmath108 along its slip line contributes an amount @xmath109 to the displacement of the half crystal , with @xmath110 the length of the crystal along the slip direction . this is illustrated for a square lattice in fig . ( [ orowan ] ) . the corresponding change in shear strain is @xmath111 , with @xmath112 the width of the crystal in the direction normal to the slip line . the change in the strain @xmath49 in a time @xmath113 due to @xmath68 dislocations is then @xmath114 , with @xmath115 the dislocation glide velocity . this is the well - known orowan law @xcite relating the plastic strain rate to the dislocation velocity . dislocation motion is driven by local stresses and we can write @xmath116 , with @xmath117 the dislocation mobility and @xmath118 the force on a dislocation due to the shear stress , @xmath119 . if we define a shear relaxation time by @xmath120 , and assume @xmath121 , we immediately obtain @xmath122 . when both glide and climb can occur the mobilities are simply additive . an alternative interpretation of eq . ( [ tau_s ] ) is as follows . the rate at which dislocations diffuse thermally ( via glide and climb ) a distance of the order of the typical separation @xmath123 between free dislocations is @xmath124 . this is just the mean time for collisions between free dislocations . the ratio of the rate of diffusion of dislocations under the influence of shear elastic forces to the rate of thermal diffusion is of the order of the ratio of a typical shear elastic energy , @xmath125 to thermal energy , @xmath126 . with this energetic argument we estimate @xmath127 which is the result given in eq . ( [ tau_s ] ) . interactions between dislocations are negligible only at time scales longer than the mean collision time , @xmath128 as a result our model applies only if @xmath129 . since @xmath130 , the condition @xmath129 simply corresponds to being above the melting temperature , @xmath31 . it is not surprising that the bulk memory time is independent of @xmath82 , since pure compressional strains can only relax through dislocation climb . also , since dislocation climb involves a change in particle number , @xmath83 is , in general , much smaller than @xmath82 , so that @xmath131 . in addition , above the melting temperature the response to shear distortions is _ purely _ viscoelastic while the response to bulk distortions is effectively still elastic at large scales , with some dissipative effects ( due to the climbing of dislocations ) which , due to the smallness of @xmath83 , are expected to be negligible . in the limit of @xmath132 , i.e. , as @xmath31 is approached from above , the shear and compressional relaxation times diverge , @xmath133 according to eq . ( [ tau_divergence ] ) , and we obtain @xmath134 in the absence of dislocations @xmath135 . integrating eq . ( [ sigma_ijtau->infty ] ) then yields the shear stress - strain relation for an elastic solid , @xmath136 similarly , to lowest order in the gradients , we can neglect the rate of change of defect density . integrating eq . ( [ sigma_kktau_b->infty ] ) and using eq . ( [ densitiesrelation ] ) , we then obtain @xmath137 and @xmath138 far above @xmath31 , free dislocations proliferate and both @xmath139 and @xmath29 become very small . in this limit eq.([totalstress ] ) then reduces to @xmath140 identifying the combinations @xmath20 with the shear viscosity , @xmath141 , and the combination @xmath142 with the bulk viscosity , @xmath143 , eq . ( [ sigma_ijtau->0 ] ) is just the stress - rate of strain relation of a viscous liquid . when only terms of lowest order in the gradients of the hydrodynamic fields are retained , it is not apparent from the viscoelastic stress - rate of strain relation that a crystal with annealed dislocations is a hexatic , rather than a isotropic viscous liquid . the presence of quasi - long - range bond orientational order in the hexatic manifests itself in a resistance to distortions that produce a non - zero gradient in the bond - angle field , @xmath144 . contributions to the hydrodynamic equations from gradients of @xmath144 correspond to spatial gradients of @xmath52 , which have been neglected in this section . such terms must be retained to see a signature of bond orientational order in the stress - rate of strain relations . this is done in the next section . in this section we incorporate all diffusive terms that were neglected in the previous section . we do so , however , only for a simplified model where dislocation climb is neglected . this corresponds to freezing out fluctuations in the defect density . it is an excellent approximation in most cases as the diffusion constant @xmath83 for dislocation climb is generally very small due to the large energy cost of creating vacancies and interstitials . in the absence of fluctuations in the defect density the relation between symmetrized strain and stress simplifies to @xmath145 the exact conservation laws for density and momentum are of course unchanged and are still given by eqs . ( [ n ] ) and ( [ g ] ) , respectively . the equation of motion for the bond angle field now contains terms due to dislocation diffusion , which couple to gradients of stress and it is given by @xmath146 > from eq . ( [ thetaeom3 ] ) , one sees that the presence of gliding dislocations leads to a dissipative term @xmath147 and a coupling to @xmath148 . in the absence of fluctuations in the defect density , the stress tensor s trace is simply @xmath149 we therefore focus on its traceless part , @xmath150 inclusion of the extra diffusive terms in the constitutive relation for @xmath148 will yield terms involving the gradient of the burgers vector charge density . these terms are easily recast as gradients of bond angle and stress , resulting in the following constitutive relation for @xmath148 @xmath151 where eq . ( [ sigma_kk ] ) relating @xmath21 and @xmath152 has been used . it is instructive to rewrite eq . ( [ sigmatilde_ij ] ) in fourier space . the equation for @xmath153 is then given by @xmath154 with @xmath155 and where we have defined @xmath156 one is naturally led to introduce a finite wavevector shear relaxation time , @xmath157 , given by @xmath158 when the constitutive relation is written in this form , it is apparent that @xmath159 is the length scale beyond which the small gradient approximation is valid . at low frequency eq . ( [ sigmatilde_ijfs1 ] ) reduces to @xmath160 by comparing eq . ( [ sigmatilde_ijfsdc ] ) to the corresponding relation for a hexatic given in zhn , @xmath161 it is natural to identify the combination @xmath162 with @xmath163 , the elastic constant associated with long wavelength distortions of the bond angle field . the length scale @xmath159 can then be written as @xmath164 and it represents the length scale below which the energy associated with bond angle distortions becomes comparable to that associated with spatially uniform shear distortions . the effective shear relaxation time , @xmath157 , is larger at short length scales , as the additional stiffness due to bond orientational order makes the system more solid - like . finally , it should be noted that the term in eq . ( [ sigmatilde_ijfs1 ] ) @xmath165 @xmath166 only contributes to the equation for the longitudinal ( @xmath167 ) component of the momentum , but not to that for the transverse part ( @xmath168 ) . this indicates that , even if dislocation climb is forbidden , at short length scales the longitudinal component of momentum is affected by dislocation glide , which yields higher order pressure - like gradients in the equations of motion . clearly a regime where the wavevector dependence of the shear relaxation time is important only exists provided @xmath169 , with @xmath77 the average distance between free dislocations . at length smaller than @xmath170 interactions among dislocations can no longer be neglected and our model becomes invalid . it is easy to see that @xmath171 with @xmath172 the melting temperature of the two - dimensional lattice . since our work always describes the region @xmath173 , where free dislocations are present in the crystal , the wavevector dependence of the relaxation time will generally be important . in a dislocation - free solid , the bond angle field is slaved to the displacement field and is given by @xmath174 in the presence of dislocations it can be rewritten as @xmath175 which , when inverted , gives eq . ( [ theta_dis ] ) . the diffusive term in the equation for the time evolution of @xmath144 originates from the dislocation diffusion current that was neglected in the other equations . this term must , however , be incorporated in the equation for the bond angle field to obtain the lowest order term in the relaxation rate of the hexatic mode .	it is shown that the dynamics of a two - dimensional crystal with a finite concentration of dislocations , as well as vacancy and interstitial defects , is governed by the hydrodynamic equations of a viscoelastic medium . at the longest length scales the viscoelasticity is described by the simplest _ maxwell _ model , whose shear and compressional relaxation times are obtained in terms of microscopic quantities , including the density of free dislocations . at short length scales , bond orientational order effects become important and lead to wavevector dependent corrections to the relaxation times . = 10000 2
recent low - temperature transport experiments , utilizes the quantum hall based interferometers to investigate the quantum nature of particles . particularly , the aharonov - bohm ( ab ) interference experiments became a paradigm @xcite , which infers the ab phases of both the electrons and the quasi- particles . the single particle edge - state picture is used to describe transport , meanwhile electrostatics is enriched by interactions and both are used to explain the observed ab oscillations @xcite . however , the actual spatial distribution of the edge - states is still under debate for real samples , since handling the full electrostatics is a formidable task , although , several powerful techniques are used @xcite . by full electrostatics we mean both handing the crystal growth parameters and the edge " definition of the interferometer , _ i.e. _ gate , etched or trench - gated . in this work , we provide a semi - analytical scheme to model ab interferometers induced on a two dimensional electron gas ( 2deg ) by solving the 3d poisson for the given hetero - structure @xcite . our calculation scheme also takes into account the lithographically defined surface patterns to obtain the electron and potential distributions under quantized hall conditions @xcite . the distinct part of our calculation is that we can handle both gate and etching defined geometries . our findings show that the etching defined samples provide a sharper potential profile than that of gate defined @xcite . in addition we can define the structure with trench gating , which is the case for the experiments , and show that the formation of the edge - states is strongly influenced . field ( upper panel ) , together with the corresponding potential ( thick black line ) , landau levels ( broken lines ) and wave functions ( red thick curves , lower panels ) , whereas @xmath1 denotes the chemical potential ( or fermi energy , at equilibrium and in 2d ) . the ellipsis indicate the @xmath2 field interval where iss become leaky ( left ) or not ( right ) . ] the high quality gaas / algaas hetero - structures provide a great laboratory for the condensed matter physicists to perform experiments inferring the phase of the particles and quasi - particles @xcite . usually , an interferometer is defined on a ultra - high mobility ( @xmath3 @xmath4/v.s ) wafer by means of etching and/or gating @xcite and so - called edge - states @xcite are utilized as phase - coherent `` beams '' to manipulate the interference pattern . these edge states are direct sequence of landau quantization due to the perpendicular magnetic field and bending of them due to the finite size of the physical system . first considerations of the edge - states neglect the electron - electron interaction and base their models on 1d ballistic channels @xcite to explain the integer quantized hall effect ( iqhe ) . however , later interactions were also taken into account and the 1d edge states were replaced by compressible / incompressible strips @xcite . essentially , chklovskii _ et.al _ attributed the properties of the 1d channels to the compressible strips where the fermi energy ( locally ) equals the landau energy . hence , the current is carried by the compressible strips @xcite . in contrast , a. m. chang and others claimed that the current is flowing from the incompressible strips due to the absence of back - scattering @xcite , since fermi energy falls in between two consequent landau levels both the electric field and conductivity vanish locally . all the above models provide a reasonable explanation to the iqhe , however , the 1d channel and compressible strip pictures both require bulk ( localized ) states to infer the transitions between the hall plateaus . meanwhile , the incompressible strip picture is almost self - standing and provides a self - consistent model both to describe electrostatics and transport . although , the incompressible picture sounds reasonable in explaining the iqhe unfortunately , it is a challenge to explain how to inject current from the contacts to these strips due to their `` incompressibility '' @xcite . moreover in the case of interference , partitioning should take place between these incompressible strips which is apparently `` difficult '' . here , we would like to answer this question as quantitative as possible . first of all , in experiments the aharonov - bohm oscillations are observed in between the plateaus @xcite . this means that the outermost edge channels are already much narrower than the magnetic length @xmath5 ( @xmath6 , where @xmath7 is the planck constant divided by two @xmath8 , @xmath9 is the charge and @xmath2 represents the magnetic field strength ) , hence become `` leaky '' . in the sense that the widths of the outermost incompressible strips are narrower than the quantum mechanical length scales . the models which consider many compressible strips utilize the thomas - fermi approximation that fail if the potential landscape vary strongly on the scale of the wave function ( approximately the magnetic length ) and this is exactly the case at the interference field interval . as an illustration we show the potential landscape at a certain cut across a hall bar in fig . [ fig : fig1 ] ( lower panels , black thick line ) , together with the hall resistance ( upper panel ) and approximate wave functions ( thick curves , lower panel ) . on one hand , once the incompressible strips become narrower than the wave extend , the strips are no longer incompressible as seen in the left panel of fig . [ fig : fig1 ] , which occurs at the lower side of the quantized hall plateau . on the other hand , within the plateau a well developed incompressible strip ( is ) exists , which decouples the hall probe contacts . this makes us to conclude that , the partitioning can take place between the `` leaky '' iss , which occurs only out of the plateau regime . the next question to be answered is `` why we do not observe interference in some samples ? '' . the answer is quite easy : one should bring these strips to close enough vicinity , so that scattering can take place between two channels . therefore , next we present some of our numerical results which treat a real interference device patterned on a 2des by means of trench gating @xcite . we obtain the actual electron density distribution starting from the real crystal growth parameters , here we use the structural information provided by v. j. goldman . the numerics is as follows : first the crystal parameters are given as an input which contains the dielectric constants of the material ( 12.4 for gaas / algaas structures ) together with the dopant density ( here , delta doped silicon ) and the location of the interface where 2des forms . next , we solve the poisson equation considering open boundary conditions , whereas the surface pattern is defined by gates @xcite , chemical etching @xcite or a combination of these @xcite . the electron density and potential distribution all over the structure is obtained self - consistently utilizing the 4@xmath0 order grid technique and fast fourier transformation . the iteration process is optimized not only by the clever choice of mesh points but also by successive over relaxation . fig [ fig : fig2 ] , depicts the electron density @xmath10 when the quantum dot is induced by trench gates under experimental conditions . the first test that our numerics has to go through is the estimation of the bulk and island electron density , reported experimentally @xcite . we , see that the central electron density is almost the same as reported in the experiment , _ i.e. _ deviates @xmath11 from the bulk density . hence , the full electrostatic treatment of the interference device yields reliable results comparable with the experimental findings . next , we calculate the spatial distribution of the iss considering a strong perpendicular magnetic field . in principle , if one knows the electron distribution at @xmath12 , it is somewhat easy to estimate the positions of the iss . however , as we have discussed it is the widths of the incompressible strips that determines whether interference take place or not . moreover , the analytical formulas given in ref . @xcite to estimate the widths can not be used here , since the potential landscape varies strongly on the scale of wave extend . therefore , we also solve the poisson and schrdinger equations self - consistently in the presence of a strong @xmath2 field , a typical filling factor distribution is shown in fig . [ fig : fig3 ] , we note that the filling factor is nothing but is the electron density normalized by the field strength and is given by @xmath13 . note the similarity between the electron density and the filling factor , the most important difference is that now at the incompressible regions ( black highlighted regions ) the electron density is constant . to investigate interference conditions we also performed calculations at various magnetic field strengths of which we present in fig . [ fig : fig4 ] for selected values . at the highest @xmath2 shown here , the two incompressible strips overlap hence there is no scattering which means we are still in the plateau regime , @xmath14 t. at a lower @xmath2 two well developed iss come close to each other , however , since they are yet incompressible , as illustrated in fig . [ fig : fig1 ] lower right panel , one can not observe interference : scattering is strongly suppressed . the interference oscillations can be observed at @xmath15 t , since the incompressible strips are leaky and also are in close proximity , so that scattering can take place . the lowest magnetic field also prevents ab oscillations , since now the two channels are far appart from each other . this scenario is repeated for the next ( even ) integer filling factor plateau . note that in our calculations we neglected the spin degree of freedom since the formation of the incompressible strips a is general situation given the single particle energy gap . in this work we reported our numerical findings concerning simulations of real interference devices . the electron density and the spatial distribution of the incompressible strips are obtained self - consistently within a mean field approximation at the hartree level . we showed that , the interference can take place only if the incompressible strips become leaky and come close to each other few magnetic lengths , so that scattering ( hence partitioning ) can take place . the actual calculation of the conductance and the ab oscillations is left untouched , however , a recent model is available to obtain these quantities which we currently work on @xcite . the scattering mechanisms and conductivities discussed here are merely phenomenological , therefore it is a great interest for us to perform calculations which also take these quantities into account in a more quantitative manner . k. it is expected that only at @xmath15 t one can observe ab oscillations , the other cases prevent scattering between the edge - states . ] a.s . would like to thank to m. heiblum , n. offek , and l. litvin for the experimental discussions , meanwhile we are also grateful to v.j . goldman for providing us the sample structure and device pattern . f. marquardt and b. rosenow are appreciated for their fruitful and critical theoretical discussions . a partial financial support is given by tubitak under grant no.109t083 and by the feza grsey institute while organizing the 3@xmath16 nano - electronics symposium .	here , we report on our results where self - consistent calculations are performed to investigate the interference conditions , numerically . we employ the successful 4@xmath0 order grid technique to obtain the actual electrostatic quantities of the samples used at the quantum hall based aharonov - bohm interferometers . by knowing the electron density distribution we calculate the spatial distribution of the edge - states , which are considered as mono - energetic current channels . our results are in accord with the experimental findings concerning the electron density distribution . finally , we also comment on the `` optimized '' sample design in which highest visibility oscillations can be measured .
the qcd lattice calculations of the equation of state ( eos ) of deconfined matter have advanced to such a level that reliable results for the pure gluon plasma are available @xcite . for the two - flavor case an estimate of the continuum extrapolation is at disposal @xcite . the eos of four light flavors @xcite are not yet continuum extrapolated . the physically interesting case of two light flavors and a medium - heavy strange quark is still in progress @xcite . all of these _ ab initio _ calculations of the eos of deconfined matter , however , are yet constrained to finite temperatures @xmath0 ( here @xmath1 is the deconfinement temperature ) and vanishing baryo - chemical potential , @xmath2 . while first attempts to calculate the eos at @xmath3 are under way @xcite , the final results can probably be expected only in a few years . it is therefore a challenge to attempt an extrapolation of the qcd lattice results into the domain of finite baryo - chemical potential . here we employ a quasi - particle model to accomplish this goal . quasi - particle models have proven powerful in describing properties of strongly correlated systems in condensed matter physics . analogously , one should expect that also strongly interacting matter can be described effectively within quasi - particle models . indeed , the investigations of @xmath4-derivable self - consistent approximations @xcite , combined with hard thermal loop resummation @xcite , delivered recently compelling support of a quasi - particle description of deconfined matter . starting from the qcd lagrangian a chain of approximations is derived @xcite which results in a quasi - particle model of deconfined matter agreeing with lattice results at @xmath5 . on the other hand , employing the hard thermal / dense loop resummation at finite baryo - chemical potential , further evidence for a quasi - particle description of cold deconfined matter is gained @xcite . what is still needed is an interpolating model , which reproduces the qcd lattice data down to @xmath1 and , at the same time , extrapolates to finite values of @xmath6 even up to @xmath7 . we present here such a model and apply it to calculate static properties of cold , pure quark stars with strange quarks properly taken into account . with increasing sophistication of qcd lattice calculations of the eos also phenomenological quasi - particle models have been developed @xcite . of central importance to our model @xcite are the baryon density @xmath8 and the entropy density @xmath9 as quantities which are dominated by the quasi - particle structure of the interacting system .- derivable approach @xcite where corrections to the quasi - particle picture arise only beyond the resummed leading - loop order contributions @xcite . within the massless @xmath10 theory the form of @xmath9 below is obtained by employing the luttinger - ward theorem @xcite with a super - daisy resummed propagator and a double - ring @xmath4 functional @xcite . ] approximating the self - energies of the quarks ( @xmath11 ) , with a current mass @xmath12 , and the gluons ( @xmath13 ) by the gauge - invariant asymptotic values of the 1-loop expressions @xmath14 g_{\rm eff}^2 , \\ \pi_g^ * & = & \frac16 \left [ \left ( 3 + \frac 12 \ , n_f \right ) t^2 + \frac{3}{2\pi^2 } \sum_q \mu_q^2 \right ] g^2_{\rm eff } , \label{pi } \end{aligned}\ ] ] the densities are given by the standard formulae of ideal gases ( labeled by the superscript i d ) of quarks and gluons with effective masses @xmath15 , @xmath16 , @xmath17 ) - n_{\bar q}^{\rm i d } ( t , \mu ; m_q[t,\mu ] ) \right\ } , \\ s & = & s_g^{\rm id}(t , \mu ; m_g[t,\mu ] ) + \sum_q s_q^{\rm id}(t , \mu ; m_q[t,\mu ] ) . \label{entropy}\end{aligned}\ ] ] beyond this resummation of the leading - order contributions , non - perturbative effects are described in the phenomenological quasi - particle model by the effective coupling @xmath18 . the requirement @xmath19 at large values of @xmath20 and/or @xmath6 ensures the smooth transition to the asymptotic regime . the corresponding pressure @xmath21 and energy density @xmath22 are @xmath23 ) - b(t , \mu)$ ] and @xmath24 ) + b(t , \mu)$ ] . the quantity @xmath25 is not an independent quantity but obtained by integrating @xmath26 @xmath27 , @xmath28 which come from the stationarity condition @xmath29 @xcite . let us mention two implications of the quasi - particle model . ( i ) according to the feynman - hellmann relation the chiral condensate is given by @xmath30 for @xmath31 , i.e. for vanishing current quark masses the chiral condensate vanishes in agreement with the chiral symmetry restoration at @xmath32 . ( ii ) in the asymptotic region , @xmath33 and @xmath34 , an expansion in the coupling yields @xmath35 thus reproducing the perturbative results @xcite in the orders of @xmath36 and @xmath37 . for the effective coupling strength @xmath18 we chose a regularized parameterization of the 1-loop running coupling strength . the resulting comparison with the qcd lattice data is displayed in figure [ comparisons ] for various flavor numbers @xmath38 . notice the almost perfect agreement with the data which ensures that the other thermodynamical quantities are also nicely reproduced ( for details cf . our model assumes the validity of the underlying quasi - particle structure at finite @xmath6 . some limitation of this assumption may arise from the recently debated supra - fluidity and color - flavor locking effects at small temperatures and asymptotically large chemical potentials @xcite . however , since many of the discussed effects @xcite , which also depend sensitively on the actual strange quark mass ( as the phase diagram does already at @xmath2 @xcite ) , are near the fermi surface , the gross properties of the eos might not be modified strongly . since the pressure @xmath21 is a potential it has to fulfill the maxwell relation @xmath40 , which results in a partial differential equation for the coupling of the form @xmath41 -5 mm a solution for the 4-flavor lattice case @xcite is displayed in the left panel of figure [ characteristics ] . the strong increase of @xmath18 towards smaller values of @xmath20 and @xmath6 can be considered as an indication of approaching the phase boundary . despite of the fact that our model does not allow to make statements on the nature of the deconfinement transition , one can safely conclude that the region of smaller values of @xmath20 and @xmath6 , not covered by the elliptic characteristics shown in the right panel of figure [ characteristics ] , may be related to confined matter . the order of magnitude of the baryo - chemical potential @xmath42 for the innermost characteristics at @xmath7 is 1.4 gev when choosing the scale @xmath43 mev . thermodynamic state variables are available in tabular form in @xcite . at small @xmath20 , the pressure in the small-@xmath6 region becomes negative giving a lower bound for the existence of conventional deconfined matter . the energy per baryon , however , at the point of vanishing pressure is quite large , typically in the order of 1.5 gev for @xmath44 . that means , even when including strangeness , no hint to stable strangelets arises from our model . to condense the properties of the predicted eos in a few concise numbers let us consider pure quark stars with locally charge - neutral deconfined matter in @xmath45 equilibrium . the chemical potentials of u , d , s quarks and leptons ( @xmath46 ) are related according to @xmath47 . unfortunately , continuum extrapolated qcd lattice data for the 2 + 1 flavor case with physical quark masses are not yet at disposal at @xmath48 . therefore , we first guess sets of eos for the 2 + 1 flavor case at @xmath2 and then extrapolate them to finite @xmath6 at @xmath7 . in doing so we are guided by the parameter range obtained in @xcite . adding to the pressure the lepton contribution it turns out that the energy density - pressure relation obtained numerically can be approximated by @xmath49 with @xmath50 and @xmath51 mev . note that @xmath52 is not longer an _ ad hoc _ parameter in the spirit of the bag model but a parameter which will be fixed by improving the qcd lattice data for the 2 + 1 flavor case @xcite . integrating the tolman - oppenheimer - volkoff equations for spherical stars in hydrodynamical equilibrium up to vanishing pressure at the surface one gets the masses and radii as displayed in figure [ q_stars ] . since the estimated range of @xmath52 is @xmath53 mev , the strange quark stars have small masses ( @xmath54 ) and radii ( @xmath55 km ) . it should be emphasized that our model does not predict absolutely stable ( strange ) quark matter . rather , at smaller values of @xmath6 and @xmath56 the transition to hadron matter is expected to happen . the static properties of the corresponding hybrid stars depend sensitively on the nature of the confinement transition and details of the hadronic eos . in summary we present a quasi - particle model for the eos of purely deconfined matter , which describes the available qcd lattice calculations at @xmath2 and @xmath0 , and map this to finite values of the baryo - chemical potential including properly the strangeness degree of freedom . the model does not point to stable strangelets . the resulting cold , pure quark stars with strangeness are light and small . at finite temperatures the strange quarks do not play any exceptional role : the self - energies mask nearly all of the finite - mass effect , even having used here @xmath57 mev , while recent lattice calculations point to smaller values @xcite . this is in agreement with qcd lattice calculations which display for @xmath58 almost the same values for the 2 , 3 and 2 + 1 flavor cases @xcite . the gluon abundance near @xmath1 is strongly suppressed . 99 boyd g , engels j , karsch f , laermann e , legeland c , ltgemeier m and petersson b 1996 _ nucl . _ b * 469 * 419 + okamoto m _ et al _ ( cp - pacs collaboration ) 1999 _ phys . _ d * 60 * 094510 karsch f 2000 _ nucl . suppl . _ * 83 - 84 * 14 engels j , joswig r , karsch , laermann e , ltgemeier m and petersson b 1997 _ phys . _ b * 396 * 210 karsch f , laermann e and peikert a 2000 _ phys . _ b * 478 * 447 engels j , kaczmarek o , karsch f and laermann e 1999 _ nucl . _ b * 558 * 307 baym g 1962 _ phys . rev . _ * 127 * 1391 andersen j o , braaten e and strickland m 2000 _ phys . rev . _ d * 61 * 074016 , _ phys . _ d * 61 * 014017 , 1999 _ phys . _ * 83 * 2139 blaizot j p , iancu e and rebhan a 1999 _ phys . _ b * 470 * 181 , _ phys . rev . lett . _ * 83 * 2906 , hep - ph/0005003 baier r and redlich k 2000 _ phys . lett . _ * 84 * 2100 biro t s , levai p and mller b 1990 _ phys . _ d * 42 * 3078 + goloviznin v and satz h 1993 _ z. phys . _ c * 57 * 671 + peshier a , kmpfer b , pavlenko o p and soff g 1994 _ phys . lett . _ b * 337 * 235 + peshier a , kmpfer b , pavlenko o p and soff g 1996 _ phys . rev . _ d * 54 * 2399 + schertler k , greiner c and thoma m h 1997 _ nucl . phys . _ a * 616 * 659 + levai p and heinz u 1998 _ phys . rev . _ c * 57 * 1879 peshier a , kmpfer b and soff g 2000 _ phys . _ c * 61 * 045203 lee t d and yang c n 1960 _ phys . rev . _ * 117 * 22 freedman b a and mclerran l d 1977 _ phys . rev . _ d * 16 * 1130 , 1147 , 1169 + kapusta j i 1989 _ finite - temperature field theory _ cambridge university press + zhai c and kastening b 1995 _ phys . _ d * 52 * 7232 riedel e 1968 _ z. phys . _ * 210 * 403 + vanderheyden b and baym g 1998 _ j. stat . * 93 * 843 luttinger j m and ward j c 1960 _ phys . rev . _ * 118 * 1417 peshier a , kmpfer b , soff g and pavlenko o p 1998 _ europhys . _ * 43 * 381 rajagopal k 1999 _ nucl . phys . _ a * 661 * 150c + alford m , rajagopal k and wilczek f 1999 _ nucl . _ b * 537 * 443 + berges j and rajagopal k 1999 _ nucl . _ b * 538 * 215 peikert a , karsch f , laermann e and sturm b 1999 _ nucl . suppl . _ * 73 * 468 www.fz-rossendorf.de/fwk/mitarb/peshier/eos.html garden j _ et al _ ( alpha and ukqcd collaboration ) 2000 _ nucl . _ b * 571 * 237	a quasi - particle model is presented which describes qcd lattice results for the 0 , 2 and 4 quark - flavor equation of state . the results are mapped to finite baryo - chemical potentials . as an application of the model we make a prediction of deconfined matter with appropriate inclusion of strange quarks and consider pure quark stars .
the electron dephasing time @xmath0 is a very important quantity that governs the mesoscopic phenomena at low temperatures . particularly , the behavior of the dephasing time near zero temperature , @xmath2 , has recently attracted vigorous experimental @xcite and theoretical @xcite attention . one of the central themes of this renewed interest is concerned with whether @xmath3 should reach a finite or an infinite value as @xmath4 0 . the connection of the zero - temperature dephasing behavior with the very fundamental condensed matter physics problems such as the validity of the fermi - liquid picture , the possibility of the occurrence of a quantum phase transition , and the persistent currents in metals , etc . , has been addressed . conventionally , it is accepted that @xmath3 should reach an infinite value if there exist only the inelastic electron - electron and electron - phonon scattering . however , several recent measurements performed on different mesoscopic conductors have revealed that @xmath3 depends only very weakly on @xmath5 , if at all , when @xmath5 is sufficiently low . there is no generally accepted process of electron low - energy - excitation interactions that can satisfactorily explain the saturation " of @xmath3 found in the experiments . it should be noted that those experiments @xcite have ruled out electron heating , external microwave noises , and very dilute magnetic impurities as the origins for the observed finite dephasing as @xmath6 . to unravel the issue of electron dephasing , systematic information of @xmath3 over a wide range of sample properties is very desirable . bearing this in mind , we have in this work performed systematic measurements of @xmath3 on several series of _ as - sputtered _ and subsequently _ annealed _ aupd and sb _ thick _ films . the low - field magnetoresistances of the as - sputtered samples are first measured . the samples are then annealed , and their magnetoresistances measured . the annealing and magnetoresistance measurement procedures are repeated a few times . @xmath0 is extracted by comparing the measured magnetoresistances with the three - dimensional ( 3d ) weak - localization ( wl ) theoretical predictions @xcite . generally , thermal annealing causes a decrease in the sample resistivity , signifying a reduction in the amount of defects in the microstructures . controlled annealing measurements are thus crucial for testing the theoretical models of electron dephasing invoking magnetic impurities and dynamical defects @xcite . thick film samples were prepared by dc sputtering deposition onto glass substrates held at room temperature . the deposition rate was varied to tune the amount of disorder , i.e. , the residual resistivity @xmath7 (= @xmath8(10 k ) ) of the films . the aupd films were typically 6000 @xmath9 @xmath10 0.3 mm @xmath10 17 mm , while the sb films were typically 3000 @xmath9 @xmath10 0.3 mm @xmath10 17 mm . thermal annealing of the aupd ( sb ) films was performed in a 99.999% pure ar atmosphere at moderate temperatures of @xmath11 100@xmath12300@xmath13c ( @xmath11 150@xmath13c ) for about one half to several hours until @xmath7 changed by a desirable amount . the use of an extremely high purity ar atmosphere greatly minimized the presence of any oxygen residual gas in the annealing . the values of the relevant parameters for our as - sputtered films are listed in table [ t.1 ] . .resistivities , diffusion constants , and dephasing times of the as - sputtered aupd and sb thick films . the value of @xmath3 is extracted by least - squares fitting the measured @xmath14 to eq . @xmath15 . [ cols="<,^,^,^,^",options="header " , ] [ t.1 ] we notice that the four aupd films listed in table [ t.1 ] were _ newly _ made from a _ new _ au@xmath16pd@xmath16 target different from that used in our previous study @xcite . moreover , a _ different _ sputtering gun and a different vacuum chamber were utilized . previously , we had studied @xmath0 in a series of dc sputtered aupd thick films prepared and measured in a different laboratory @xcite . by applying these new samples , we are able to perform a close comparison study of @xmath0 in the same material prepared under different conditions . such a comparison is indispensable for clarifying the possible role , if at all , of magnetic scattering on @xmath3 . if there were any noticeable magnetic contamination during this experiment , it is natural to expect an _ unintentional _ magnetic concentration , @xmath17 , that differs from that in our previously samples @xcite . consequently , a distinct value of @xmath3 should be observed . on the other hand , if a similar value of @xmath3 is measured , this result must then bear important information about an intrinsic material property . in addition to the newly prepared aupd samples , we have studied two aged " sb films . the two sb films listed in table [ t.1 ] were first deposited and studied two years ago in ref . @xcite . during the past two years , they were exposed to air all the time . one might have naively speculated that these two samples must be heavily contaminated by ( magnetic ) impurities and , thus , have a shorter @xmath0 with a much weaker @xmath5 dependence , compared with that measured two years ago . to the contrary , this experiment indicates that 3d aupd and sb are _ not _ as vulnerable to contamination as speculated . our results point to an experimental fact that suggests that the observed saturation of @xmath3 can not be readily explained by magnetic scattering . we have measured the magnetoresistances and compared with 3d wl predictions @xcite to extract the values of @xmath0 . our experimental method and data analysis procedure had been discussed previously @xcite . here we emphasize that , in the limit of strong spin - orbit scattering ( which applies for both aupd and sb ) , @xmath0 is the _ only _ adjusting parameter in the least - squares fits of the measured magnetoresistances with wl predictions . this great advantage makes the extraction of @xmath0 highly reliable . empirically , @xmath0 can be written in the form @xmath18 where @xmath3 dominates at the lowest measurement temperatures , and @xmath19 is the relevant inelastic scattering time which is usually important at a ( few ) degree(s ) kelvin and higher . in three dimensions , electron - phonon scattering is the predominant inelastic process while the nyquist electron - electron scattering is negligibly small @xcite , i.e. @xmath20 in eq . the electron - phonon scattering rate @xmath21 varies as @xmath22 , with 2 @xmath23 4 . figure [ f.1](a ) shows our measured @xmath0 as a function of temperature for four as - sputtered aupd films . this figure demonstrates that @xmath0 first increases with decreasing @xmath5 at a few degrees kelvin , where the electron - phonon scattering dominates the total dephasing and @xmath24 in aupd @xcite . below about 2 k , the inelastic process is much less effective and a new mechanism progressively takes over , resulting in a very weak temperature dependence of @xmath0 as @xmath25 . to our knowledge , there is no generally accepted process of electron low - energy - excitation interactions that can account for such a weak @xmath5 behavior . a weak temperature dependence of @xmath3 is also observed in the two sb thick films listed in table [ t.1 ] . in fact , an ( almost ) absence of temperature dependence of @xmath0 has previously been found in numerous three - dimensional polycrystalline metals @xcite . we notice that the values of @xmath3 in fig . [ f.1](a ) follow _ closely _ the scaling relation of @xmath26 s established in fig . 3 of @xcite , where the diffusion constant @xmath27 is in @xmath28/s . this result suggests that the behavior of @xmath0 found in fig . [ f.1](a ) is material intrinsic . it should be emphasized that the weak @xmath5 dependence or the so - called saturation " of @xmath3 in fig . [ f.1](a ) is observed in a temperature regime where the sample resistance varies as @xmath29 all the way down to our lowest measurement temperatures ( fig . [ f.1](b ) ) . this @xmath29 dependence of resistance is well described by the 3d electron - electron interaction effects @xcite . this result indicates that the saturation of @xmath3 is _ not _ caused by electron heating . a similar assertion of non - hot - electron effects has also been reached in previous experiments @xcite . information about the effect of annealing is crucial for clarifying the nature of magnetic scattering and dynamical defects . figure [ f.2](a ) shows the variation of @xmath0 with temperature for the as - prepared and subsequently annealed aupd1e thick film . this figure clearly indicates that @xmath0 increases with annealing . similar behavior of increasing @xmath0 with annealing has also been found in the as - prepared and annealed aupd6e thick film . at first glance , this observation is easily explained . suppose that annealing results in the rearrangement of lattice atoms and relaxation of grain boundaries and , hence , makes the film less disordered . because two - level systems ( tls ) are closely associated with the presence of defects in the microstructures , their number concentration would be reduced by annealing . by assuming that dynamical defects are effective scatterers , one can then understand fig . [ f.2](a ) in terms of a reducing tls picture . however , our further measurements indicate that the nature of low - temperature dephasing in real metals is not so straightforward . we find that the effect of annealing on @xmath0 is distinctly different in strongly disordered samples ( fig . [ f.3 ] ) . for the convenience of discussion , the aupd1e and aupd6e ( aupd4a and aupd5e ) thick films are referred to as moderately ( strongly ) disordered , because they have @xmath30 120 ( 470 ) @xmath31 cm before annealing . figure [ f.2](b ) shows the variation of @xmath0 with @xmath5 for the as - prepared and annealed sb01b thick film . this figure clearly indicates that @xmath0 increases with annealing . similar effect of annealing has also been found in the sb12 thick film . the results of figs . [ f.2](a ) and [ f.2](b ) suggest that an enhancement of @xmath0 by thermal annealing is common to different _ moderately _ disordered metals . ( we notice that the high resistivities in sb films arise from a low carrier concentration instead of a short electron mean free path @xcite . our sb thick films are thus moderately disordered . ) one of the widely accepted explanations for the saturation " behavior of @xmath3 invokes magnetic spin - spin scattering due to a low level contamination of the sample . this explanation has been challenged in several recent careful experiments @xcite . however , despite this experimental situation , there is still an insisting opinion that argues for non - zero magnetic scattering in the sample . to completely reject such an opinion is non - trivial , because it is argued that the level of unintentional contamination is so low that it can not be detected by the state - of - the - art material analysis techniques . the situation becomes more serious when reduced - dimensional systems are involved . in the case of low - dimensional structures , surface effects due to interfaces , substrates , and paramagnetic oxidation are non - negligible . therefore , it is not straightforward to ascribe the observed saturation of @xmath3 to either intrinsic material properties or surface / interface effects . on the other hand , this kind of ambiguity does _ not _ occur in our _ three - dimensional _ measurements . in fact , we believe that magnetic scattering can at most play a subdominant role in our experiment . our reasons are given as follows . ( i ) suppose that there is a low level of magnetic contamination in our as - sputtered films . upon annealing , the magnetic impurity concentration @xmath17 should be left unchanged . if the original saturation " in our as - sputtered samples is caused by spin - spin scattering , one should then expect the same value of @xmath3 ( @xmath32 ) after annealing . however , we find increasing @xmath3 with annealing . our result is thus in disagreement with this assumption . ( ii ) our argument for a non - magnetic origin is supported by the observation of an _ increased _ @xmath3 in the aged and annealed sb films . since our sb01b and sb12 thick films were aged in air for two years , one might have naively expected a large decrease in @xmath3 due to magnetic contamination . nevertheless , this is not the case found in fig . [ f.2](b ) . ( iii ) moreover , if our samples do contain an appreciable level of unintentional magnetic impurities , the contaminated concentration @xmath17 should be basically the same in all films , because similar fabrication and measurement procedures were involved . one should then expect a similar @xmath3 in all _ as - prepared _ samples , regardless of disorder . this is certainly inconsistent with the observed scaling relation @xmath33 discussed above . therefore , magnetic scattering in its current form can not easily explain our overall results in a consistent manner . in order to explain the widely observed saturation behavior of @xmath3 , it has recently been proposed that dynamical defects can be important @xcite . the low - energy excitations of the dynamical defects are usually modelled by two - level systems . we already discussed that tls might be partly responsible for the saturated dephasing found in our moderately disordered films . however , it is impossible to perform a quantitative comparison of our experiment with the tls theories . the difficulties lie on the facts that ( i ) the number concentration of tls in a particular sample is not known , ( ii ) the strength of coupling between conduction electrons and a tls is poorly understood , and ( iii ) the dynamical properties of real defects ( impurities , grain boundaries , etc . ) are even less clear . experimentally , we also find other features of thermal annealing ( fig . [ f.3 ] ) that seem incompatible with a tls picture of dephasing . in addition to the moderately disordered samples , we have performed measurements on thick films containing much higher levels of disorder . surprisingly , we discover that the effect of annealing is completely different . in the _ strongly _ disordered aupd4a and aupd5e thick films , we find that annealing causes _ negligible _ effect on @xmath0 . figure [ f.3 ] shows the variation of @xmath0 with @xmath5 for the as - prepared and annealed aupd4a thick film . this figure clearly demonstrates that the values of @xmath0 for the as - prepared and annealed samples are essentially the same , even though the resistance , and hence diffusion constant @xmath27 changed by a factor of more than 6 . the absence of an appreciable annealing effect in this case implies that , in addition to the usual tls addressed above , these two films contain other defects that can not be readily cured by thermal annealing . such a null effect of annealing seems to suggest that , despite a large effort in this direction , no real defects of any nature can be found to have dynamical properties that may explain the saturation behavior of @xmath3 . a comparison of figs . [ f.2 ] and [ f.3 ] strongly indicates that low - temperature dephasing is very sensitive to the microstructures . the observation of fig . [ f.3 ] deserves further discussion . first , we recall that our measured @xmath0 in the as - sputtered , strongly ( and moderately ) disordered films follows the scaling relation of @xmath33 mentioned above @xcite , implying that the result of fig . [ f.3 ] is material intrinsic . secondly , in the context of magnetic scattering , blachly and giordano recently found that kondo effect is very _ sensitive _ to disorder , namely , an increase in disorder suppresses kondo effect @xcite . along this line , if the original saturated @xmath3 found in fig . [ f.3 ] were due to magnetic scattering , one should argue that thermal annealing that suppresses disorder should enhance kondo effect . then , a decreased @xmath3 should be expected with annealing . since our @xmath3 does not change even when the sample resistivity is reduced by a factor of more than 6 by annealing , fig . [ f.3 ] thus can not be easily reconciled with a magnetic scattering scenario . besides , this picture of a weakened kondo effect by disorder is also incompatible with our result for the moderately disordered films ( fig . [ f.2 ] ) where an increased , instead of a decreased , @xmath3 with annealing is found . thirdly , one might argue that annealing in the presence of oxygen can lead to the oxidation of magnetic impurities , and hence to the disappearance of kondo resistivity @xcite . we argue that this is unlikely in our case , since we had taken precautions by annealing our films in a 99.999% pure ar atmosphere , where the amount of oxygen residual gas was greatly minimized . moreover , since we do not find a rapid increase in the @xmath5 dependence of @xmath3 after annealing , there is thus no clear evidence of the presence of oxygen or magnetic impurities in our films . in any case , it would be interesting to study if annealing in oxygen would have a drastic effect on @xmath3 . lastly , we discuss the advantage of using three- , instead of lower - dimensional , mesoscopic structures for @xmath0 measurements . in 3d , the dominating inelastic process is the electron - phonon scattering for which @xmath34 obeys a strong @xmath35 ( 2 @xmath23 4 ) dependence @xcite . such a temperature variation is much stronger than the dominating @xmath36 in one dimension and @xmath37 in two dimensions ( which are both due to electron - electron scattering @xcite ) . inspection of the solid lines , which are drawn proportional to @xmath38 , in figs . [ f.1 ] and [ f.3 ] reveals that our measured @xmath3 at 0.3 k is already @xmath11 _ two orders of magnitude _ lower than that as would be extrapolated from the measured @xmath34 at a few degrees kelvin . such a huge discrepancy is well outside any experimental uncertainties . on the contrary , in the case of narrow wires , the dominating inelastic dephasing time obeys a much weaker @xmath39 law just mentioned . in this case , any discrepancy between the measured and extrapolated values of @xmath3 would be less dramatic in the attainable experimental temperature range , rendering a discrimination of the presence or absence of a saturated @xmath3 less clear - cut . we have studied the influence of thermal annealing on low - temperature electron dephasing in polycrystalline aupd and sb thick films . we find that @xmath3 reveals an extremely weak temperature dependence in both as - sputtered and annealed samples . the effect of annealing is non - universal , depending strongly on the amount of disorder quenched in the microstructures during deposition . the observed saturation behavior of @xmath3 can not be easily explained by magnetic - scattering in its current form . we also find that the disorder behavior of @xmath3 in as - prepared and annealed samples is very different . a complete theoretical explanation would need to take the microstructures into account . the authors are grateful to m. bttiker , c. s. chu , p. mohanty , r. rosenbaum , j. von delft , g. y. wu , and a. zawadowski for valuable discussion . this work was supported by the taiwan nsc through grant no . nsc 89 - 2112-m-009 - 054 . mohanty p. _ lett . * 78 * 3366 , 1997 ; webb r. a. _ et al . _ , fortschr . phys . * 46 * 779 , 1998 ; webb r. a. _ et al . _ , in _ quantum coherence and decoherence _ , ed . by k. fujikawa and y. a. ono ( elsevier , amsterdam , 1999 )	we have studied the effect of thermal annealing on electron dephasing times @xmath0 in three - dimensional polycrystalline metals . measurements are performed on as - sputtered and annealed aupd and sb thick films , using weak - localization method . in all samples , we find that @xmath0 possesses an extremely weak temperature dependence as @xmath1 . our results show that the effect of annealing is non - universal , and it depends strongly on the amount of disorder quenched in the microstructures during deposition . the observed saturation " behavior of @xmath0 can not be easily explained by magnetic scattering . we suggest that the issue of saturation can be better addressed in three - dimensional , rather than lower - dimensional , structures .
in broad terms , there are two very different ways in which galaxies could have formed . the first , which we shall refer to as the monolithic formation idea , is that a large cloud of primordial gas collapsed rapidly to produce either an elliptical or lenticular galaxy , or perhaps the spheroidal bulge of a spiral galaxy @xcite . the main advantage of monolithic formation is that it seems consistent with the observational facts that most such systems have relatively little angular momentum and that they contain mostly old stars . the second way galaxies might have formed is through the hierarchical accumulation of smaller clumps of gas and stars into larger structures that became the galaxies we see today @xcite . within that view , there would be a continuum of mergers between clumps of varying masses and angular momenta . some mergers would lead to bursts of massive star formation , while others would not . galaxy formation would be drawn out over a long period , perhaps continuing to the present epoch , where we still see mergers and violent interactions between galaxies , along with the possible formation of tidal dwarfs . the red colours of today s e and s0 galaxies would be explained if most of the stars were produced in the original merger participants . determining the observational properties of the different classes of galaxies offers a way to discriminate between the possibilities . until the iras database was available , the conventional wisdom was that elliptical and lenticular galaxies had little or no gaseous interstellar medium this was reinforced by the general paucity of hi emission . for example , @xcite and @xcite find that most s0s have @xmath1 in solar units , but with large scatter , which is 3 - 6 times lower than spirals @xcite . the surveys of atomic gas in early - type galaxies would of course miss the molecular content , which we now know dominates the ism near the centers of spirals . the iras results motivated a few surveys of co emission ( @xcite ; @xcite ; ( * ? ? ? * hereafter wh ) ; @xcite ; @xcite ) . while the co surveys were much less complete in terms of sample size than those for hi , the general conclusion was that in the iras - selected galaxies the interstellar medium at least near the centers was not too different from that in normal , isolated spirals . comparison of the molecular and far - infrared properties showed that the process of star formation in lenticulars is broadly similar to that in field spirals . it seems unlikely , however , that the gaseous contents of the galaxies selected using far - infrared criteria are representative of the bulk of lenticulars , because most normal s0s have little or no far - infrared emission . @xcite looked at 10 galaxies lacking optical emission lines , to investigate a more typical sample . their detection of six sources led them to suggest that most normal s0s contain significant amounts of cool , dense gas , which for unexplained reasons is not producing massive new stars . although s0s are dominated by @xmath2 year old stars , their low gas content can not merely be due to early , very efficient astration , where the gas is simply locked into stars or blown out by winds from massive stars and supernovae produced during a huge initial starburst . in the simplest ( closed box ) view of galaxy evolution , one expects to find gas that has subsequently been returned to the ism by evolving low - mass stars , and the latter contribution should be very large . in general , the total mass of returned gas m@xmath3 scales with the present blue luminosity as m@xmath3 = kl@xmath4 , with both quantities in solar units . * hereafter fg ) adopt a reasonable return rate for a 10 gyr old population , that is , the product of an assumed planetary nebula formation rate and mass loss by stars near 1 @xmath5 . assuming the rate to be constant over the galaxy lifetime implies that recycled gas alone should presently give k@xmath6 . more complex models of gas return within the monolithic galaxy evolution scenario ( e.g. @xcite ) predict values more than an order of magnitude larger , owing to the early contributions of massive stars during and after the spike of initial star formation . in summary , there is very little understanding of how much gas is present in field s0s , of its properties , and of how these properties might have arisen . to improve this situation , and we hope thereby our understanding of what s0 galaxies are , has been the primary motivation for this study . the initial sample comprises all entries in the nearby galaxies catalog @xcite having type -3 to 0 inclusive , declination @xmath7 , and distance @xmath8 20 mpc . we have omitted known peculiar galaxies , cases of probable interaction ( based on inspection of the field on poss prints ) , members of the virgo cluster , and objects assigned class sa in the revised shapley ames catalog ( * ? ? ? * rsac ) or the carnegie atlas @xcite . the final sample of 27 galaxies is listed in table 1 . all but six of our sample are contained in the rsac . most ( 24/27 ) are classified as either s0 , s0/sa , or as their barred counterparts . our final sample is free of the fir bias inherent in earlier s0 surveys , but still contains the biases of the nearby galaxies catalog . the most obvious of these are that optically small galaxies ( diameter @xmath8 1.5 @xmath9 -2 @xmath9 on the poss ) are excluded , and hi - poor systems are under - represented . nonetheless , we believe that table 1 constitutes the best sample now available for probing the cool ism in normal s0 galaxies within low - density environments . we report here on a single - antenna survey of co in these galaxies ; a complementary hi study is underway . in addition to its lack of fir bias , the present work departs in two important ways from published ones , which have been essentially flux - limited , and have usually looked only at the center of the galaxy ( see , however , * ? ? ? * ) . first , we define a target sensitivity for each galaxy using the faber - gallagher relation for an assumed age of 10 gyr to compute an expected ism mass , @xmath10 . the goal is to provide a 5-sigma detection , or upper limit , of 0.02@xmath10 per pointing . since @xmath10 scales with the total blue luminosity , long integrations are needed for faint galaxies . secondly , we observe additional points ( along the major axis ) in the larger galaxies , even if no emission is detected at the center , as there are several early - type galaxies with molecular gas outside their centers @xcite . spectra of the j=2 - 1 and j=1 - 0 transitions of co in 20 galaxies were obtained in october 1998 with the iram 30 meter telescope . the nrao 12 meter telescope was used in november 1998 , january 2001 , and in february and april 2002 to observe the remainder of the sample in the j=1 - 0 transition , and also the galaxies having the largest angular sizes . the iram telescope main beam is assumed to have fwhm = 20.9@xmath11 and 10.4@xmath11 in the two transitions , whereas we take fwhm = 55@xmath11 for the 12 meter telescope at co ( j=1 - 0 ) . sufficiently sensitive co observations have been published for ngc 404 , 7013 , and 7077 ( wh ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , so we have not observed these galaxies . the iram observations employed beam switching in azimuth with a throw of 240@xmath11 at 1 hz . dual polarization measurements were obtained simultaneously in both transitions , and directed both to filterbanks and autocorrelator backends . the filterbanks provided spectral windows 1332 km s@xmath12 and 666 km s@xmath12 wide for j=1 - 0 and j=2 - 1 , respectively , with resolutions of 2.60 km s@xmath12 and 1.30 km s@xmath12 wide , respectively . corresponding values for the autocorrelators were 1326 km s@xmath12 and 666 km s@xmath12 , and 3.25 km s@xmath12 and 1.63 km s@xmath12 , respectively . telescope pointing and focus were checked about every 2 hours using a bright quasar or the planet mars . over the observing session , median corrections to the pointing model in azimuth and altitude were 0.9@xmath11 and 1.4@xmath11 , respectively . system temperatures were typically 300 - 400 k for j=1 - 0 and 600 - 700 k for j=2 - 1 . an initial 30 m spectrum was obtained at the center of the target . additional positions along the major axis were observed in 10 arcsecond steps , unless either no emission was seen or it appeared that the j=2 - 1 transition was at least as strong as j=1 - 0 . because previous observers have generally reported and i@xmath13 represent the integrated intensities ( area under the line ) of the two co transitions ] i@xmath14/i@xmath15 1 ( see below ) , the latter case was assumed to result from beam filling effects produced by a source smaller than the j=2 - 1 beam . the large beam of the 12 m telescope is useful in searching for extended gas in larger galaxies . beam switching in azimuth was also employed at that telescope , using a throw of 180@xmath11 . filterbanks collected data in both polarizations , providing windows 1330 km s@xmath12 wide and individual channel widths of 5.2 km s@xmath12 . system temperatures were usually 400 - 500 k. pointing and focus checks were made at roughly 2-hour intervals , using mars , saturn , jupiter , and several bright quasars . galaxies were observed on center and at 55 arcsecond steps along the major axis . the observations typically encompass @xmath0 0.5d@xmath16 and @xmath0 0.1d@xmath16 along the major axis , for the 12 m and 30 m telescopes , respectively , making the present survey unique for its extended spatial coverage . the class software package was used for data reduction . occasional bad channels were first removed in the individual raw scans , which represented 4 and 6 minute integrations for the 30 m and 12 m telescopes , respectively . all scans from a given pointing were then summed and hanning smoothed to typical resolutions of 13.0 and 20.8 km s@xmath12 for the 30 m and 12 m telescopes , respectively . other resolutions are noted in tables 2 and 3 . line windows were chosen as discussed below or in footnotes to tables 2 and 3 , and linear baselines fitted to the remainder of the spectrum . second order baselines were subtracted in a few cases , as noted in the tables . data were obtained on the main beam temperature scale from the 30 m telescope , and on the t@xmath17 scale at kitt peak . the latter was transformed to t@xmath18 using a corrected main beam efficiency of @xmath19 = 0.82 . when emission was not clearly detected , we defined spectral - line windows for upper limit estimates using published kinematics of atomic gas ; otherwise , the optical rotation curve was employed . in only two cases , ngc 4143 and ngc 4346 , were we unable to find any guiding observations ; for these galaxies we used windows 400 km s@xmath12 wide , centered on the optical velocity , as that represents the linewidth of a typical galaxy in our sample . the final summed spectrum at each center pointing is presented in figure 1 . integrated intensities on the t@xmath18 scale from 30 m and 12 m observations appear in tables 2 and 3 , respectively . standard deviations , which serve as 1-sigma upper limits for non - detections , are calculated following the prescription of @xcite . the rms noise in the binned spectrum , required for that procedure , is given in the table . observed positions are identified by specifying the offsets in arcseconds from the coordinates given in table 1 . excluding objects already known to contain co ( ngc 404 , 7013 , 7077 ) , we have detected 75 percent ( 18/24 ) of our sample with a formal confidence of 3@xmath20 or higher , 42 percent ( 10/24 ) with 5@xmath20 or higher . the detection rate for the entire sample is 78 percent . only the j=2 - 1 transition was detected in ngc 2787 , 3115 , 3384 , and 3990 ; although formally significant at 3.6@xmath20 and 4.8@xmath20 , respectively , the lines in ngc 3115 and ngc 3990 are complex and not convincing . emission ( at 4.9@xmath20 ) from ngc 4026 was detected only by the 12 m telescope at offset ( 2 , -55 ) . in the following summary , we include these galaxies among our detections ; better data are clearly needed for them . within our admittedly small sample of 27 objects , types s0 and s0/a are detected with similar frequency ( 11/15 and 6/8 , respectively ) . somewhat surprisingly , s0s with optical dust features and those without are also detected at about the same rate : 7/10 galaxies classified s0@xmath21 or s0@xmath22 ( or as their barred counterparts ) are detected , and 4/6 of those classified as s0@xmath23 , s0@xmath24 or s0@xmath25 . finally , sb0 galaxies are somewhat less likely to be detected ( 2/4 ) than s0 galaxies ( 8/10 ) , though high inclinations might obscure bars in several of the latter . in summary , except for a possible anti correlation with the presence of a bar , our sample does not show a relation between the gross optical properties of s0 galaxies and their detectability in co emission . the present observations can be used to estimate typical source sizes . the iram 30 m data show that i@xmath14/i@xmath261 in 7 galaxies where only the centre pointing was observed . even when the gas is quiescent , large intensity ratios can result when the source size is comparable to or smaller than the j=2 - 1 beam . assuming that to be the case , we derive a median linear size of 0.7 kpc for those 7 galaxies . an additional 7 galaxies have multiple - pointing 30 m data showing strong radial intensity declines which indicate that we have seen most of the total co emission . ( this conclusion is less clear for ngc 4460 ; nonetheless , we include this galaxy in the discussion ) . we estimate the source size in those cases by summing the maximum separation between pointings with detections and adding the fwhm of the appropriate telescope beam , finding a median of 2.8 kpc , or 0.32d@xmath16 . therefore , our work reinforces the view ( e.g. * ? ? ? * ) that most of the molecular gas in s0 galaxies is near their centres . line ratios ( i.e. ratios of integrated intensities in two transitions ) are used to investigate the physical state of molecular gas , though the interpretation is complicated because of the different ways the telescope beams couple to the source , and by the inevitable mixing of emission from dense cloud cores and more diffuse surrounding molecular gas . the few published multi - transition co studies of early - type galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) report 0.5 @xmath27i@xmath14/i@xmath281 , which is assumed to indicate the gas is cool ( t@xmath2910k ) , optically thick and generally quiescent . without maps of our sources we can not correct for different beam coupling . instead , we compare in figure 2 the 30 m telescope results with predictions which assume the same radiation temperature in both transitions , and that the source is either a point ( top line ) or that it uniformly fills both beams . the factor of 4 vertical difference between the lines is therefore the ratio of beam areas . the results shown in the figure are broadly consistent with previous studies . total h@xmath23 masses for all detected galaxies , or their upper limits , are presented in table 4 . whenever possible we use the relation given by @xcite with a conversion constant of 2.3 @xmath30 10@xmath31 @xmath32(k km s@xmath12)@xmath12 . uncertainties and upper limits are obtained by adding in quadrature the appropriate uncertainties from individual pointings . molecular masses for ngc 2787 , 3115 , 3384 , 3990 are estimated from the j=2 - 1 line intensities , assuming i@xmath14/i@xmath13=0.5 because , except perhaps for ngc 2787 , no j=1 - 0 emission is seen . we transform to our assumptions the published mass estimates for 6 galaxies observed by others and present them in table 5 . values obtained by ourselves from the 30 m and 12 m telescopes show pleasing consistency , which is generally present throughout the results contained in tables 2 - 4 . agreement with other studies is usually satisfactory , though the values reported by the nrao 12 m and nobeyama 45 m telescopes for ngc 7457 are the inverse of what one might expect from comparing the two beam sizes . another noteworthy disparity is the factor of 2 difference between 30 m observations towards the centre of ngc 5866 . most of the disagreement comes from the integrated intensities , which we believe are both on the t@xmath18 scale but which differ by a factor of 1.6 . inspection of table 2 shows that i@xmath13 changes by large amounts between positions separated by only @xmath00.5@xmath30fwhm or @xmath010@xmath11 near the centre of the galaxy , leading us to suspect that telescope pointing is the culprit , as previously suggested by @xcite for a different set of galaxies . in table 6 and figure 3 we present estimates of the total cool gas content of galaxies in our sample , obtained by combining molecular and atomic hydrogen masses from table 4 and from the literature . the table also shows the predicted mass of gas returned by stars according to the analytical expression of @xcite , i.e. the case k = 1.2 as discussed below . together , table 6 and figure 3 encapsulate the major results of our study . we now turn to the question of whether they support current ideas of the nature of lenticular galaxies . it has generally been thought that s0 galaxies formed most of their stars more or less monolithically 10 gyr or more ago , because of their typically red colours and late - type stellar spectra ( e.g. @xcite ) . that view has been reinforced by the paucity of atomic hydrogen @xcite , relatively low levels of far - infrared emission @xcite , and by the almost complete absence of optical dust lanes @xcite and hii regions @xcite associated with star formation in spiral galaxies . in the monolithic paradigm , a large cloud of primordial gas collapsed fairly rapidly to produce the galaxy , which has evolved passively since then . more rapid collapses produced elliptical galaxies , in which disks are insignificant ; less rapid collapses led to gas - rich spirals @xcite . but the simple monolithic picture has come under increasing pressure over the last decade , because the predicted bright , massive protogalaxies that would be the precursors of today s ellipticals have not been observed at any redshift or spectral band in the necessary numbers @xcite , although the nature of the unidentified galaxies seen through their sub - millimeter emission remains unclear @xcite . those deficiencies have stimulated investigations of the hierarchical formation idea , whereby galaxies are assembled from smaller pieces through a series of mergers @xcite . we initially hoped that our data would reveal evidence to prefer one path over the other . but as we show below , the most startling observational result remains difficult to understand within either picture . our most significant result is contained in figure 3 , which displays the total mass of atomic and molecular gas versus blue luminosity . the wide range in gas content among galaxies of similar luminosity is immediately evident . the most striking aspect , though , is the upper mass cutoff , which extends over the entire luminosity range spanned by our data , a factor of @xmath33 in luminosity and presumably in total mass . within that range , no galaxy contains more than about 10 percent of the gas predicted by recent estimates of stellar mass return ( see below ) , shown as a solid line in the figure . stated differently , even the most gas - rich s0s have lost 90 percent or more of all the gas ever returned by their stars . we find ourselves unable to use the data to distinguish clearly between the two evolutionary paths . as we outline below , with a judicious tuning of parameters it is possible to explain aspects of our results within either evolutionary scenario , but in both cases the tuning is contrived and the correspondence with the data incomplete . perhaps the difficulty arises from an assumption underlying our work - that all s0 galaxies have evolved along very similar paths . van den bergh ( 1990,1994 ) suggests that lenticulars comprise at least two distinct kinds of objects , one of these being the true s0s , i.e. galaxies forming the transition between ellipticals and spirals . although it is not yet possible to identify the pedigree of an individual galaxy , van den bergh suggests that true s0s are hiding among brighter , rounder galaxies , and that they tend to have @xcite sub - type i , i / s , or s. we have tried to test that suggestion by searching for some trend in optical properties among the galaxies in our sample . we have asked , for example , if the upper cutoff might be delineated by true s0s , but have not found any evidence to support the idea . while a wide range of gas content is certainly consistent with the view of s0s as a heterogeneous collection , the upper cutoff probably is not . in the following sections we attempt to lay out the issues for the two galaxy formation paradigms , without providing a coherent explanation for our results . faber and gallagher realized that any isolated , passively evolving galaxy which converted almost all of its gas into stars early in its existence would over a hubble time build up an embarrassingly massive ism , unless most of the returned gas could be disposed of . the simple picture of stellar mass return used by fg has been extensively revised , most recently by @xcite , and @xcite , without qualitatively altering the original conclusion . we remind the reader of the model predictions that the mass of returned gas is given by m@xmath3 = kl@xmath4 , with all the relevant and variable physics being subsumed into the factor k. the study by fg includes only gas return from stars with mass @xmath01m@xmath34 . when updated with more recent planetary nebula formation rates @xcite , it predicts that k = 0.07 , shown by a dashed line in figure 3 . the assumption underlying the fg approach is that gas returned to the ism during the first few gyr of the galaxy s existence was blown into intergalactic space by type ii supernovae . more complex descriptions of stellar gas return , which include contributions from a range of masses described by salpeter or miller - scalo initial mass functions , predict larger values of k , i.e. k@xmath01 . those descriptions have been incorporated in models of ism evolution which attempt to explain the x - ray emission from plasma within the deep potential wells of optically luminous e and s0 galaxies . the assumed masses are therefore considerably larger than appropriate for most of the galaxies in our sample . with that caveat in mind , the studies most relevant for interpreting our data are brighenti & mathews s ( 1997 ) work on flattened ellipticals , and dercole & ciotti s ( 1998 ) investigation of lenticular galaxies . @xcite and @xcite take the initial state of the ism to be a hot , outflowing wind - the result of heating by type ii supernovae in the initial starburst . if all gas returned after the first @xmath35 gyr is retained , the analytic approximation of @xcite yields k = 1.2 for galaxies assumed to be 10 gyr old today , implying the presence of a massive ism . the amount of gas actually retained , however , is very sensitive to the depth and shape of the gravitational potential , and also to the assumed rate of energy input from type ia supernovae . in all cases , though , the results are qualitatively consistent with our observations : most of the gas ever returned to the ism is lost , much of it in the first 1 gyr of the galaxy s life . models of more massive galaxies show the development of central cooling flows , also termed decoupled inflows , which accumulate 10@xmath36m@xmath34 of cool gas . no attempt has yet been made to explore the fate of the cooling gas . the amount of cool gas expected at the present time depends on when the initial outflow reverses , i.e. on the assumed sn ia rate for a given gravitational potential . uncertainty in the supernova rate is a major impediment to understanding what happens to the gas , but using rates consistent with recent work ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) the models of dercole & ciotti predict that the most massive galaxies in the present sample , are likely to have developed inflows after several gyr . models are not yet available of flattened galaxies with m@xmath37m@xmath34 , which is more typical of our sample , but the trends make it appear that , due to their low masses , few of our galaxies can have developed inflows , so they should be essentially free of cool gas . we therefore anticipate that models with galaxy masses more appropriate to our sample will predict that effectively no gas will be retained in the smaller galaxies . what is particularly striking , in the light of that expectation , is that we see molecular gas even in some of the lowest - mass galaxies in our sample . figure 3 reveals that those galaxies have gas deficiencies equal to ( and in some cases much less than ) the deficiencies of the most massive galaxies we have observed - objects roughly 100 times more massive . varying the energy equilibrium of the gas(e.g . the sn ia rate ) can explain the range of gas mass at any particular blue luminosity . but it seems excessively contrived to explain the upper cutoff by such parameter adjustments . the optical colours and spectra of s0 disks are consistent with them being a few gyr younger than the accompanying bulges @xcite , although most observations refer to s0s in clusters . relatively young disks can be readily explained by one of the central pillars of the hierarchical paradigm : spheroids formed in major mergers can later acquire disk components from surrounding gas reservoirs ( kauffmann et al . 1993 ; steinmetz & navarro 2002 ) . while the evolution of the morphology and gas - content of such disks has not been clarified by current hierarchical studies , the idea of gas infall over extended periods seems to offer no ready explanation for the upper cutoff in figure 3 . the fundamental reason why hierarchical models can not yet address our results is that the fate of the gas in a _ single _ merger can not be predicted with confidence . interactions between galaxies bring in new gas . they also dispose of it in star formation , by blasting it into intergalactic space through type ii supernovae , or by tidal stripping . the outcome therefore depends on many parameters , including the structure of the colliding galaxies , the encounter geometry , the prescription for star formation , and the effectiveness of stellar feedback(e.g . * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? it is again conceivable that judicious parameter tuning could explain the observed range of gas deficiency . we doubt the same is true for the upper cutoff . we report a survey of molecular gas in a sample of 27 normal , field s0 galaxies within 20 mpc . the present survey differs from earlier ones in that it is volume - limited rather than iras flux - limited , looks for gas outside the centres of the galaxies , and has sensitivity limits based on fundamental considerations of stellar evolution - that aging stars return gas to the interstellar medium . our results provide a unique basis for understanding the properties , origin , and fate of dense gas in normal s0 galaxies residing outside of rich clusters . we confirm earlier studies of different samples , finding that molecular gas is common in s0 galaxies ; 78 percent of our sample is detected in either the j=1 - 0 or j=2 - 1 transition of co. the gas is centrally located in regions with sizes up to 1 - 3 kpc ; we do not find significant amounts of gas at larger distances . we have combined observations with published hi data to estimate the total mass of cold gas in each galaxy . the results of that exercise are presented in figure 3 , which contains two important results : first , a wide range of gas content is present at a fixed optical luminosity . second , an upper cutoff in gas content is present across the entire luminosity range spanned by our sample : -17 @xmath38m@xmath39 - 21 . those findings , but especially the second , present a significant challenge for our ideas of how s0 galaxies have evolved . if s0s formed during a single monolithic collapse early in the history of the universe , and have subsequently evolved as _ closed systems _ , then all now contain much less gas than expected from stellar mass loss ; figure 3 indicates that 10@xmath4010@xmath12 . that fact alone , though , does not necessarily present a serious problem for models of ism evolution based on the monolithic formation paradigm . the expected amount of cool gas is extremely sensitive to poorly known quantities such as the assumed rate of type ia supernovae , and the shape and depth of the gravitational potential well , so that a wide range of gas content can be explained . it is very hard to understand , however , how those poorly known factors , and perhaps others , could conspire to produce the upper cutoff of gas content so clearly evident in figure 3 .. within the paradigm of hierarchical galaxy formation , s0 bulges may have formed through major mergers , followed by the slow accretion of surrounding gas to form their disks . although that idea can explain optical observations which imply relatively young ages for s0 disks , it might be difficult to understand how the upper cutoff would arise from an extended period of gas accretion . in summary , our study poses broad and significant questions for the two popular galaxy formation paradigms , and it is not yet clear which , if either , will be more successful at answering them . it will certainly be helpful to substantially deepen our understanding of the properties of the cold gas , and we are embarking on projects for this purpose . we are grateful to dave hogg and luca ciotti for their perceptive comments on early versions of this paper , and to bill mathews for insights into ism evolution in galaxies . comments from the referee helped us to improve our presentation . this work was supported by a canadian natural sciences and engineering research council research grant to gaw . appleton , p. n. & davies , r. d. 1982 , , 201 , 1073 . arimoto , n. , matsushita , k. , ishimaru , y. , ohashi , t. & renzini , a. 1997 , , 477 , 128 . arimoto , n. & yoshii , y. 1987 , , 173 , 23 . bally , j. & thronson , h. a. 1989 , , 97 , 69 . barger , a. j. , cowie , l. l. , smail , i. , ivison , r. j. , blain , a. w. , and kneib , j .- 1999 , , 117 , 2656 . barnes , j. e. 2002 , , 333 , 481 . barnes , j. e. & hernquist , l. 1996 , , 471 , 115 . barnes , j. e. & hernquist , l. 1991 , , 370 , l65 . baugh , c. m. , cole , s. & frenk , c. s. 1996 , , 283 , 1361 . bieging , j. h. & biermann , p. 1977 , , 60 , 361 . bekki , k. & noguchi , m. 1994 , , 290 , 7 . bekki , k. 1998 , , 334 , 814 . bekki , k. 1998 , , 496 , 713 . bekki , k. 1998 , , 502 , l133 . biermann , p. , clarke , j. n. & fricke , k. j. 1979 , , 75 , 7 . bothun , g. d. & gregg , m. d. 1990 , , 350 , 73 . bottinelli , l. , gouguenheim , l. & heidmann , j. 1973 , , 22 , 281 . bower , r. g. , lucey , j. r. , & ellis , r. s. 1992 , , 254 , 601 . bregman , j.n . , hogg , d. e. & roberts , m. s. 1992 , , 387 , 484 . brighenti , f. & mathews , w. g. 1997 , , 490 , 592 . bruzual a , g. , & charlot , s. 1993 , , 405 , 538 . burstein , d. , krumm , n. & salpeter , e. e. 1987 , , 94 , 883 . buson , l. m. , sadler , e. m. , zeilinger , w. w. , bertin , g. , bertola , f. , danziger , j. , dejonghe , h. , saglia , r. p. , & de zeeuw , p. t. 1993 , , 280 , 409 . caldwell , n. 1983 , , 268 , 90 . capaccioli , m. , cappellaro , e. , held , e. v. & vietri , m. 1993 , , 274 , 69 . chamaraux , p. , balkowski , c. & fontanelli , p. 1987 , , 69 , 263 . ciotti , l. , dercole , a. , pellegrini , s. & renzini , a. 1991 , , 376 , 380 . dercole , a. & ciotti , l. 1998 , , 494 , 535 . dressler , a. & sandage , a. 1983 , , 265 , 664 . ebneter , k. , davis , m. & djorgovski , s. 1988 , , 95 , 422 . faber , s. & gallagher , j.s . 1976 , , 204 , 365 . fisher , d. , franx , m. & illingworth , g. 1996 , , 459 , 110 . fisher , d. 1997 , , 113 , 950 . gerin , m. & casoli , f. 1994 , , 290 , 49 . giovanardi , c. , krumm , n , & salpeter , e. e. 1983 , , 88 , 1719 . hernquist , l. & mihos , j. c. 1995 , , 448 , 41 . hogg , d. e. , roberts , m. s. & sandage , a. 1993 , , 106 , 907 . howard , s. , keel , w. c. , byrd , g. & burkey , k. 1993 , , 417 , 502 . huchtmeier , w. k. 1982 , , 110 , 121 . kauffmann , g , , white , s. d. m. & guiderdoni , b. 1993 , , 264 , 201 . kaufmann , g. & charlot , s. 1998 , , 297 , l23 . kennicutt , r. c. jr . , tamblyn , p. , congdon , c. w. 1994 , , 435 , 22 . kent , s. m. 1987 , , 93 , 1062 . kim , d .- w . 1989 , , 346 , 653 . knapp , g. r. , kerr , f. g. & henderson , a. p. 1979 , , 234 , 448 . knapp , g. r. & gunn , j. e. 1982 , , 87 , 1634 . knapp , g. r. , van driel , w. , schwarz , u. j. , van woerden , h. & gallagher , j. s. iii 1984 , , 133 , 127 . larson , r. b. 1974 , mnras , 173 , 671 . lees , j. , knapp , g. r. , rupen , m. p. & phillips , t. g. 1991 , , 379 , 177 . lewis , b. m. 1987 , , 63 , 515 . li , j.g . , seaquist , e.r . & sage , l.j . 1993 , , 411 , l71 . lilly , s. j. , eales , s. a. , gear , w. k. p. , hammer , f. , lefevre , o. , crampton , d. , bond , j. r. , and dunn , l. 1999 , , 518 , 641 . madau , p. , ferguson , h. c. , dickinson , m. e. , giavalisco , m. , steidel , c. c. , & fruchter , a. 1996 , , 283 , 1388 . matsumoto , h. , koyama , k. , awaki , h. , tsuru , t. , loewenstein , m. & matsushita , k. 1997 , 482 , 133 . mihos , j. c. , richstone , d. o. , & bothun , g. d. 1992 , , 333 , 481 . mihos , j. c. & hernquist , l. 1996 , , 464 , 641 . munn , j. a. 1992 , , 399 , 444 . peimbert , m. 1993 , i.a.u . 155 , planetary nebulae , r. weinberger & a. acker , dordrecht : kluwer , 1993 , 523 . pogge , r. w. & eskridge , p. b. 1993 , , 106 , 1405 . richter , o .- g . , & huchtmeier , w. k. 1987 , , 68 , 427 . roberts , m. s. 1975 , stars and stellar systems ix , a. r. sandage , m. sandage , and j. kristian , eds , ( chicago : university of chicago press ) , p. 309 . roberts , m. s. , hogg , d. e. , bregman , j. n. , forman , w. r. & jones , c. 1991 , , 75 , 751 . rocca - volmerange , b. & guiderdoni , b. 1988 , , 75 , 93 . roukema , b. f. , ninin , s. , devriendt , j. , bouchet , f. r. , guiderdoni , b. & maman , g. a. 2001 , , 373 , 494 . sage , l.j . & wrobel , j. 1989 , , 344 , 204 . sage , l.j . , 1990 , , 239 , 125 . sage , l.j . & isbell , d. w. 1991 , , 247 , 320 . sage , l.j . , salzer , j.j . , loose , h. & henkel , c. 1992 , , 265 , 19 . sage , l.j . 1993 , , 239 , 125 . sage , l. j. & galletta , g. 1994 , , 108 , 1633 . sancisi , r. , van woerden , h. , davies , r. d. & hart , l. 1984 , , 210 , 497 . sandage , a.r . 1986 , revised shapley ames catalog , ( rsac ) ( washington : carnegie institution ) . sandage , a.r . 1994 , carnegie atlas , ( washington : carnegie institution ) . schweizer , f. & seitzer , p. 1992 , , 104 , 1039 . shostak , g. s. 1987 , , 175 , 4 . steinmetz , m. & navarro , j. f. 2002 , new astronomy , 7 , 155 . tacconi , l. j. , tacconi - garman , l. e. , thornley , m. & van woerden , h. 1991 , , 252 , 541 . taniguchi , y. et al . 1994 , , 108 , 468 . thronson , h. a. jr . , tacconi , l. , kenney , j. , greenhouse , m. a. , margulis , m. , tacconi - garman , l. & young , j. s. 1989 , , 344 , 747 . tully , r.b . 1988 , nearby galaxies catalog , ( cambridge : cambridge university press ) . turatto , m. , cappellaro , e. & benetti , s. 1994 , , 108 , 202 . van den bergh , s. 1990 , , 348 , 57 . van den bergh , s. 1994 , , 107 , 153 . van der burg , g. 1985 , , 62 , 147 . van driel , w. van woerden , h. , gallagher , j. s. iii , & schwarz , u , j. 1988 , , 191 , 201 . veron - cetty , m .- p . & veron , p. 1988 , , 204 , 28 . wardle , m. & knapp , g. r. 1986 , , 91 , 23 . welch , g. a. , mitchell , g. f. , & yi , s. 1996 , , 470 , 781 . white , s. d. m. & rees , m. j. 1978 , , 183 , 341 . white , s. d. m. & frenk , c. s. 1991 , , 379 , 52 . wiklind , t. & henkel , c. 1989 , , 225 , 1 ( wh ) . wiklind , t. , & henkel , c. 1990 , , 227 , 394 . wiklind , t. , combes , f. & henkel , c. 1995 , , 297 , 643 . wiklind , t. , combes , f. , henkel , c. & wyrowski , f. 1997 , , 323 , 727 . young , j. s. , xie , s. , tacconi , l. , knezek , p. , viscuso , p. , tacconi - garman , l. , scoville , n. , schneider , s. , schloerb , f. p. , lord , s. , lesser , a. , kenney , j. , huang , y .- l . , devereux , n. , claussen , m. , case , j. , carpenter , j. , berry , m. , & allen , l. 1995 , , 98 , 219 . lcccccc ngc 404 & 01:06:39.3 & 35:27:06 & -35 & s03(0 ) & 2.4 & 10.92 + ngc 936 & 02:25:04.7 & -01:22:42 & 1340 & sb02/3/sba & 10.6 & 10.98 + ngc 1023 & 02:37:15.9 & 38:50:55 & 637 & sb01(5 ) & 10.5 & 10.08 + ngc 2787 & 09:14:49.7 & 69:24:50 & 696 & sb0/a & 13 & 11.61 + ngc 3115 & 10:02:44.7 & -07:28:31 & 670 & s01(7 ) & 6.7 & 9.74 + + ngc 3384 & 10:45:38.7 & 12:53:41 & 735 & sb01(5 ) & 8.1 & 10.75 + ngc 3412 & 10:48:14.6 & 13:40:41 & 865 & sb01/2(5 ) & 8.1 & 11.34 + ngc 3489 & 10:57:40.0 & 14:10:15 & 708 & s03/sa & 6.4 & 11.15 + ngc 3607 & 11:14:16.1 & 18:19:35 & 935 & s03(3 ) & 19.9 & 10.79 + ngc 3870 & 11:43:16.9 & 50:28:43 & 750 & [ s0 ] & 17 & 13.47 + + ngc 3941 & 11:50:19.4 & 37:15:55 & 928 & sb01/2/a & 18.9 & 11.25 + ngc 3990 & 11:55:01.0 & 55:44:15 & 696 & s06/sa & 17 & 13.46 + ngc 4026 & 11:56:51.1 & 51:14:25 & 930 & s01/2(9 ) & 17 & 11.59 + ngc 4111 & 12:04:30.6 & 43:20:43 & 807 & s01(9 ) & 17 & 11.6 + ngc 4143 & 12:07:05.0 & 42:48:52 & 784 & s01(5)/sa & 17 & 11.9 + + + ngc 4150 & 12:08:01.3 & 30:40:47 & 244 & s03(4)/sa & 9.7 & 12.4 + ngc 4203 & 12:12:34.1 & 33:28:33 & 1086 & s02(1 ) & 9.7 & 11.61 + ngc 4251 & 12:15:36.8 & 28:27:11 & 1014 & s01(8 ) & 9.7 & 11.43 + ngc 4310 & 12:19:56.1 & 29:29:10 & 913 & [ sab0 ? ] & 9.7 & 13.08 + ngc 4346 & 12:21:01.2 & 47:16:16 & 783 & sb01(8 ) & 17 & 12.17 + + ngc 4460 & 12:26:20.2 & 45:08:21 & 528 & s0/sc & 8.1 & 12.25 + ngc 4880 & 12:57:41.0 & 12:45:10 & 1491 & [ e4/s01(4 ) & 15.7 & 12.35 + ngc 5866 & 15:05:07.8 & 55:57:16 & 745 & s03(8 ) & 15.3 & 10.83 + ngc 7013 & 21:01:26.1 & 29:41:51 & 779 & [ s0/a ] & 14.2 & 10.7 + ngc 7077 & 21:27:27.5 & 02:11:39 & 1146 & [ s0 pec ? ] & 13.3 & 13.94 + + ngc 7332 & 22:35:01.2 & 23:32:16 & 1300 & s02/3(8 ) & 18.2 & 11.93 + ngc 7457 & 22:58:36.2 & 29:52:31 & 824 & s01(5 ) & 12.3 & 11.76 + ngc 2787 & 0 , 0 & 493 - 906 & 1.02 0.35 & 3.92 & & 1.93 0.46 & 3.78 & & & & & & & + ngc 3115 & -14 , -17 & 360 - 450 & 0.216 0.190 & 5.36 & & 0.953 0.251 & 6.81 & & 360 - 960 & -0.423 0.725 & 6.02 & & & + & 0 , 0 & 500 - 840 & & & & 1.95 0.54 & 5.58 & & 360 - 960 & 0.232 0.584 & 4.85 & & & + & 14 , 17 & 870 - 960 & 0.023 0.191 & 5.37 & & 0.320 0.233 & 6.32 & & 360 - 960 & -1.42 0.67 & 5.55 & & & + ngc 3384 & 0 , 0 & 535 - 935 & -0.796 0.260 & 3.00 & & 2.49 0.56 & 4.75 & & & & & & & + ngc 3412 & 0 , 0 & 727 - 986 & 0.788 0.204 & 3.18 & & 0.990 0.307 & 4.20 & & & & & & & + ngc 3489 & 0 , 0 & 521 - 879 & 3.49 0.39 & 4.21 & & 4.86 0.41 & 3.97 & & & & & & & + ngc 3607 & -9.5 , 5.5 & 631 - 1238 & 2.59 0.37 & 3.16 & & & & & 839 - 1221 & & & & 2.44 0.46 & 4.17 + & 0 , 0 & 670 - 1227 & 8.74 0.97 & 4.16 & & 9.85 0.56 & 2.35 & & & & & & & + & 9.5 , -5.5 & 634 - 1225 & 10.46 0.36 & 3.05 & & & & & 645 - 1219 & & & & 21.62 0.49 & 1.84 + & 19.1 , -11.0 & 632 - 1161 & 5.35 0.45 & 4.14 & & & & & 670 - 845 & & & & 2.55 0.31 & 5.08 + ngc 3870 & -4.7 , -10 & 714 - 900 & 1.13 0.25 & 3.61 & & & & & 736 - 880 & & & & 1.55 0.36 & 8.13 + & 0 , 0 & 700 - 850 & 1.60 0.15 & 2.38 & & & & & 700 - 800 & & & & 2.46 0.24 & 6.90 + & 4.7 , 10 & 701 - 817 & 0.894 0.120 & 2.33 & & & & & 708 - 797 & & & & 0.780 0.240 & 7.17 + ngc 3941 & 0 , 0 & 728 - 1128 & 0.729 0.295 & 2.41 & & 0.267 0.716 & 4.30 & & & & & & & + ngc 3990 & 0 , 0 & 625 - 865 & 0.328 0.177 & 1.95 & & 0.863 0.181 & 1.81 & & & & & & & + ngc 4026 & 0 , 0 & 780 - 1080 & 0.183 0.206 & 2.02 & & 0.072 0.588 & 4.85 & & & & & & & + ngc 4111 & 0 , 0 & 607 - 1007 & 1.29 0.29 & 2.38 & & 3.80 0.35 & 2.09 & & & & & & & + ngc 4143 & 0 , 0 & 584 - 984 & 0.003 0.359 & 3.26 & & & & & 700 - 984 & & & & 1.17 0.91 & 6.07 + ngc 4150 & -7.5 , 8 & 25 - 320 & 3.04 0.26 & 3.54 & & 1.28 0.70 & 8.03 & & & & & & & + & 0 , 0 & 22 - 362 & 6.61 0.26 & 3.41 & & 4.65 0.36 & 3.66 & & & & & & & + & 7.7 -8 & 22 - 362 & 6.74 0.39 & 4.97 & & 8.96 0.48 & 4.90 & & & & & & & + & 15 , -16 & 180 - 364 & 2.50 0.31 & 5.66 & & & & & 262 - 336 & & & & 0.798 0.280 & 8.57 + ngc 4203 & 0 , 0 & 868 - 1300 & 1.88 0.21 & 1.59 & & & & & 891 - 1261 & & & & 4.52 0.35 & 3.24 + ngc 4251 & 0 , 0 & 914 - 1114 & 0.094 0.195 & 3.51 & & -0.346 0.417 & 6.79 & & & & & & & + ngc 4310 & -26 , 20 & 881 - 1024 & 0.408 0.270 & 4.34 & & 0.744 0.290 & 5.94 & & & & & & & + & -17 , 14 & 881 - 1024 & 3.12 0.22 & 4.78 & & 2.48 0.37 & 7.63 & & & & & & & + & -9 , 7 & 853 - 1005 & 5.70 0.23 & 4.88 & & 6.86 0.39 & 7.62 & & & & & & & + & 0 , 0 & 813 - 996 & 6.69 0.31 & 5.86 & & 7.17 0.37 & 6.35 & & & & & & & + & 9 , -7 & 799 - 993 & 3.60 0.24 & 4.45 & & 2.91 0.36 & 6.05 & & & & & & & + & 17 , -14 & 813 - 929 & 0.847 0.184 & 4.53 & & 0.396 0.290 & 6.71 & & & & & & & + ngc 4346 & 0 , 0 & 583 - 983 & 0.407 0.413 & 3.77 & & -0.196 0.726 & 6.90 & & & & & & & + ngc 4460 & -11 , -19 & 390 - 500 & 1.10 0.25 & 5.04 & & 0.234 0.274 & 5.22 & & & & & & & + & -6 , -10 & 410 - 544 & 0.988 0.218 & 3.85 & & 0.622 0.354 & 5.97 & & & & & & & + & 0 , 0 & 402 - 564 & 2.12 0.20 & 3.16 & & 0.676 0.315 & 4.63 & & & & & & & + & 6 , 10 & 454 - 584 & 2.24 0.25 & 4.39 & & -0.339 0.305 & 5.23 & & & & & & & + & 11 , 19 & 514 - 565 & 0.560 0.118 & 3.54 & & 0.588 0.193 & 5.67 & & & & & & & + ngc 4880 & 0 , 0 & 1342 - 1413 & 0.823 0.110 & 3.52 & & 0.999 0.110 & 4.72 & & & & & & & + ngc 5866 & -35 , 27 & 479 - 606 & 0.441 0.170 & 4.02 & & 0.810 0.360 & 7.94 & & & & & & & + & -26 , 20 & 479 - 606 & 1.68 0.24 & 5.64 & & 1.04 0.28 & 6.27 & & & & & & & + & -17 , 14 & 466 - 641 & 4.57 0.19 & 4.01 & & 4.03 0.33 & 6.18 & & & & & & & + & -9 , 7 & 458 - 752 & 10.97 0.28 & 3.95 & & 12.53 0.65 & 7.59 & & & & & & & + & 0 , 0 & 432 - 1051 & 21.61 0.52 & 4.23 & & & & & 455 - 1014 & & & & 18.79 0.69 & 2.89 + & 9 , -7 & 410 - 1087 & 28.06 0.37 & 2.98 & & & & & 472 - 1040 & & & & 32.79 0.83 & 3.23 + & 17 , -14 & 708 - 1047 & 13.83 0.35 & 4.31 & & & & & 797 - 1022 & & & & 13.69 0.50 & 7.43 + & 26 , -20 & 816 - 1048 & 5.01 0.29 & 4.46 & & & & & 834 - 1027 & & & & 6.94 0.36 & 5.99 + & 35 , -27 & 866 - 1100 & 1.68 0.35 & 5.38 & & & & & 854 - 1014 & & & & 0.449 0.390 & 7.44 + & -9 , -7 & 465 - 1042 & 6.27 0.82 & 6.63 & & & & & 521 - 1036 & & & & 2.94 0.93 & 5.00 + & 9 , 7 & 455 - 1061 & 15.55 0.82 & 6.59 & & & & & 479 - 1037 & & & & 25.15 1.21 & 5.03 + ngc 936 & -39 , 39 & 1490 - 1580 & -0.021 & 0.134 & 2.98 + & 0 , 0 & 1190 - 1490 & 0.058 & 0.408 & 4.52 + & 39 , -39 & 1100 - 1190 & -0.037 & 0.181 & 4.02 + ngc 1023 & -220 , -12 & 430 - 760 & 0.087 & 0.486 & 5.07 + & -220 , -12 & 400 - 1000 & 0.792 & 0.721 & 4.74 + & -165 , -9 & 430 - 760 & 0.217 & 0.224 & 2.33 + & -165 , -9 & 400 - 1000 & & 0.357 & 2.33 + & -110 , -6 & 430 - 800 & -0.366 & 0.395 & 3.80 + & -110 , -6 & 400 - 1000 & -0.195 & 0.519 & 3.38 + & -55 , -3 & 530 - 800 & 0.283 & 0.292 & 3.46 + & -55 , -3 & 400 - 1000 & -0.141 & 0.476 & 3.11 + & 0 , 0 & 530 - 800 & -0.014 & 0.149 & 1.77 + & 0 , 0 & 400 - 1000 & 0.449 & 0.235 & 1.54 + & 55 , 3 & 530 - 800 & 0.056 & 0.251 & 2.97 + & 55 , 3 & 400 - 1000 & 1.24 & 0.45 & 2.91 + & 110 , 6 & 530 - 850 & 1.02 & 0.53 & 5.62 + & 110 , 6 & 400 - 1000 & 1.10 & 0.80 & 5.19 + & 165 , 9 & 600 - 850 & 0.080 & 0.221 & 2.75 + & 165 , 9 & 400 - 1000 & -0.911 & 0.381 & 2.49 + & 220 , 12 & 600 - 850 & -0.338 & 0.152 & 1.90 + & 220 , 12 & 400 - 1000 & -0.293 & 0.275 & 1.79 + ngc 2787 & 0 , 0 & 580 - 690 & 0.355 & 0.072 & 1.37 + ngc 3115 & -71 , -84 & 360 - 450 & -0.312 & 0.238 & 5.31 + & -35 , -42 & 360 - 450 & 0.029 & 0.129 & 2.87 + & 0 , 0 & 360 - 960 & -0.335 & 0.582 & 3.80 + & 35 , 42 & 870 - 960 & -0.021 & 0.162 & 3.60 + & 71 , 84 & 870 - 960 & -0.111 & 0.188 & 4.19 + ngc 3384 & 0 , 0 & 535 - 935 & -0.136 & 0.406 & 3.66 + & 44 , 33 & 535 - 935 & -0.547 & 0.312 & 2.94 + ngc 3412 & 0 , 0 & 727 - 986 & -0.420 & 0.250 & 3.03 + ngc 3607 & 0 , 0 & 634 - 1221 & 8.72 & 0.96 & 6.42 + ngc 3870 & 0 , 0 & 635 - 937 & 1.50 & 0.36 & 4.11 + ngc 3941 & 0 , 0 & 728 - 1128 & 0.620 & 0.197 & 1.79 + ngc 4026 & -2 , 55 & 630 - 920 & 0.532 & 0.241 & 2.73 + & 0 , 0 & 630 - 1180 & 0.043 & 0.325 & 2.29 + & 2 , -55 & 900 - 1180 & 1.03 & 0.21 & 2.47 + ngc 4111 & 0 , 0 & 607 - 1007 & -0.793 & 0.510 & 4.64 + ngc 4251 & 0 , 0 & 864 - 1164 & -0.336 & 0.227 & 2.52 + ngc 4310 & -43 , 34 & 881 - 1024 & 0.462 & 0.310 & 5.28 + & 0 , 0 & 813 - 996 & 1.56 & 0.71 & 1.07 + & 43 , -34 & 813 - 929 & -0.088 & 0.230 & 4.48 + ngc 4346 & 0 , 0 & 583 - 983 & -0.475 & 0.324 & 2.94 + ngc 4460 & 0 , 0 & 392 - 586 & 0.676 & 0.200 & 3.22 + & 35 , 42 & 434 - 606 & 0.948 & 0.340 & 5.16 + ngc 5866 & -86 , 68 & 479 - 606 & -0.190 & 0.110 & 4.25 + & -43 , 34 & 479 - 606 & -0.250 & 0.130 & 4.85 + & 0 , 0 & 455 - 1014 & 5.14 & 0.49 & 3.14 + & 43 , -34 & 931 - 1062 & 0.767 & 0.100 & 3.68 + & 86 , -68 & 931 - 1062 & -0.462 & 0.110 & 3.95 + ngc 7332 & -23 , 50 & 1100 - 1500 & -1.17 & 0.31 & 2.81 + & 0 , 0 & 1100 - 1500 & 0.186 & 0.089 & 0.81 + & 0 , 0 & 1370 - 1520 & 0.119 & 0.045 & 0.75 + & 23 , -50 & 1100 - 1500 & 0.351 & 0.272 & 2.48 + ngc 7457 & -42 , 35 & 874 - 974 & 0.086 & 0.241 & 5.07 + & 0 , 0 & 695 - 945 & 0.234 & 0.073 & 0.913 + & 42 , -35 & 674 - 774 & 0.348 & 0.184 & 3.89 + ngc 936 & & & & @xmath411.55@xmath3010@xmath42 + ngc 1023 & & & & @xmath414.91@xmath3010@xmath42 + ngc 2787 & & ( [email protected])@xmath3010@xmath43 & & ( [email protected])@xmath3010@xmath42 + ngc 3115 & & ( [email protected])@xmath3010@xmath43 & & + ngc 3384 & & ( [email protected])@xmath3010@xmath43 & & + + ngc 3412 & ( [email protected])@xmath3010@xmath43 & ( [email protected])@xmath3010@xmath43 & & + ngc 3489 & ( [email protected])@xmath3010@xmath43 & ( [email protected])@xmath3010@xmath43 & & + ngc 3607 & ( [email protected])@xmath3010@xmath44 & ( [email protected])@xmath3010@xmath44 & & ( [email protected])@xmath3010@xmath45 + ngc 3870 & ( [email protected])@xmath3010@xmath42 & ( [email protected])@xmath3010@xmath42 & & ( [email protected])@xmath3010@xmath44 + ngc 3941 & & & & ( [email protected])@xmath3010@xmath42 + + ngc 3990 & & ( [email protected])@xmath3010@xmath43 & & + ngc 4026 & & & & ( [email protected])@xmath3010@xmath42 + ngc 4111 & ( [email protected])@xmath3010@xmath42 & ( [email protected])@xmath3010@xmath42 & & + ngc 4143 & @xmath414.45@xmath3010@xmath43 & @xmath415.55@xmath3010@xmath43 & & + ngc 4150 & ( [email protected])@xmath3010@xmath42 & ( [email protected])@xmath3010@xmath42 & & + + ngc 4203 & ( [email protected])@xmath3010@xmath43 & ( [email protected])@xmath3010@xmath43 & & + ngc 4251 & @xmath417.87@xmath3010@xmath46 & @xmath418.35@xmath3010@xmath46 & & @xmath416.35@xmath3010@xmath43 + ngc 4310 & ( [email protected])@xmath3010@xmath42 & ( [email protected])@xmath3010@xmath42 & & + ngc 4346 & @xmath415.12@xmath3010@xmath43 & @xmath414.46@xmath3010@xmath43 & & @xmath412.78@xmath3010@xmath42 + ngc 4460 & ( [email protected])@xmath3010@xmath42 & & & ( [email protected])@xmath3010@xmath42 + + ngc 4880 & ( [email protected])@xmath3010@xmath43 & ( [email protected])@xmath3010@xmath43 & & + ngc 5866 & ( [email protected])@xmath3010@xmath44 & ( [email protected])@xmath3010@xmath44 & & ( [email protected])@xmath3010@xmath44 + ngc 7332 & & & & @xmath414.16@xmath3010@xmath42 + ngc 7457 & & & & ( [email protected])@xmath3010@xmath43 + ngc 1023 & @xmath81.69@xmath3010@xmath43 & 15 & @xmath87.71@xmath3010@xmath43 & 55 + ngc 3115 & @xmath84.35@xmath3010@xmath43 & 15 & 1.864@xmath3010@xmath43 & 10 + & & & @xmath83.37@xmath3010@xmath43 & 21 + & & & @xmath82.33@xmath3010@xmath42 & 55 + ngc 4251 & @xmath81.1@xmath3010@xmath42 & 21 & @xmath82.36@xmath3010@xmath43 & 21 + & & & @xmath81.94@xmath3010@xmath42 & 55 + ngc 4310 & 6.30@xmath3010@xmath42 & 21 & 2.702@xmath3010@xmath42 & 21 + & & & @xmath85.96@xmath3010@xmath42 & 55 + ngc 5866 & 8.34@xmath3010@xmath42 & 15 & 2.17@xmath3010@xmath44 & 21 + & 1.02@xmath3010@xmath44 & 21 & 2.17@xmath3010@xmath44 & 21 + & 5.09@xmath3010@xmath44 & 45 & 3.58@xmath3010@xmath44 & 55 + & 4.37@xmath3010@xmath44 & 3 @xmath3045 & 4.38@xmath3010@xmath44 & 5 @xmath30 21 + ngc 7457 & 2.64@xmath3010@xmath42 & 15 & 3.30@xmath3010@xmath43 & 55 + ngc 404 & 4.48@xmath3010@xmath44 & & 1.01@xmath3010@xmath44 & 1 + ngc 936 & 8.28@xmath3010@xmath45 & @xmath81.55@xmath3010@xmath42 & @xmath87.33@xmath3010@xmath44 & 1 + ngc 1023 & 1.86@xmath3010@xmath47 & @xmath84.91@xmath3010@xmath42 & @xmath82.49@xmath3010@xmath45 & 2 + ngc 2787 & 6.97@xmath3010@xmath45 & 1.78@xmath3010@xmath42 & 1.10@xmath3010@xmath45 & 3 + ngc 3115 & 1.04@xmath3010@xmath47 & 1.86@xmath3010@xmath43 & @xmath81.86@xmath3010@xmath42 & 4 + + ngc 3384 & 5.97@xmath3010@xmath45 & 3.48@xmath3010@xmath43 & @xmath82.22@xmath3010@xmath42 & 5 + ngc 3412 & 3.47@xmath3010@xmath45 & 2.22@xmath3010@xmath43 & @xmath81.82@xmath3010@xmath42 & 5 + ngc 3489 & 2.58@xmath3010@xmath45 & 6.13@xmath3010@xmath43 & 1.67@xmath3010@xmath42 & 5 + ngc 3607 & 3.47@xmath3010@xmath47 & 1.03@xmath3010@xmath45 & @xmath81.50@xmath3010@xmath45 & 6 + ngc 3870 & 2.15@xmath3010@xmath45 & 1.28@xmath3010@xmath44 & @xmath81.86@xmath3010@xmath45 & 7 + + ngc 3941 & 2.05@xmath3010@xmath47 & 6.58@xmath3010@xmath42 & 1.94@xmath3010@xmath45 & 8 + ngc 3990 & 2.17@xmath3010@xmath45 & 5.30@xmath3010@xmath43 & & + ngc 4026 & 1.21@xmath3010@xmath47 & 8.80@xmath3010@xmath42 & @xmath82.15@xmath3010@xmath44 & 9 + ngc 4111 & 1.20@xmath3010@xmath47 & 1.59@xmath3010@xmath42 & 1.13@xmath3010@xmath45 & 10 + ngc 4143 & 9.12@xmath3010@xmath45 & @xmath84.45@xmath3010@xmath43 & @xmath81.00@xmath3010@xmath45 & 8 + + ngc 4150 & 1.87@xmath3010@xmath45 & 3.81@xmath3010@xmath42 & @xmath86.27@xmath3010@xmath42 & 11 + ngc 4203 & 3.88@xmath3010@xmath45 & 7.61@xmath3010@xmath43 & 7.43@xmath3010@xmath44 & 12 + ngc 4251 & 4.58@xmath3010@xmath45 & @xmath86.35@xmath3010@xmath43 & @xmath83.89@xmath3010@xmath42 & 13 + ngc 4310 & 1.00@xmath3010@xmath45 & 4.30@xmath3010@xmath42 & 9.59@xmath3010@xmath42 & 14 + ngc 4346 & 7.11@xmath3010@xmath45 & @xmath82.78@xmath3010@xmath42 & @xmath85.17@xmath3010@xmath44 & 15 + + ngc 4460 & 1.50@xmath3010@xmath45 & 1.32@xmath3010@xmath42 & & + ngc 4880 & 5.14@xmath3010@xmath45 & 8.70@xmath3010@xmath43 & @xmath82.69@xmath3010@xmath42 & 9 + ngc 5866 & 1.98@xmath3010@xmath47 & 4.39@xmath3010@xmath44 & @xmath81.32@xmath3010@xmath45 & 16 + ngc 7013 & 1.92@xmath3010@xmath47 & & 1.57@xmath3010@xmath45 & 17 + ngc 7077 & 8.53@xmath3010@xmath44 & & 1.22@xmath3010@xmath44 & 18 + + ngc 7332 & 1.02@xmath3010@xmath47 & @xmath84.16@xmath3010@xmath42 & @xmath81.98@xmath3010@xmath44 & 19 + ngc 7457 & 5.43@xmath3010@xmath45 & 3.30@xmath3010@xmath43 & @xmath89.28@xmath3010@xmath42 & 14 +	lenticular galaxies remain remarkably mysterious as a class . observations to date have not led to any broad consensus about their origins , properties and evolution , though they are often thought to have formed in one big burst of star formation early in the history of the universe , and to have evolved relatively passively since then . in that picture , current theory predicts that stellar evolution returns substantial quantities of gas to the interstellar medium ; most is ejected from the galaxy , but significant amounts of cool gas might be retained . past searches for that material , though , have provided unclear results . we present results from a survey of molecular gas in a volume - limited sample of field s0 galaxies , selected from the nearby galaxies catalog . co emission is detected from 78 percent of the sample galaxies . we find that the molecular gas is almost always located inside the central few kiloparsecs of a lenticular galaxy , meaning that in general it is more centrally concentrated than in spirals . we combine our data with hi observations from the literature to determine the total masses of cool and cold gas . curiously , we find that , across a wide range of luminosity , the most gas rich galaxies have @xmath010 percent of the total amount of gas ever returned by their stars . that result is difficult to understand within the context of either monolithic or hierarchical models of evolution of the interstellar medium .
in this paper we study the dynamics of two different incompressible fluids with the same viscosity in a bounded porous medium . this is known as the confined muskat problem . for this problem we show that there are global in time lipschitz continuous solutions corresponding to initial data that fulfills some conditions related to the amplitude , slope and depth . this problem is of practical importance because it is used as a model for a geothermal reservoir ( see @xcite and references therein ) or a model of an aquifer or an oil well ( see @xcite ) . the velocity of a fluid flowing in a porous medium satisfies darcy s law ( see @xcite ) @xmath0 where @xmath1 is the dynamic viscosity , @xmath2 is the permeability of the medium , @xmath3 is the acceleration due to gravity , @xmath4 is the density of the fluid , @xmath5 is the pressure of the fluid and @xmath6 is the incompressible velocity field . to simplify the notation we assume @xmath7 the motion of a fluid in a two - dimensional porous medium is analogous to the hele - shaw cell problem ( see @xcite and the references therein ) . let us consider the spatial domain @xmath8 for @xmath9 . we assume impermeable boundary conditions for the velocity in the walls . in this domain we have two immiscible and incompressible fluids with the same viscosity and different densities ; @xmath10 fills the upper subdomain and @xmath11 fills the lower subdomain ( see figure [ ivscheme ] ) . the graph @xmath12 is the interface between the fluids . it is well - known that the system is in the ( rayleigh - taylor ) stable regime if the denser fluid is below the lighter one in every point @xmath13 , _ i.e. _ @xmath14 . conversely , the system is in the unstable regime if there is at least a point @xmath13 where the denser fluid is above the lighter one . if the fluids fill the whole plane the contour equation satisfies ( see @xcite ) @xmath15 for this equation the authors show the existence of classical solution locally in time ( see @xcite and also @xcite ) in the rayleigh - taylor stable regime , and maximum principles for @xmath16 and @xmath17 ( see @xcite ) . moreover , in @xcite the authors show the existence of turning waves and finite time singularities . in @xcite the authors show an energy balance for the @xmath18 norm and some results concerning the global existence of solutions corresponding to _ small _ initial data . furthermore , they show that if initially @xmath19 , then there is global lipschitz solution and if the initial data has small @xmath20 norm then there is global classical solution . the case where the fluid domain is the strip @xmath8 , with @xmath9 , has been studied in @xcite . in this domain the equation for the interface is @xmath21 for equation the authors in @xcite obtain local existence of classical solution when the system starts its evolution in the stable regime and the initial interface does not reach the walls , and the existence of initial data such that @xmath22 blows up in finite time . the authors also study the effect of the boundaries on the evolution of the interface , obtaining the maximum principle and a decay estimate for @xmath16 and the maximum principle for @xmath17 for initial data satisfying the following hypotheses : @xmath23 @xmath24 and @xmath25 these hypotheses are smallness conditions relating @xmath26 , @xmath27 and the depth . we define @xmath28 as the solution of the system @xmath29 then , for initial data satisfying @xmath30 the authors in @xcite show that @xmath31 these inequalities define a region where the slope of the solution can grow but it is bounded uniformly in time . this region only appears in the finite depth case . in this paper the question of global existence of weak solution ( in the sense of definition [ ivdefi ] ) for in the stable regime is adressed . in particular we show the following theorem : [ ivglobal ] let @xmath32 be the initial datum satisfying hypotheses , and or in the rayleigh - taylor stable regime . then there exists a global solution @xmath33 moreover , if the initial data satisfy , and the solution fulfills the following bounds : @xmath34 while , if the initial datums satisfy , the solution satisfies the following bounds : @xmath35 this result excludes the formation of cusps ( blow up of the first and second derivatives ) and turning waves for these initial data , remaining open the existence ( or non - existence ) of corners ( blow up of the curvature with finite first derivative ) during the evolution . notice that in the limit @xmath36 we recover the result contained in @xcite . in this paper and the works @xcite the effect of the boundaries over the evolution of the internal wave in a flow in porous media has been addressed . when these results for the confined case are compared with the known results in the case where the depth is infinite ( see @xcite ) three main differences appear : 1 . the decay of the maximum amplitude is slower in the confined case . 2 . there are smooth curves with finite energy that turn over in the confined case but do not show this behaviour when the fluids fill the whole plane . 3 . to avoid the turning effect in the confined case you need to have smallness conditions in @xmath27 and @xmath26 . however , in the unconfined case , only the condition in the slope is required . moreover , in the confined case a new region without turning effect appears : a region without a maximum principle for the slope but with an uniform bound . in both cases ( the region with the maximum principle and the region with the uniform bound ) , theorem [ ivglobal ] ensures the existence of a global lipschitz continuous solution . keeping these results in mind , there are some questions that remain open . for instance , the existence of a wave whose maximum slope grows but remains uniformly bounded , or the existence of a wave with small slope such that , due to the distance to the boundaries , its slope grows and the existence ( or non - existence ) of corner - like singularities when the initial data considered is small in @xmath37 . the proof of theorem [ ivglobal ] is achieved using some lemmas and propositions . first , we define _ hoc _ diffusive operators and the regularized system ( see section [ ivsec1 ] ) . for this regularized system , we show some _ a priori _ bounds for the amplitude and the slope . with these _ a priori _ bounds we show global existence of @xmath20 solution ( see section [ ivsecglobal ] ) . then , we obtain the weak solution to , @xmath38 , as the limit of the regularized solutions ( see sections [ ivsec4 ] and [ ivsec5 ] ) . on the rest of the paper we take @xmath39 and @xmath40 and we drop in the notation the @xmath41 dependence . we write @xmath42 for a universal constant that can change from one line to another . we denote @xmath43.$ ] in this section we define the regularized system and obtain some useful _ a priori _ bounds for the amplitude and the slope . to clarify the exposition we write @xmath44 for the solution of the regularized system . we remark that the term @xmath45 in is a singular integral operator , while @xmath46 is not if the curve does not reach the boundaries . in order to remove the singularity while preserving the inner structure , we put a term @xmath47 for @xmath48 in both kernels . we define @xmath49 and @xmath50 to pass to the limit we use compactness coming from an uniform bound in@xmath51,w^{1,\infty}({\mathbb r}))$ ] . thus , we need to obtain _ priori _ bounds for the amplitude and the slope . we define @xmath52 @xmath53 positive constants that will be fixed below depending only on the initial datum considered . taking derivatives in @xmath54 , we obtain some terms with positive contribution . so , we attach some diffusive operators to the regularized system . given a smooth function @xmath55 , we define @xmath56 we notice that , if the depth is not @xmath57 , the previous operators should be rescaled and we write the subscript @xmath58 to keep this dependence in mind . these operators are finite depth versions of the classical @xmath59 . roughly speaking , there are three different types of _ extra _ terms appearing in the derivatives of and that we need to control to obtain the _ a priori _ bound for the slope : 1 . there are terms which have an integrable singularity and they appear multiplied by @xmath60 . in order to handle these terms we add @xmath61 and @xmath62 . these two scales @xmath63 , @xmath64 , appear naturally due to the nonlinearity present in . there are terms which are nonlinear versions of @xmath65 and @xmath66 . these terms go to zero due to the convergence of the operators but they are not multiplied by @xmath60 . in order to handle these terms we add @xmath67 and @xmath68 . 3 . to absorb the nonsingular terms we add @xmath69 . we notice that , as @xmath70 , the square root converges to zero less than linearly . this factor will be used because the contribution of some terms is @xmath71 with @xmath72 . once the _ a priori _ bounds are achieved , we should prove global solvability in @xmath20 for the regularized system . to get this bound we add @xmath73 . we also regularize the initial datum . we take @xmath74 , @xmath75 and @xmath76 , a symmetric mollifier and define @xmath77 . given @xmath78 we define the initial datum for the regularized system as @xmath79 putting all together , we define the regularized system @xmath80 where @xmath81 are universal constants that will be fixed below depending only on the initial datum @xmath82 . we remark that @xmath83 for all @xmath84 . notice that , due to the continuity of @xmath82 , @xmath85 uniformly on any compact set in @xmath86 . since @xmath87 , we get @xmath88 and then , as as @xmath89 , we have @xmath90 a.e . thus , we have @xmath91 and @xmath92 . furthermore , we have that if @xmath82 satisfies the hypotheses , and , @xmath93 also satisfy these hypotheses if @xmath60 is small enough . moreover , if @xmath94 satisfy the same remains valid for @xmath95 and @xmath96 if @xmath60 is small enough . we use some properties of the operators @xmath97 . for the reader s convenience , we collect them in the following lemma : [ ivops ] for the operators @xmath97 ( see ) , the following properties hold : 1 . @xmath97 is @xmath18-symmetric . @xmath97 is positive definite . 3 . let @xmath55 be a schwartz function . then , they converge acting on @xmath55 as @xmath60 goes to zero : @xmath98 4 . let @xmath55 be a schwartz function . then , the derivative can be written in two different forms as @xmath99 the proof of the first two statement follows from . for the proof of the third part we recall some useful facts : if @xmath100 , due to the mean value theorem , we get @xmath101 and @xmath102 now the proof follows in a straightforward way . for the last statement we use the cancellation coming from the principal value to define @xmath103 using the uniform convergence of the derivative , we conclude the result . in this section we prove an _ a priori _ bound for @xmath95 . to simplify notation we define @xmath104 [ ivmpf ] let @xmath78 be the initial datum in , define @xmath93 as in and let @xmath95 be the classical solution of corresponding to the initial datum @xmath93 . then @xmath95 verifies @xmath105 moreover , if @xmath82 has a sign then this sign is preserved during the evolution of @xmath95 . changing variables and taking the derivative we obtain that is equivalent to @xmath106 if @xmath107 we define @xmath108 . then we have @xmath109 ( see @xcite for the details ) . if @xmath110 we write @xmath111 and we get @xmath112 . we compute @xmath113 by notational convenience we use the notation @xmath114 and we define @xmath115 evaluating in @xmath116 we have @xmath117 using the definition of @xmath118 and classical trigonometric identities we have @xmath119 putting together all the terms in @xmath120 , we obtain @xmath121[\tan^2(\sigma)+\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]^{-1}}{[(\tan(\sigma)+\tan(\theta))^2+(1-\tan(\sigma)\tan(\theta))^2\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}\\ + \frac{2\tan^2(\sigma)\tan(\theta)[1-\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}{[\tan^2(\sigma)+\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)][(\tan(\sigma)+\tan(\theta))^2+(1-\tan(\sigma)\tan(\theta))^2\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}\\ + \frac{\tan^2(\sigma)\tan(\theta)[1+\tan^2(\theta)\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}{[\tan^2(\theta)+\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)][(\tan(\sigma)+\tan(\theta))^2+(1-\tan(\sigma)\tan(\theta))^2\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}\\ + \frac{2\tan(\sigma)\tan^2(\theta)[1-\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}{[\tan^2(\theta)+\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)][(\tan(\sigma)+\tan(\theta))^2+(1-\tan(\sigma)\tan(\theta))^2\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)]}\\ + \frac{(\tan(\sigma)+\tan(\theta))\tan(\sigma)\tan(\theta)}{(\tan(\sigma)+\tan(\theta))^2+(1-\tan(\sigma)\tan(\theta))^2\|\tanh\|^{2 - 2\epsilon}\left(\frac{\eta}{2}\right)}.\end{gathered}\ ] ] assuming that @xmath122 , then @xmath123 and we obtain @xmath124 and @xmath125 . in the case @xmath126 , we have @xmath127 and we get @xmath128 and @xmath129 . integrating this in time , we get @xmath130 where in the last step we use the definition . in order to prove that the initial sign propagates we observe that if @xmath82 is positive ( respectively negative ) the same remains valid for @xmath93 . assume now that @xmath131 and suppose that the line @xmath132 is reached ( if this line is not reached at any time @xmath41 we are done ) . we write @xmath133 . we have @xmath134 , @xmath135 and we get @xmath128 and @xmath129 . if @xmath136 we denote @xmath137 . we have @xmath138 and @xmath124 . integrating in time we conclude the result . in this section we prove an _ a priori _ bound for @xmath96 . we define @xmath139 where @xmath140 and @xmath118 are defined in and @xmath116 is a critical point for @xmath141 . we will use some bounds for @xmath142 and , for the reader s convenience , we collect them in the following lemma : [ ivlemmu ] let @xmath82 be an initial datum that fulfills , and ( or ) , and let @xmath95 be the solution with initial datum @xmath93 defined in . then for @xmath142 the following inequalities hold 1 . if @xmath143 , due to , we have @xmath144 2 . if @xmath145 , we get latexmath:[\[\label{ivmu1 } 3 . if @xmath145 and @xmath116 is the point where @xmath96 reaches its maximum , @xmath147 4 . if @xmath145 and @xmath148 @xmath149 to prove this lemma we use the following splitting @xmath150 taylor s theorem and the appropriate bounds using proposition [ ivmpf ] . first , we assume @xmath151 . notice that we can take @xmath48 small enough to ensure that @xmath152 defined in also fulfills the hypotheses , and . from , taking one derivative and using lemma [ ivops ] , we get @xmath153 where @xmath154 is the integral corresponding to @xmath155 , @xmath156 is the integral corresponding to @xmath157 and @xmath158 this extra term appear from the regularization present in both @xmath54 . we have @xmath159 where @xmath160 with @xmath161 and @xmath162 the second term is given by @xmath163 where @xmath164 we compute @xmath165 with @xmath166 where @xmath167 and @xmath168 the second term is given by @xmath169 we need to obtain the local decay @xmath170 for @xmath171 . assuming the classical solvability for with an initial datum @xmath82 fulfilling the hypotheses , and we have that @xmath172 also fulfills , and if @xmath173 is small enough . recall that @xmath174 and @xmath175 . the linear terms in have the appropriate sign and they will be used to control the the positive contributions of the nonlinear terms . we need to prove that @xmath176 . for the sake of simplicity , we split the proof of this inequality in different lemmas . [ ivlemdf1 ] if @xmath177 , we have @xmath178 using the linear term @xmath179 to control , we have @xmath180 if @xmath181 . due to @xmath174 , we have @xmath182 . then , the term is @xmath183 the term is @xmath184 this kind of terms will be absorbed by @xmath185 . we have to deal with @xmath154 . we start with the term corresponding to @xmath186 in . we write @xmath187 [ ivlemdf2 ] if @xmath177 , we have @xmath188 we split @xmath189 since @xmath190 is small enough to ensure that the hypotheses , and hold at time @xmath191 , we have that , if @xmath192 , @xmath193 is not singular and can be bounded using and : @xmath194 we compute @xmath195 with @xmath196 and @xmath197 using the mean value theorem , we bound the inner term @xmath198 as @xmath199 due to , the outer term is @xmath200 putting all together , we obtain @xmath201 then , using the diffusion given by @xmath97 to control @xmath202 , we get @xmath203 due to @xmath204 and @xmath48 , some terms have the appropriate sign : @xmath205 thus we can neglect their contribution . furthermore , we have @xmath206 taking @xmath207 and using the mean value theorem , we get @xmath208 combining these terms we conclude this result . the term corresponding to @xmath209 in is @xmath210 [ ivlemdf3 ] if @xmath211 , we have @xmath212 the proof follows the same ideas as in lemma [ ivlemdf2 ] . we are done with @xmath209 , thus , using the previous bound for @xmath213 , we are done with @xmath214 in . the terms in @xmath215 are not multiplied by @xmath60 and we have to obtain this decay from the integral . we write @xmath216 [ ivlemdf4 ] we have @xmath217 we have @xmath218 with @xmath219 @xmath220 and @xmath221 the term @xmath222 is not singular and can be bounded using and as follows : @xmath223 we can bound @xmath224 in the same way , @xmath225 we split the term @xmath226 as follows @xmath227 where @xmath228 and @xmath229 to bound @xmath230 we need to use the diffusion coming from @xmath65 . notice that , according to lemma [ ivops ] , we have @xmath231 and , when evaluating in the point where @xmath232 reaches its maximum , the first two terms are positive and they can be neglected . we get @xmath233 where in the last step we have used the previous splitting in @xmath234 and @xmath235 , and . this concludes the result . now that we have finished with @xmath236 , the term with @xmath237 is @xmath238 we have [ ivlemdf5 ] if @xmath239 , we have @xmath240 the proof is similar to the proof of lemma [ ivlemdf4 ] and , for the sake of brevity , omit it . in order to finish bounding @xmath215 in , we have to bound the term @xmath241 this term , akin to the singular term in @xcite , is bounded using the hypotheses and . [ ivlemdf6 ] using , and , we obtain @xmath242 using classical trigonometric identities we can write @xmath243 @xmath244 and @xmath245 therefore , as in @xcite , the sign of @xmath246 is the same as the sign of @xmath247 the roots of @xmath248 are @xmath249 and @xmath250 , so , if we have @xmath251 then we can ensure that this contribution is negative . since , we get @xmath252 using the cancellation when @xmath253 , we obtain @xmath254 where @xmath255 we remark that @xmath256 . we consider the cases given by the sign and the size of @xmath257 . _ 1 . case @xmath258 : _ in this case , we have @xmath259 and @xmath260 . using the definition of @xmath140 in and the fact that @xmath261 , we have ( see lemma [ ivlemmu ] ) . notice that , in this case , we have @xmath262 and we get . due to and we obtain @xmath263 _ 2 . case @xmath264 : _ in this case we have @xmath265 and @xmath266 . therefore , we get @xmath267 and we can neglect it . case @xmath268 : _ we remark that in this case we have @xmath265 and @xmath269 . we split @xmath270 the last term is now positive due to the definition of @xmath271 . then , in this case , we have @xmath272 and we can neglect its contribution . using taylor s theorem in we obtain the bound and . we are done with @xmath154 in and now we move on to @xmath156 . these terms are easier because the integrals are not singular . with the same ideas as before we can bound the term involving @xmath273 : [ ivlemdf7 ] the contribution of @xmath273 is bounded by @xmath274 the proof is straightforward . we are left with @xmath275 in . first , we consider @xmath276 [ ivlemdf8 ] the term @xmath277 is bounded as @xmath278 using classical trigonometric identities , we compute @xmath279 we have to bound the terms containing @xmath280 . these terms are @xmath281 to obtain the decay with @xmath60 we split the integral in the regions @xmath234 and @xmath282 as before . [ ivlemdf9 ] the terms @xmath283 and @xmath284 are bounded by @xmath285 using this splitting , @xmath48 , , and , we get @xmath286 with the same ideas and using , we have @xmath287 in order to estimate the decay with @xmath60 of these integrals we compute @xmath288 and @xmath289 we have the following result concerning the evolution of the slope : [ ivmpdf ] let @xmath290 be the initial datum in satisfying , and , define @xmath93 as in and let @xmath95 be the classical solution of corresponding to the initial datum @xmath93 . then @xmath95 verifies @xmath291 for the sake of simplicity we split the proof in different steps . * step 1 ( local decay ) : * combining @xmath292 in and @xmath277 in lemma [ ivlemdf8 ] , and using the bounds and and the hypothesis we obtain @xmath293 we take @xmath294 , @xmath295 , @xmath296 . since we have a term @xmath297 and @xmath48 , we can compare the bounds in lemmas [ ivlemdf1]- [ ivlemdf9 ] with @xmath298 if @xmath299 is chosen big enough . the universal constant @xmath42 in all these bounds can be @xmath300 . we have shown that for every @xmath190 small enough , there is local in time decay . as @xmath191 is positive and arbitrary , we have @xmath301 * step 2 ( from local decay to an uniform bound ) : * then , in the worst case , we have @xmath302 these inequalities ensure that the hypotheses , and hold at time @xmath303 and @xmath304 decays again . * step 3 ( the case where @xmath305 ) : * this case follows the same ideas , and we conclude , thus , the result . [ ivubdf ] let @xmath290 be the initial datum in satisfying and define @xmath93 as in . let @xmath95 be the classical solution of corresponding to the initial datum @xmath93 . then , @xmath95 verifies @xmath306 the region delimited by @xmath28 is below the region with maximum principle ( see @xcite ) . then , in the worst case , at some @xmath307 we have that@xmath308 fulfills the hypotheses , and . from them the result follows . in this section we obtain _ a priori _ estimates in @xmath309 that ensure the global existence for the regularized systems for initial data satisfying hypotheses , and or . first , notice that if the initial datum satisfies hypotheses , and , by propositions [ ivmpf ] and [ ivmpdf ] , the solution satisfies @xmath310 if the initial datum satisfies , by propositions [ ivmpf ] and [ ivubdf ] , the solution to the regularized system again satisfies the bounds . then we have the following proposition : [ existence ] let @xmath290 be the initial datum in satisfying , and or and define @xmath93 as in . then for every @xmath311 and @xmath312 there exists a solution @xmath313,h^3({\mathbb r}))$ ] . we have to bound the @xmath18 norm of the function and its third derivative . we split the proof in different steps . * step 0 ( local existence ) : * the local existence follows by classical energy methods as in @xcite . * step 1 ( the function ) : * we have @xmath314 using we get @xmath315 and we obtain that the contribution of the linear terms is negative . the nonlinear term @xmath155 defined in is @xmath316 using the cancellation coming from the principal value we have @xmath317 inserting and in the expression for @xmath318 we obtain @xmath319 the second term in @xmath320 is @xmath321 using the cauchy schwarz inequality , the equality @xmath322 and integrating by parts we get @xmath323 to finish with the @xmath18 norm we have to deal with @xmath324 . we have @xmath325 where @xmath118 is defined in . using the same ideas as in @xmath320 and @xmath326 we conclude the bound @xmath327 putting all these bounds together we get @xmath328 * step 2 ( the third derivative ) : * to study the @xmath18 norm of the third derivative , we compute @xmath329 the term @xmath330 is positive due to lemma [ ivops ] : @xmath331 the nonlinear terms related to @xmath140 are @xmath332 dx = a_3+a_4.\ ] ] the term @xmath333 is not singular if @xmath311 and can be bounded using hlder and nirenberg interpolation inequalities . for the sake of brevity , we write some terms detailedly , being the rest analogous to them . we have @xmath334 using @xmath335 we obtain @xmath336 the second term is @xmath337 and using the classical interpolation inequality @xmath338 we get @xmath339 we split the term @xmath340 as follows @xmath341 these terms are not singular because of the domain of integration . we have to deal with the integrability at infinity in @xmath342 . we compute @xmath343 the integrability at infinity is obtained using and . we only bound the more singular terms in @xmath344 and @xmath345 . the most singular term in @xmath344 is @xmath346 using , and , we obtain @xmath347 analogously , the more singular term in @xmath345 is @xmath348 using the same bounds as in @xmath349 , we get @xmath350 using classical trigonometric identities , we obtain @xmath351 and the most singular term in @xmath222 is @xmath352 using the cancellation of the principal value integral we obtain @xmath353 thus , @xmath354 integrating by parts in @xmath355 , we obtain the required decay at infinity and we conclude @xmath356 putting all together , we get @xmath357 the nonlinear terms related to @xmath118 are @xmath358 we observe that , due to @xmath359 and @xmath360 , this integral is not singular . thus the inner part @xmath361 can be bounded following the same ideas as for @xmath333 . the integrability at infinity is obtained with the following splitting @xmath362 the term @xmath363 is @xmath364 and it can be handled as @xmath222 . the terms @xmath224 and @xmath226 have a term @xmath365 and they can be bounded following the steps in @xmath344 and @xmath345 by using and . putting all the estimates together we obtain @xmath366 using , young s inequality and the dissipation given by the laplacian we get the _ a priori _ estimate @xmath367 a classical continuation argument shows the global existence . in this section we study the limit of @xmath95 as @xmath89 . [ ivlemafeps ] the regularized solutions @xmath95 corresponding to an initial datum satisfying the hypotheses , and , or , converge ( up to a subsequence ) weakly- * to @xmath368,w^{1,\infty}({\mathbb r}))$ ] . moreover , up to a subsequence , @xmath369 in @xmath370 for all compact set @xmath371 . first , notice that , due to propositions [ ivmpf ] , [ ivmpdf ] and [ ivubdf ] and hypotheses , and , the regularized solutions satisfy @xmath372 while , if the initial datum , instead of hypotheses , and , satisfies then @xmath373 due to the banach - alaoglu theorem , these bounds imply that there exists a subsequence such that @xmath374 and @xmath375 with @xmath368,w^{1,\infty}({\mathbb r}))$ ] , any @xmath376\times{\mathbb r})$ ] and every @xmath312 . fixing @xmath41 , due to the uniform bound in @xmath37 and the ascoli - arzela theorem we have that , up to a subsequence , @xmath377 uniformly on any bounded interval @xmath378 . moreover , for all @xmath379 , we have @xmath380)}\rightarrow0.\ ] ] in order to prove this uniform convergence on compact sets we use the spaces and results contained in @xcite . for @xmath381 , we define the norm @xmath382 we define the banach space @xmath383 as the completion of @xmath384 with respect to the norm . we have @xmath385 the embedding @xmath386 is continuous and , due to the ascoli - arzela theorem , the embedding @xmath387 is compact . we use the following lemma [ ivlemapaco ] consider a sequence @xmath388\times b(0,n))$ ] that is uniformly bounded in the space @xmath51,w^{1,\infty}(b(0,n)))$ ] . assume further that the weak derivative @xmath389 is in @xmath51,l^\infty(b(0,n)))$ ] ( not necessarily uniform ) and is uniformly bounded in . finally suppose that @xmath390\times b(0,n))$ ] . then there exists a subsequence of @xmath391 that converges strongly in @xmath51\times b(0,n)).$ ] due to this lemma we only need to bound @xmath392 in @xmath51\times b(0,n))$ ] ( not uniformly ) and in @xmath51,w^{-2,\infty}_(b(0,n)))$ ] ( uniformly ) . using that @xmath393,h^3({\mathbb r}))$ ] , the linear terms in can be bounded easily with a bound depending on @xmath60 . to bound the nonlinear terms we split the integral @xmath394 and we compute @xmath395 where we have used @xmath396 , , and @xmath397 the second term with the kernel involving @xmath140 is @xmath398 the terms with the kernel involving @xmath118 are not singular and can be bounded following the same ideas @xmath399 and @xmath400 putting together all these estimates we get @xmath401 thus we conclude with the bound in @xmath51\times b(0,n))$ ] . to obtain the bound in @xmath51,w^{-2,\infty}_(b(0,n)))$ ] we extend @xmath402 by zero outside of this ball of radius @xmath379 . then , using lemma [ ivops ] , we _ integrate by parts _ and obtain @xmath403 @xmath404 @xmath405 and @xmath406 using @xmath407 we bound the linear terms in as @xmath408 being @xmath42 a universal constant . the nonlinear terms are @xmath409 and @xmath410 using the boundedness of @xmath411 , we get @xmath412 the outer part is not singular and can be bounded ( as it was done before ) applying @xmath70 . we get @xmath413 putting together all these bounds we obtain @xmath414}\\|{\partial_t}f(t)\\|_{w^{-2,\infty}_(b(0,n))}\leq c\left(\\|f_0\\|_{l^\infty({\mathbb r})}\right).\ ] ] using lemma [ ivlemapaco ] , we conclude the result . looking at we give the following definition first , we deal with the linear terms . using the weak- convergence in @xmath51,w^{1,\infty}({\mathbb r}))$ ] and lemma [ ivops ] , we obtain @xmath418 @xmath419 @xmath420 @xmath421 and @xmath422 where , in the last step , we use the dominated convergence theorem and the @xmath423 convergence of the mollifier . to deal with the nonlinear terms we split the integrals @xmath424 for sufficiently small @xmath191 and large enough @xmath379 . these parameters , @xmath425 , that will be fixed below , can depend on @xmath82 but they do nt depend on @xmath60 . for the inner part of the integrals , we get @xmath426\times{\mathbb r})}.\end{gathered}\ ] ] the outer integral goes to zero as @xmath379 grows . we compute @xmath427 as @xmath428 , the integrals are not singular and we only have to deal with the decay at infinity . using , , , the bound @xmath70 , integrating by parts and using the extra decay coming from the principal value at infinity ( see , for instance , the term @xmath429 in proposition [ existence ] in section [ ivsecglobal ] ) , we have @xmath430 the only thing to check is the convergence of @xmath431 . due to the compactness of the support of @xmath55 , we have @xmath432 with @xmath433 large enough to ensure @xmath434 . since we have ( up to a subsequence ) that @xmath369 uniformly on compact sets ( see lemma [ ivlemafeps ] ) , the uniform convergence @xmath435 if @xmath436 and the continuity of all the functions in this integral , the limit in @xmath60 and the integral commute and we get @xmath437	in this paper we show global existence of lipschitz continuous solution for the stable muskat problem with finite depth ( confined ) and initial data satisfying some smallness conditions relating the amplitude , the slope and the depth . the cornerstone of the argument is that , for these _ small _ initial data , both the amplitude and the slope remain uniformly bounded for all positive times . we notice that , for some of these solutions , the slope can grow but it remains bounded . this is very different from the infinite deep case , where the slope of the solutions satisfy a maximum principle . our work generalizes a previous result where the depth is infinite . * keywords * : darcy s law , inhomogeneus muskat problem , well - posedness . * acknowledgments * : the author is supported by the grant mtm2011 - 26696 from ministerio de economa y competitividad ( mineco ) . the author thanks david paredes and professors diego crdoba and rafael orive for comments that greatly improved the manuscript . the author is grateful to reviewers for their helpful suggestions .
the lattice dimensionality is a key parameter in materials science . one dimensional ( 1d ) systems with half filled band undergo a structural distortion @xcite which increases the elastic energy and opens a gap at the fermi surface thus lowering the electronic energy . the competition between lattice and electronic subsystems stabilizes the 1d structure which accordingly acquires semiconducting properties whereas the behavior of the 3d system would be metallic like . conjugated polymers , take polyacetylene as prototype , show anisotropic electrical and optical properties @xcite due to intrinsic delocalization of @xmath2 electrons along the chain of ch units . as the intrachain bonding between adjacent ch monomers is much stronger than the interchain coupling the lattice is quasi-1d . hence , as a result of the peierls instability , polyacetylene shows an alternation of short and long neighboring carbon bonds , a dimerization , accompanied by a two fold degenerate ground state energy . the su - schrieffer - heeger ( ssh ) model hamiltonian @xcite has become a successful tool in polymer physics as it hosts the peculiar ground state excitations of the 1d conjugated structure and it accounts for a broad range of polymer properties @xcite . as a fundamental feature of the ssh hamiltonian the electronic hopping integral linearly depends on the relative displacement between adjacent atomic sites thus leading to a nonlocal _ e - ph _ coupling with vertex function depending both on the electronic and the phononic wave vector . the latter property induces , in the matsubara formalism @xcite , an electron hopping associated with a time dependent lattice displacement . as a consequence time retarded electron - phonon interactions arise in the system yielding a source current which depends both on time and on the electron path coordinates . this causes large _ e - ph _ anharmonicities in the equilibrium thermodynamics of the ssh model @xcite . hopping of electrons from site to site accompanied by a coupling to the lattice vibration modes is a fundamental process @xcite determining the transport @xcite and equilibrium properties @xcite of many body systems . a variable range hopping may introduce some degree of disorder thus affecting the charge mobility @xcite and the thermodynamic functions . this paper focusses on this issue , dealing with the thermodynamical properties of the ssh model in two dimensions and comparing them with the results obtained in one dimension @xcite . only a few extensions of the ssh hamiltonian to higher dimensionality appear in the literature @xcite mainly concerning the phase diagrams @xcite and the ground state excitations @xcite . we apply a path integral method @xcite which fully accounts for the time retarded _ e - ph _ interactions and , exploiting the above mentioned hamiltonian linear dependence on the atomic displacement , allows us to derive the electron - phonon source action in two dimensions . the general formalism is outlined in section ii while the results are reported on in section iii . the conclusions are drawn in section iv . in a square lattice with isotropic nearest neighbors hopping integral @xmath3 , the ssh hamiltonian for electrons plus _ e - ph _ interactions reads : @xmath4 \ , \nonumber \\ ( j_{r , s})_x=\ , & & - { 1 \over 2}\bigl [ j - \alpha \delta u_x \bigr ] \ , \nonumber \\ ( j_{r , s})_y=\ , & & - { 1 \over 2}\bigl [ j - \alpha \delta u_y \bigr ] \ , \nonumber \\ \delta u_x=\ , & & u_x(r+1,s ) - { u_{x}(r , s ) } \ , \nonumber \\ \delta u_y=\ , & & u_y(r , s+1 ) - { u_{y}(r , s ) } \ , \nonumber \\\end{aligned}\ ] ] where @xmath5 is the electron - phonon coupling , @xmath6 is the dimerization coordinate indicating the displacement of the monomer group on the @xmath7 lattice site , @xmath8 and @xmath9 create and destroy electrons ( i.e. , @xmath2 band electrons in polyacetylene ) . the phonon hamiltonian is given by a set of 2d classical harmonic oscillators . the two addenda in ( 1 ) deal with one dimensional _ e - ph _ couplings along the _ x _ and _ y _ axis respectively , with first neighbors electron hopping . second neighbors hopping processes ( with overlap integral @xmath10 ) may be accounted for by adding to the hamiltonian the term @xmath11 such that @xmath12 \ , \nonumber \\\end{aligned}\ ] ] the real space hamiltonian in ( 1 ) can be transformed into a time dependent hamiltonian @xcite by introducing the electron coordinates : i ) @xmath13 at the @xmath14 lattice site , ii ) @xmath15 at the @xmath16 lattice site and iii ) @xmath17 at the @xmath18 lattice site , respectively . @xmath19 and @xmath20 vary on the scale of the inverse temperature @xmath21 . the spatial _ e - ph _ correlations contained in ( 1 ) are mapped onto the time axis by changing : @xmath22 , @xmath23 and @xmath24 . now we set @xmath25 , @xmath26 , @xmath27 . accordingly , ( 1 ) transforms into the time dependent hamiltonian : @xmath28 \ , \nonumber \\ j_{y}(\tau)&=&\,- { 1 \over 2}\bigl[j - \alpha u_{y}(\tau ) \bigr ] \ , \nonumber \\\end{aligned}\ ] ] while the ground state of the 1d ssh hamiltonian is twofold degenerate , the degree of phase degeneracy is believed to be much higher in 2d @xcite as many lattice distortion modes contribute to open the gap at the fermi surface . nonetheless , as in 1d , these phases are connected by localized and nonlinear excitations , the soliton solutions . thus , also in 2d both electron hopping between solitons @xcite and thermal excitation of electrons to band states may take place within the model . these features are accounted for by the time dependent version of the hamiltonian . as @xmath19 varies continuously on the @xmath21 scale and the @xmath19-dependent displacement fields are continuous variables ( whose amplitudes are in principle unbound in the path integral ) , long range hopping processes are automatically included in @xmath29 which is therefore more general than the real space ssh hamiltonian in ( 1 ) ( and ( 2 ) ) . thus by means of the path integral formalism we look at the low temperature thermodynamical behavior both in 1d and 2d searching for those features which may be ascribable to some local disorder related to the variable range of the hopping processes . the semiclassical nature of the model is evident from ( 3 ) in which quantum mechanical degrees of freedom interact with the classical variables @xmath30 . averaging the electron operators over the ground state we obtain the time dependent semiclassical energy per lattice site @xmath31 : @xmath32 \cosh(\epsilon_{\bf k } \tau ) n_f(\epsilon_{\bf k } ) \ , \nonumber \\\end{aligned}\ ] ] with @xmath33 and @xmath34 in the first and second addendum respectively . @xmath35 is the electron dispersion relation and @xmath36 is the fermi function . eq.(4 ) can be rewritten in a way suitable to the path integral approach by defining @xmath37 @xmath38 and @xmath39 are the effective terms accounting for the @xmath19 dependent electronic hopping while @xmath40 is interpreted as the external source @xcite current for the oscillator field @xmath41 . averaging the electrons over the ground state we neglect the fermion - fermion correlations @xcite . this approximation however is not expected to affect substantially the following calculations . taking a bath of @xmath42 2d oscillators , we generally write the ssh electron path integral , @xmath43 , as : @xmath44 \ , \nonumber \\ & \cdot & exp\biggl[- \int_0^{\beta}d\tau \biggl({m \over 2 } \dot{\bf \zeta}^2(\tau ) + v\bigl({x}(\tau)\bigr ) + v\bigl({y}(\tau)\bigr ) \biggr)\ , \nonumber \\ & + & \sum_{i=1}^{\bar n } { \bf u}_i(\tau ) \cdot { \bf j}({\bf v}(\tau ) ) \biggr ] \ , \nonumber \\\end{aligned}\ ] ] where @xmath45 is the electron mass , @xmath46 is the atomic mass and @xmath47 is the oscillator frequency . as a main feature we notice that the interacting energy is linear in the atomic displacement field . then , the electronic path integral can be derived after integrating out the oscillator degrees of freedom which are decoupled along the @xmath48 and @xmath49 axis . thus , we get : @xmath50 \biggr]^2 , \ , \nonumber \\ & & a ( { x}(\tau))=\ , -{{\hbar^2 } \over { 4m}}\sum_{i=1}^{\bar n } { 1 \over { \hbar \omega_i \sinh(\hbar\omega_i\beta/2 ) } } \ , \nonumber \\ & \cdot & \int_0^{\beta } d\tau j({x}(\tau ) ) \int_0^{\beta}d{\tau '' } \cosh\bigl(\hbar\omega_i \bigl ( \|\tau - { \tau''}\| - \beta/2 \bigr ) \bigr ) j({x}({\tau '' } ) ) , \ , \nonumber \\ & & z_i=\ , \bigl [ { 1 \over { 2\sinh(\hbar\omega_i\beta/2 ) } } \bigr]^2 \ , \nonumber \\\end{aligned}\ ] ] and the 2d electron path integral is obtained after squaring the sum over one dimensional electron paths . this permits to reduce the computational problem which is nonetheless highly time consuming particularly in the low temperature limit . note infact that the source current @xmath51 requires integration over the 2d brillouin zone ( bz ) according to ( 4 ) and ( 5 ) and this occurs for any choice of the electron path coordinates . the quadratic ( in the coupling @xmath5 ) source action @xmath52 is time retarded as the particle moving through the lattice drags the excitations of the oscillator fields which take a time to readjust to the electron motion . when the interaction is sufficiently strong the conditions for polaron formation @xcite may be fulfilled in the system according to the degree of adiabaticity @xcite . however the present path integral description is valid independently of the existence of polarons as it applies also to the weak coupling regime . assuming periodic conditions @xmath53 , the particle paths can be expanded in fourier components @xmath54 and the open ends integral over the paths @xmath55 transforms into the measure of integration @xmath56 . taking : @xmath57\ ] ] which normalizes the kinetic energy term in ( 7 ) , we proceed to integrate ( 7 ) in order to derive the full partition function of the system versus temperature . taking a dimensionless path ( in units of the lattice constant @xmath58 ) with : @xmath59 and @xmath60 the functional measure in ( 9 ) can be rewritten as : @xmath61 where @xmath62 is the cutoff on the fourier components in ( 8) . at a given temperature and for a given set of particle path coefficients we first carry out the momentum integrations required by ( 4 ) summing over 1600 points in the reduced 2d bz . this prepares the source current @xmath51 to be used in the double time integration which yields , according to ( 7 ) , the source action @xmath52 . afterwards we sum the exponential of the total action over an appropriate set of particle paths . numerically stable results are achieved by tuning the number of pairs @xmath62 in ( 10 ) , the cutoff ( @xmath63 ) on the integration range of the fourier coefficients in ( 10 ) and the related number of points ( @xmath64 ) in the measure of integration . @xmath65 suffices in the fourier expansion of the path . the argument to set the cutoff @xmath63 on the integration range is based on the fact that the functional measure normalizes the kinetic term in ( 7 ) : @xmath66 = \,1\ ] ] from ( 11 ) we get , @xmath67 hence , @xmath63 scales versus temperature as @xmath68 . computing ( 7 ) with ( 10 ) we find that infact a smaller cutoff , @xmath69 , guarantees convergence in the interacting partition function . accordingly the number of points in each integration range has to increase by decreasing @xmath0 as @xmath70 . in the 2d problem , @xmath71 points for each fourier coefficient are required in ( 10 ) . then we carry out the total 2d path integral by evaluating the contribution of @xmath72 paths selecting the integer part of @xmath64 at any temperature . computation of the second order derivatives of the free energy in the range @xmath73k$ ] , with a spacing of 3k , takes 55hours and 15 minutes on a pentium 4 . the physical input parameters characterizing both the 1d and 2d models are : the bare hopping integral @xmath3 , the oscillator frequencies @xmath47 and the effective coupling @xmath74 ( in units @xmath75 ) . we take here a narrow band system with @xmath76 . the crossover between weak and strong _ e - ph _ coupling is given in our model by @xmath77 @xcite , where @xmath78 is the largest energy in the oscillators bath . let s assume a bath of @xmath79 low energy oscillators with @xmath80 thus setting the crossover value at @xmath81 . a higher number of oscillators in the same energy range would not significantly modify the results hereafter presented whereas lower @xmath47 values would yield a larger contribution to the phonon partition function mainly at low @xmath0 . in fig.1 , a comparison between the 1d and the 2d free energies is presented for two values of @xmath82 , one lying in the weak and one in the strong _ e - ph _ coupling regime . the oscillator free energies @xmath83 are plotted separately while the free energies arising from the total action in ( 7 ) , shortly termed @xmath84 , are due to the purely electronic plus _ e - ph _ contributions . in general , the 2d free energies have a larger gradient ( versus temperature ) than the corresponding 1d terms . the @xmath83 lie above the @xmath84 both in 1d and 2d because of the choice of the @xmath85 ; lowering the energy of some oscillator modes would produce a crossing point between @xmath83 and @xmath84 with temperature location depending on the value of @xmath82 . note that the @xmath84 plots have a positive temperature derivative in the low temperature regime and this feature is more pronounced in 2d . infact , at low @xmath0 , the source action is dominated by the hopping potential @xmath86 while , at increasing @xmath0 , the _ e - ph _ effects become progressively more important ( as the bifurcation between the @xmath87 and @xmath88 curves shows ) and the @xmath84 get a negative derivative . in 2d , the weight of the @xmath86 term is larger because there is a higher hopping probability . this physical expectation is taken into account by the path integral method . at any temperature , we monitor the ensamble of relevant particle paths over which the hopping potential is evaluated . for a selected set of fourier components in ( 10 ) the hopping decreases by lowering @xmath0 but its value is still significant at low @xmath0 . considered that : i ) the total action is obtained after a @xmath89 integration of @xmath86 and ii ) the @xmath89 integration range is larger at lower temperatures , we explain why the overall hopping potential contribution to the total action is responsible for the anomalous free energy behavior at low t. in fig.2 , the heat capacity contributions due to the oscillators ( @xmath90 ) and electrons plus _ e - ph _ coupling ( @xmath91 ) are reported on . the values are normalized over @xmath42 . the dimensionality effects are seen to be large and , for a given dimensionality , the role of the _ e - ph _ interactions is magnified at increasing @xmath0 . the total heat capacity ( @xmath90 + @xmath91 ) over @xmath0 ratios are plotted in fig.3 in the low @xmath0 regime to emphasize the presence of an anomalous upturn which appears at @xmath92 in 1d and @xmath93 in 2d . this feature in the heat capacity linear coefficient is ultimately related to the above discussed overall hopping potential effects while the strength of the coupling @xmath82 has a minor role in the low @xmath0 limit . finally we emphasize that the degeneracy of the ground state scarcely affects the finite temperature equilibrium properties . this can be shown in 1d by treating the ssh hamiltonian in mean field approximation with a staggered dimerization amplitude which yields the size of the lattice distortion . by diagonalizing the hamiltonian in momentum space we see that the competition between electron and lattice energy determines the contribution to the partition function and shifts downwards the free energy versus temperature . hence , the heat capacity due to the degeneracy is very small ( in comparison with the values given in fig.2 ) and strictly linear in @xmath0 . accordingly , the peculiar and sizeable upturn in the heat capacity over t ratio does not depend on the nature of the ground state . the inclusion of the quantum lattice fluctuations beyond the mean field approximation is expected to reinforce our conclusion as the fluctuations tend to suppress the long range order associated with the dimerized structure @xcite . we have investigated the thermodynamics of a semiclassical su - schrieffer - heeger ( ssh ) hamiltonian extended to two dimensions in which the lattice displacements are taken as classical variables . by mapping the real space interactions onto the time ( @xmath19 ) scale we deal with an hamiltonian accounting for a variable range of the electronic transfer integral . such a model can be suitably treated in the path integral formalism as both the electron and lattice degrees of freedom are continuous functions of @xmath19 which varies up to the inverse temperature . although we are thus assuming a continuum model the present path integral calculation is valid for any range of _ e - ph _ coupling and for any value of physical parameters . the two dimensional path integral for the ssh particle in a bath of harmonic oscillators has been derived by summing over a set of electron paths that becomes increasingly large at decreasing temperatures . thus , the computation of the total action and of the free energy derivatives is highly time consuming in the low @xmath0 regime where the quantum mechanical nature of the electronic subsystem is fully accounted for . at low @xmath0 , the electron hopping integral largely shapes the free energy which displays an anomalous positive derivative more pronounced in 2d than in 1d . this behavior is mirrored in the upturn of the _ heat capacity over t _ ratio . for a given set of input parameters characterizing electron band , oscillator energies and _ e - ph _ couplings , we find that the upturn shows up at @xmath93 in the square lattice and at @xmath92 in the linear chain . also the size of the upturn is larger in two dimensions consistently with the fact that the 2d hopping potential contribution to the total action is higher than in 1d . the presented results have been obtained assuming a narrow electron band ( @xmath94 ) but the upturn in the _ heat capacity over t _ ratio appears also in the case of wide band systems . the degenerate ground state of the ssh model is not responsible for the anomalous behavior of the heat capacity over @xmath0 ratio . infact , closing the linear chain in a ring geometry , one can easily diagonalize the 1d hamiltonian in mean field approximation and show that the contribution to the total heat capacity , due to the two fold degeneracy of the ground state , is small and linear versus temperature . a quantum description of the lattice displacement fields should further reduce the role of the dimerized ground state as the fluctuations tend to suppress the order parameter . similar conclusions are expected to hold in the 2d case where , however , the degeneracy of the ground state is manyfold . in conclusion , while in the low @xmath0 regime the thermodynamics is essentially governed by the electronic term of the hamiltonian , at increasing @xmath0 the strength of _ e - ph _ coupling progressively determines the slope of the free energy and heat capacity . our results suggest that a glassy like thermodynamical behavior is a peculiar feature of hamiltonian models with a variable range for the electronic hopping . such a behavior is more likely to occur in higher dimensionality as the structure is more closely packed and , accordingly , the hopping probability is larger . r.e.peierls , quantum theory of solids , clarendon ( oxford ) 1955 . yu lu , _ solitons and polarons in conducting polymers _ world scientific , singapore ( 1988 ) . w.p.su , j.r.schrieffer , a.j.heeger , phys . lett . * 42 * , 1698 ( 1979 ) . a.j . heeger , s.kivelson , j.r.schrieffer , w .- p.su , rev . phys . * 60 * , 781 ( 1988 ) . g.d.mahan , _ many particle physics _ , plenum press , n.y . m.zoli , phys.rev.b * 70 * , 184301 ( 2004 ) . j.t.devreese , _ polarons _ in encyclopedia of applied physics ( vch publishers , ny ) * 14 * , 383 ( 1996 ) . i.j.lang , yu.a.firsov , sov . jetp * 16 * , 1301 ( 1963 ) . raedt , a.lagendijk , phys . rev . b * 27 * , 6097 ( 1983 ) . i.i.fishchuk , a.kadashchuk , h.bssler , s.nepurek , phys.rev . b * 67 * , 224303 ( 2003 ) . m.zoli , phys.rev.b * 67 * , 195102 ( 2003 ) . s.tang , j.e.hirsch , phys . b * 37 * , 9546 ( 1988 ) . k.k.voo , c.y.mou , cond - mat/0309298 m.t.figge , m.mostovoy , j.knoester , phys . . lett . * 86 * , 4572 ( 2001 ) . n.miyasaka,y.ono , j.phys.soc.jpn . * 70 * , 2968 ( 2001 ) . m.zoli , phys . b * 66 * , 012303 ( 2002 ) . r.p.feynman , phys . rev . * 97 * , 660 ( 1955 ) . d.r.hamann , phys . b * 2 * , 1373 ( 1970 ) . y.ono , t.hamano , j.phys.soc.jpn . * 69 * , 1769 ( 2000 ) . s.kivelson , phys . 46 * , 1344 ( 1981 ) . h.kleinert , _ path integrals in quantum mechanics , statistics and polymer physycs _ world scientific publishing , singapore ( 1995 ) . j.e.hirsch , phys . 51 * , 296 ( 1983 ) . v.m.stojanovi , p.a.bobbert , m.a.j.michels , phys . b * 69 * , 144302 ( 2004 ) . d.w.brown , k.lindenberg , y.zhao , j.chem.phys . * 107 * , 3179 ( 1997 ) . v.perebeinos , p.b.allen , j.napolitano , solid state commun . * 118 * , 215 ( 2001 ) .	the equilibrium thermodynamics of the two dimensional su - schrieffer - heeger model is derived by means of a path integral method which accounts for the variable range of the electronic hopping processes . while the lattice degrees of freedom are classical functions of time and are integrated out exactly , the electron particle paths are treated quantum mechanically . the free energy of the system and its temperature derivatives are computed by summing at any @xmath0 over the ensemble of relevant particle paths which mainly contribute to the total partition function . in the low @xmath0 regime , the _ heat capacity over t _ ratio shows an upturn peculiar of a glassy like behavior . this feature is more sizeable in the square lattice than in the linear chain as the overall hopping potential contribution to the total action is larger in higher dimensionality . #
for many years , observations of the diffuse interstellar medium ( ism ) have traced a whole hierarchy of structures on spatial scales @xmath2 pc ( cf . @xcite ) , often attributed to the ism turbulence . however , the extremely small - scale end of this spectrum , reaching down to scales of tens to hundreds of aus , has been largely unexplored . as the kinetic energy cascades from larger to smaller scales , it is expected that the turbulent spectrum will reach its end at so - called `` dissipation scales . '' understanding of the dissipative properties of turbulence , such as its scale , efficiency , dissipation rate , and ubiquity , are important since the delicate balance between energy injection and dissipation has a profound impact on the stellar ism recycling chain . however , many questions still await both theoretical understanding and observational constraints . for example , how is the balance between turbulent energy injection and dissipation maintained ? what is the efficiency of the dissipation process ? does it vary with time and galactocentric radius ? is dissipation located only in pockets of `` active turbulence '' , or it is ubiquitous ? how do dissipative properties vary with galactic environments ? starting in 1976 @xcite , structure in the diffuse ism has been observed sporadically on scales down to @xmath3 au . these findings caused much controversy as the observationally inferred properties were not in accord with the standard ism picture . with a typical , observationally inferred , hi volume density of @xmath4 @xmath5 , and the thermal pressure of @xmath6 @xmath5 k ( assuming temperature of @xmath74070 k ) , the au - scale `` tiny '' features appeared significantly over - dense and over - pressured compared with the traditional cold neutral medium ( cnm ) clouds . since the initial observations , the au - sized structures came to be rather frequently detected , with claims that up to 15% of the cnm could be in this form @xcite . there has been considerable progress in recent years in understanding of the ism turbulence and the small - scale structure in the ism , from both theoretical and observational perspectives [ for reviews see @xcite ] . nevertheless , it remains unclear whether the observed au - scale features represent the dissipation scales of the @xmath2 pc spatial scale turbulence . at the same time , the observational picture has become even more complicated with improved telescope sensitivity , resulting in the detection of new populations of possibly related sub - parsec ism clouds @xcite . the `` low column density clouds '' observed by stanimirovic & heiles ( 2005 ) have a peak optical depth of only @xmath8 to @xmath9 and an inferred size of 5005000 au . motivated by the recent theoretical and observational efforts in understanding the nature and origin of the tiny - scale ism structure , we have undertaken multi - epoch observations of the hi absorption against a set of bright pulsars . because of pulsars relatively high proper motion and transverse speeds ( typically @xmath10 au yr@xmath0 ) , pulsar hi absorption profiles obtained at different epochs sample cnm structure on au spatial scales . in addition , the pulsed nature of pulsars emission allows spectra to be obtained _ on _ and _ off _ source without moving the telescope , therefore sampling both emission and absorption along almost exactly the same line of sight . @xcite summarized preliminary results from this project for three pulsars , b0823 + 26 , b1133 + 16 and b2016 + 28 . here we present results for all pulsars with full analysis . the structure of this paper is organized in the following way . section [ s : background ] provides a short summary of the main observational and theoretical results of the au - scale structure in neutral gas . in section [ s : obs ] we summarize our observing and data processing strategies . results from our investigation of changes in absorption profiles _ integrated _ optical depth properties ( c.f . equivalent widths ) are presented in section [ s : ew ] , while in section [ s : results ] we search for hi absorption _ feature _ optical depth variability . to facilitate further data analysis we derive the spin temperature and hi column density in the direction of all pulsars in section [ s : gaussian_results ] . in section 7 , we investigate in detail the properties of the ism along lines of sight toward our pulsars . in section [ s : sig_features ] we study the properties of the most significant au - scale features found in this study . we discuss our results and compare them with several theoretical models in section [ s : discussion ] , and summarize our findings in section [ s : conclusions ] . while structure on au scales has been observed in neutral , ionized , and molecular flavors of diffuse gas , it is still not clear whether all these features trace a single or multiple phenomena . our work focuses only on the tiny - scale _ atomic _ structure ( tsas ; heiles 1997 ) and we summarize results from two different observational approaches used to study tsas in the cnm below in [ s : obstsas - vlbi ] and [ s : obstsas - psr ] , as well as several theoretical considerations of the tsas phenomenon in [ s : theorytsas ] . in addition to observational methods discussed here , tsas in the cnm has been also observed through temporal ( probing scales of @xmath11 to tens of aus ) and spatial ( probing scales of a few thousands of au ) variability of optical absorption lines @xcite . the reader is referred to excellent articles in the asp conf . series vol . 365 @xcite for complementary details on optical absorption line studies , or the galactic tiny - scale structure in the ionized and molecular gas . finally , there are also some indications that even damped lyman @xmath12 absorbing systems at high redshift may contain small - scale hi structure on scales of @xmath13 au ( kanekar & chengalur 2001 ) . single baseline very long baseline interferometric ( vlbi ) observations of the quasar 3c147 by @xcite were the first to infer the existence of a tiny - scale atomic cloud , having a size of 70 au and hi volume density of @xmath14 @xmath5 , based on variations in absorption line profiles in the directions of separate source components . subsequent vlbi observations by @xcite found evidence for 25-au hi clouds in directions toward 3c138 , 3c380 , and 3c147 . particularly large variations were noticed in the case of 3c138 . the first _ images _ of hi optical depth distribution in the direction of extragalactic sources were obtained by @xcite toward 3c138 and 3c147 , using the merlin array and the european vlbi network . @xcite and @xcite used the very long baseline array ( vlba ) to image hi absorption toward seven sources . significant variations in hi optical depth with @xmath15 of 0.1 and @xmath11 , respectively , were found toward only two sources , 3c147 and 3c138 . the latest results on 3c138 by @xcite , in a three - epoch series of observations , show clear evidence for spatial variations of @xmath16 on scales of 25 au . in addition , _ temporal _ changes in hi optical depth have been found over a period of 7 years , with implied transverse velocities of order 20 ( see @xcite for a thorough review of these measurements ) . some 10% of pixels in their optical depth images have @xmath17-@xmath18 . assuming that most of the cnm along the line of sight contains small - scale structure , they estimated the volume filling factor of about 1% . they proposed that tsas is ubiquitous , and that the lower measured levels of variations in the other cases result from a selection effect due to incomplete sampling of the relevant angular and spatial scales . recently , @xcite used merlin observations of 3c161 and found @xmath191.0 - 1.5 on scales of @xmath20 au . a slightly smaller level of variations ( typically @xmath21 ) was found in the case of 3c147 , on scales of about 10 au @xcite . in the late 1980s , sufficiently accurate and repeated measurements of hi absorption profiles against pulsars began to be made , and it was noticed that pulsar ism spectra changed over time in some cases , suggesting inhomogeneities in the intervening gas . for example , @xcite found that the hi absorption spectrum of psr b1821 + 05 changed significantly between 1981 and 1988 , with the appearance at the latter epoch of a previously unobserved feature with @xmath22 and @xmath23 . @xcite showed that between @xmath24 and 1981 , hi absorption toward b1154 - 62 did _ not _ change significantly ; while toward b1557 - 50 , a variation with @xmath25 was interpreted as a cloud of size in the 1000 au range . the early pulsar hi results inspired @xcite to undertake a dedicated multi - epoch pulsar hi absorption experiment at arecibo . six pulsars were observed at three epochs , with time baselines ranging from 0.71.7 yr . these authors reported the presence of _ pervasive _ variations with @xmath26 , and associated hi column densities of @xmath27 @xmath28 @xmath29 ; on scales of 5100 au . they indicated that 10 - 15% of cold hi is in the tiny structures , and additionally detected a correlation between absorption equivalent width variations and equivalent width itself . these results appeared to buttress the earlier vlbi findings of tsas , and they provided a strong impetus for further experimental and theoretical work . the recent era of pulsar tsas experiments began with the parkes observations of @xcite . surprisingly , these investigators found no significant optical depth variations in their three - epoch , 2.5-yr observations of three pulsars , and were able to place an upper limit on column density variations toward psr b1641 - 45 of @xmath27 @xmath29 , significantly below the @xcite detections . they showed that the earlier experiment did not fully account for the large increases in noise in absorption spectra at the line frequency , so that some of the apparently significant variations actually were not . while no significant variations were seen by @xcite during the 2.5 years of their tsas experiment , they detected variations in the spectrum of psr b1557 - 50 when compared with measurements made five years earlier . this is the same pulsar whose spectrum was noted earlier by @xcite to vary on similar timescales in the late 1970s . in combining the results from four measurements over twenty - five years , @xcite concluded that the cloud causing the variations is @xmath30 au in size , with a density of @xmath31 @xmath5 . @xcite performed a particularly exhaustive tsas study on psr b0329 + 54 , which is very bright and almost circumpolar at the green bank telescope . the investigators observed the pulsar continuously for up to 20 hr in eighteen observing sessions over a period of 1.3 yr . they detected no hi optical depth variations ( @xmath32 in most cases ) for pulsar transverse offsets ranging from 0.00525 au . we and our colleagues published early results from a new set of multi - epoch arecibo observations @xcite . based on non - detections of tsas in the direction of three pulsars ( b0823 + 26 , b1133 + 16 , and b2016 + 28 ) , we suggested that tsas is not ubiquitous in the ism and could be related to isolated events . the current work elaborates these themes . several explanations were proposed to reconcile the observations of tsas with theory . @xcite proposed that tsas features could be curved filaments and/or sheets that happen to be aligned along our line - of - sight . by combining low temperature and a geometrical elongation , the apparently problematic volume density and thermal pressure can be significantly reduced . we discuss this model in some detail in section [ s : heiles ] . @xcite suggested that tsas represents the tail - end of the turbulent ism spectrum which exists on larger spatial scales , and showed that the observed optical depth differences are consistent with a single power - law description of the hi optical depth distribution as a function of spatial scale . it was also pointed out that the over - density of tsas is due to a misinterpretation of observations : the observed variations in optical depth sample the square - root of the structure function of the hi optical depth , not directly the power spectrum of optical depth fluctuations . if the power spectrum of the hi optical depth distribution in the direction of cas a , with a measured slope of 2.75 @xcite over a range of spatial scales from 0.02 to 4 pc , is extrapolated to au scales then @xmath33 - 0.4 is expected on scales of 50 - 100 au . we note that this slope was confirmed recently in an independent experiment by @xcite . @xcite proposed that optical depth fluctuations seen in multi - epoch pulsar observations are a scintillation phenomenon combined with a velocity gradient across the absorbing hi . while in the case of @xcite optical depth variations are expected to increase with the size of structure , the scintillation phenomenon predicts maximum variations on the very small spatial scales probed by interstellar scintillation . however , dedicated observations of psr j1456 - 6843 by @xcite failed to detect any scintillation effects on the hi absorption line . one of the main problems regarding tsas is its over - pressure relative to the traditional thermal pressure of the ism , which suggests that tsas can not persist nor be pervasive . however , recent numerical simulations of the dynamic and turbulent ism find substantial departures from pressure equilibrium . instead of well - defined thermal phases , gas with a wide range of densities and temperatures is found in simulations @xcite . such results are supported by observations of the ci fine - structure lines by @xcite who found evidence for the existence of over - pressured gas ( @xmath34 @xmath5 k ) . while these studies suggest that high - pressure tsas could be a natural product of the highly dynamic ism , the fraction of high - pressure gas in the ism , as well as the physical processes responsible for their formation , vary hugely across simulations . the only way to constrain tsas properties and abundance is through sensitive observations . [ t : observed_psrs ] ccccccccc psr & @xmath35 & dm & s@xmath36 & d & v@xmath37 & l@xmath38 & @xmath39 & reference + ( b1950 ) & deg & ( pc@xmath5 ) & ( mjy ) & ( kpc ) & ( ) & ( au yr@xmath0 ) & ( k ) & + b0540 + 23 & 184.36/@xmath40 & 77.7 & 9 & 3.5 & 377 & 80 & 2.7 & @xcite + b0823 + 26 & 196.96/31.74 & 19.5 & 10 & 0.4 & 192 & 39 & 6.9 & @xcite + b1133 + 16 & 241.90/69.19 & 4.8 & 32 & 0.4 & 631 & 135 & 9.0 & @xcite + b1737 + 13 & 37.08/21.68 & 48.9 & 3.9 & 4.7 & 672 & 142 & 0.7 & @xcite + b1929 + 10 & 47.38/@xmath41 & 3.2 & 36 & 0.3 & 177 & 37 & 8.3 & @xcite + b2016 + 28 & 68.10/@xmath42 & 14.2 & 30 & 1.0 & 30 & 7 & 8.4 & @xcite + we used the arecibo telescope and the caltech baseband recorder ( cbr ) backend @xcite , to measure hi absorption spectra of six pulsars at multiple epochs . we chose to observe the same sources as frail et al . ( 1994 ) in order to enhance the number of available time baselines for comparison . table [ t : observed_psrs ] lists several basic parameters for our targets : galactic coordinates ( @xmath35 ) , the dispersion measure ( dm ) , continuum equivalent flux density at 20 cm ( s@xmath43 ) , distance ( d ) , the transverse velocity ( v@xmath44 ) , the transverse distance traveled in a year ( l@xmath38 ) , and the pulsar - on antenna temperature ( @xmath39 ) . the tabulated flux density and @xmath39 represent average values obtained from long integrations , whereas the actual pulsar flux density and @xmath39 vary on timescales from several minutes to several hours due to interstellar scintillation . distances for all six pulsars ( column 5 ) were determined from interferometric parallax measurements . the pulsar - on antenna temperature @xmath45 , with the l - band gain @xmath46 k mjy@xmath0 ( heiles et al . 2001 ) , and pulsar period @xmath47 and 50% pulse width @xmath48 from the atnf pulsar catalogue ( manchester et al . 2005 ) . the observing and basic data processing procedures were described in @xcite and @xcite . relative to the previous large multi - epoch experiment by @xcite , our four - level spectrometer is less prone to systematic errors when observing these strong and highly - variable sources . we had four observing sessions , referred to below as s1 , s2 , s3 , s4 ; for epochs 2000.62 , 2000.95 , 2001.70 , and 2001.86 , respectively . the typical integration time per pulsar was about 8 hours for each of four sessions . the raw complex voltage samples recorded with the cbr were processed at caltech s center for advanced computation and research to obtain a data cube of pulsar intensity as a function of pulsar rotational phase and radio frequency . for each scan , portions of the data cube during and off the pulsar pulse were accumulated to obtain the ` pulsar - on ' and ` pulsar - off ' spectra . the pulsar absorption spectrum represents the ` pulsar - on ' ` pulsar - off ' spectrum for each scan , whose baseline was then flattened by performing frequency switching and normalizing by the off line pulsar intensity . in accumulating pulsar absorption spectra from different @xmath49min scans , each spectrum was given a weight proportional to @xmath50 , where @xmath39 is the antenna temperature of the pulsar . the frequency - switched ` pulsar - off ' spectra correspond to the emission spectra in pulsar directions . as no temperature calibration was performed during the observations , we have scaled these spectra to match the brightness temperature from the leiden - dwingeloo survey @xcite . final absorption and emission spectra for all pulsars , from the first observing epoch ( 2000.62 ) , are shown in figures [ f : all_abs ] and [ f:0540_1929em ] . our spectrometer , consisting of 4096 frequency channels across 10 mhz total bandwidth , has a velocity channel spacing of 0.52- . however , adjacent spectral channels are not independent , which can lead to distortion of narrowband signals such as the ones we are studying @xcite . to ameliorate this problem , we hanning smoothed all processed absorption spectra , yielding a final velocity resolution of 1.04 . note that hanning smoothing was not applied in our preliminary analyses [ stanimirovic et al . ( 2003 ) ; weisberg & stanimirovic ( 2007 ) ] . to flatten the baseline of the final absorption spectra we fitted a polynomial function to the unabsorbed portions of the spectrum . for pulsars b0823 + 26 , b1133 + 16 , b1737 + 13 and b2016 + 28 generally a 1st or 2nd order polynomial function was fitted and then the full spectrum was divided by the fitted function . a higher order polynomial was required for the 3rd session data for b2016 + 28 due to ripples in the baseline structure , most likely cased by interstellar scintillation . psr b0540 + 23 data were heavily affected by interstellar scintillation . this pulsar s scintillation bandwidth at 1.4 ghz is 0.3 mhz @xcite , which resulted in numerous ripples across our band ( see figure [ f:0540_1929em ] , left middle panel ) . we needed to fit and divide by a 15th order polynomial to flatten the baseline for this pulsar ( see bottom panel in figure [ f:0540_1929em ] ) . however , as emphasized in section 4.1 , we do not use this pulsar s data for our analysis due to large scintillation effects which presumably led to the unphysically ( @xmath51 ) deep absorption features . noise and calibration errors in the pulsar absorption spectrum are magnified by a factor of @xmath52 when the absorption spectrum is expressed in terms of the ultimately desired optical depth @xmath53 . with this pulsar s optical depth near infinity , these errors become so large as to render our measurements unusable . b1929 + 10 data , as shown in figure [ f:0540_1929em ] ( right , middle panel ) , suffer from the ` ghost effect ' @xcite . the main absorption feature is at 5 . however several additional features are visible from 20 to 80 which represent a false image of the pulsar on emission spectrum . the usual reason for the ghosts is an incorrect scaling of total power received by the spectrometer in the case of strong and variable sources . as the ghost is a faint copy of the pulsar on emission spectrum in the pulsar absorption spectrum , we can exorcize the ghost by fitting for and removing the former from the latter . to do so we have performed a least - squares fit of the @xmath54 absorption spectrum to a ghost of fitted amplitude @xmath55 , and a fitted linear baseline slope @xmath56 plus constant @xmath57 , all to spectral channels lacking true absorption : @xmath58 where @xmath59 is the measured pulsar - on spectrum . the exorcized absorption spectrum is then : @xmath60/c.\ ] ] the exorcized psr b1929 + 10 absorption spectrum is shown in figure [ f:0540_1929em ] , right bottom panel . the fitted parameters are listed in table [ t : ghost ] . [ t : ghost ] cccc session & @xmath55 & @xmath56 & @xmath57 + & ( k@xmath0 ) & ( @xmath61 km@xmath0 s ) & + 1 & @xmath62 & @xmath63 & @xmath64 + 2 & @xmath65 & @xmath66 & @xmath67 + 3 & @xmath68 & @xmath69 & @xmath70 + 4 & @xmath71 & @xmath72 & @xmath73 + for each absorption spectrum we have calculated an expected noise envelope which will be used to assess the significance of any observed temporal variations . the expected noise level @xmath74 is proportional to the overall system temperature @xmath75 , which has the following contributions . ( 1 ) the receiver temperature @xmath7630 k. ( 2 ) the sky background , which is typically of order of 12 k , based on the all - sky survey at 408 mhz by @xcite and after scaling to 1.4 ghz using a spectral index of @xmath77 . ( 3 ) the contribution from the hi _ emission _ which can be very significant and which varies with velocity . for example the rms noise on line is three times higher than the rms noise off line for the case of b0540 + 23 . ( 4 ) at arecibo , the pulsar continuum emission itself is also sometimes sufficiently strong to provide the principal contribution to the system temperature . since an observed pulsar - on spectrum is identically the sum of these four contributions , the expected noise in the pulsar - on spectrum should be proportional to the pulsar - on spectrum . furthermore , owing to pulsar - off integration times being @xmath13 times longer than pulsar - on integration times , the pulsar - off spectrum adds negligible additional noise to the absorption spectrum . hence our expected noise envelopes ( shown in figures [ f:0823_6plot][f:2016_6plot ] ) are generated by scaling the observed pulsar - on spectra by a multiplicative factor that matches them to the observed noise in the off - line portions of the absorption spectra . we have then propagated uncertainties associated with baseline fitting and ghost removal procedures to derive the final expected noise envelopes . to estimate significance envelopes for _ difference _ spectra shown in section [ s : results ] ( figures 3 - 8 ) or for the equivalent width analysis in section [ s : ew ] , appropriate noise envelopes for individual absorption profiles were added in quadrature . to investigate the properties of noise in emission - free channels , we have plotted the histogram of data points located off - line for each difference spectrum ( not shown in the paper ) . most histograms are consistent with gaussian statistics with a standard deviation in agreement or close to our derived 1-@xmath18 noise level . in several cases there is a small discrepancy between the measured off - line noise and derived 1-@xmath18 noise envelopes . this is not surprising after the propagation of errors due to baseline , and especially ghost , fitting . for at least three of our pulsars we have achieved excellent sensitivity . we discuss the rms on - line noise level of the final absorption spectra in section [ s : results ] . our main goal is to search for variations in the hi absorption profiles obtained at different observing epochs . we have employed two methods : ( i ) in this section we look for changes in integrated properties such as the equivalent width , and ( ii ) in section [ s : results ] we investigate spectra differenced between each pair of observing epochs . we have measured the equivalent width @xmath78 in units of , for each observing epoch . the optical depth profiles were integrated over the ` on - line ' velocity regions . to avoid low signal / noise parts of absorption spectra , as well as line wings which may be affected by small baseline imperfections , we have defined the ` on - line ' regions as roughly encompassing velocities with @xmath792-@xmath18 . for each pulsar a single region was used for all observing epochs : @xmath80 $ ] for b0540 + 23 , @xmath81 $ ] for b0823 + 26 , @xmath82 $ ] for b1133 + 16 , @xmath83 $ ] for b1737 + 13 , @xmath84 $ ] for b1929 + 10 and @xmath85 $ ] for b2016 + 28 . in addition , by averaging optical depth profiles for all epochs we have estimated the mean equivalent width for each pulsar ( @xmath86 ) . to compare ew from one epoch to the next we have calculated @xmath87 $ ] . these values , as a function of time baselines @xmath88 ( bottom axis ) or traversed spatial scales @xmath89 ( shown on the top axis ) , are shown for all pulsars in figure [ f : comparison_ew ] . b0540 + 23 was excluded from this analysis as its optical depth profile is uncertain around 0 . to convert from time intervals @xmath88 to spatial scales @xmath89 , we use @xmath90 , where @xmath91 is provided in table 1 . our spatial scales given below should be multiplied by an ( unknown ) fractional distance factor @xmath92 to account for the fact that absorbing hi is located between us and the pulsar . note that some authors have arbitrarily fixed @xmath93 ; e.g. for @xcite @xmath94 . ( when we assemble various workers results in this paper , we redefine all such values onto our scale . ) as shown in figure [ f : comparison_ew ] , in the case of b0823 + 26 , b1133 + 16 , b1737 + 13 and b2016 + 28 all variations in @xmath95 are essentially within 2-@xmath18 and are most likely not significant . however , in the case of b1929 + 10 , baselines of 0.33 and 1.24 yrs show changes in @xmath95 at a @xmath96-@xmath18 and @xmath97-@xmath18 level , respectively . in addition , the 0.16 yr baseline shows a variation at a @xmath98-@xmath18 level , while the 0.75 yr baseline shows a 2.5-@xmath18 change . to quantify the significance of @xmath95 variability we test the null hypothesis of @xmath95 being constant with time . we fit a constant @xmath95 model to measurements in figure [ f : comparison_ew ] and find @xmath990.4 ( b0823 + 26 ) , 0.8 ( b1133 + 16 ) , 1.3 ( b1737 + 13 ) , 6.0 ( b1929 + 10 ) , and 1.2 ( b2016 + 28 ) . we conclude that @xmath95 changes for b1929 + 10 can not be fit with a constant model , while for all other pulsars no significant variability of @xmath95 is observed . this @xmath95 variability , as further analyzed in section 7.1 , can be interpreted as being due to tsas . in figure [ f : like_fig4 ] we plot @xmath95 , separately for individual pulsars , but as a function of @xmath100 . this plot can be directly compared with figure 4 from frail et al . our results for b0823 + 26 , b1133 + 16 and b1929 + 10 agree roughly with frail et al.s , @xmath101 . however @xmath95 for b1737 + 13 and b2016 + 28 found in frail et al . ( 1994 ) were significantly larger than what we find . a particularly striking difference is the case of b2016 + 10 , where frail et al . ( 1994 ) found @xmath1025 , while we find only @xmath1030.3 . we believe that at least some of the variations frail et al . saw were due to small calibration errors . looking at our whole sample , the fractional variation @xmath104 spans a range from 0% to 11% , with a median of 4% . this is significantly smaller from a median value of 13% found in frail et al . ( 1994 ) . as a conclusion , we find significant variations in @xmath95 only for b1929 + 10 , with two largest changes being at a fractional level of 11% . the frequency of variations and the level of variations ( as well as upper limits ) are significantly smaller than what was found previously for the same sources by frail et al . our results do not rule out the finding by @xcite that @xmath105 increases with ew . however , if this relationship exists , it is at a much smaller level than suggested by frail et al . considering that tsas may have temperature as low as 15 k ( heiles 1997 ) , the change in absorption profiles may occur over @xmath106 and appear in only one or two velocity channels in our spectra . such narrow changes may get washed out when calculating the equivalent width . we therefore in this section compare all pairs of absorption profiles for a given pulsar , and investigate absorption spectrum differences , @xmath107 . with four observing epochs we have six difference spectra ( figures [ f:0823_6plot ] to [ f:2016_6plot ] ) . overlaid atop each difference spectrum is a @xmath1082-@xmath18 significance envelope calculated from noise envelopes discussed in section [ s : noise ] . in summary , we see changes in difference spectra at a 3-@xmath18 level only in the case of b1929 + 10 and only for the two baselines where we also find changes in @xmath95 at a @xmath109-@xmath18 level . we note that b1133 + 16 has one feature above 3-@xmath18 ( s4-s3 baseline ) at a velocity of @xmath110 , however this likely comes from rfi as is seen in _ emission _ in the absorption spectrum from session 3 . similarly , b1737 + 13 has a significant variation at a velocity of @xmath111 on 3 baselines , all involving the session 1 absorption spectrum ; this corresponds to an ( unphysical ) _ emission _ feature and is most likely due to rfi . for three of our pulsars ( b0823 + 23 , b1133 + 16 and b1929 + 10 ) we have excellent sensitivity for tsas experiments , @xmath112 - 0.02 ( 2-@xmath18 in the on - line portion of the spectrum ) . only one recent experiment ( minter at al . 2005 ) reached similar sensitivity but primarily over smaller spatial scales . for b1737 + 13 and b2016 + 28 we have reached 2-@xmath18 sensitivity of 0.1 - 0.2 . although this is slightly higher , several pulsar and vlba tsas detections have been claimed at this level . we provide below brief comparison between our observations for individual pulsars and previous tsas experiments . the corresponding tranverse distances traveled by b0540 + 23 during the course of our observations are 13 , 26 , 60 , 73 , 86 and 100 aus . frail et al . ( 1994 ) concluded that there were clear signs of significant optical depth variations in some absorption components for this pulsar . in fact both our and frail et al.s spectra show ( @xmath113 ) slightly less than zero near 0 as well as scintillation ripples . as already discussed in section 3.2 , we do not use data for this pulsar for our further analyses and do not show difference spectra . with @xmath114 au yr@xmath0 , the time intervals covered by our experiment translate to spatial scales of 6 , 13 , 29 , 35 , 42 and 48 aus . there are no data points with @xmath15 above 3-@xmath18 in difference spectra for all six time baselines . we conclude that _ no _ significant change in absorption profiles is found down to a @xmath115 level of 0.03 ( a 2-@xmath18 level on line ) . frail et al . ( 1994 ) detected only marginal changes over 0.6 yr , but reported significant variations of @xmath116 in 1.1 yr . we do not confirm this reported variation on any of our temporal baselines although our 2-@xmath18 sensitivity level on all baselines is at least two times smaller than the level of variations found by frail et al . we discussed this further in @xcite and concluded that calibration and noise issues affected the frail et al . analyses . as this pulsar has @xmath117 au yr@xmath0 a wide range of spatial scales is probed by our observations , from 20 au to 170 au . no variability at @xmath118-@xmath18 was found and we conclude that _ no _ significant variations were detected in the direction of this pulsar down to a @xmath115 level of about 0.02 . frail et al . ( 1994 ) reported @xmath119 on a 1.1 yr baseline . on all of our difference spectra ( time baselines of 0.33 , 0.75 and 1.08 yr ) our 2-@xmath18 sensitivity is half the level of variations found in frail at al . ( 1994 ) , @xmath120 at most . we note that no variability was detected at this level in any of later variability experiments . with @xmath121 au yr@xmath0 the range of probed spatial scales is from 20 au to 175 au . _ no _ significant variations are found . frail et al . ( 1994 ) reported @xmath122 at 4 ( which corresponds to @xmath123 in this relatively deep line ) over 0.65 yr . our 2-@xmath18 sensitivity is @xmath124 for half of baselines , and 0.3 for the rest . several tsas features were detected at a similar level in recent experiments , both pulsar and interferometric . with @xmath125 au yr@xmath0 , the range of probed spatial scales is very narrow , from 6 to 45 au . the absorption profile for this pulsar consists of at least three different velocity components ( at @xmath126 , 5 and 8 ) . the component at 5 , which is the one with the deepest hi absorption , is the only one exhibiting significant changes . difference plots show line variations at 3.2 and 3.0-@xmath18 levels ( @xmath1270.025 ) for time baselines of 0.33 and 1.24 years . we consider these detections to be significant , especially as the same time baselines showed significant variations in ew . this level of variability is the lowest ever detected in tsas experiments . both detected difference features are sharp and appear in the same single velocity channel at 5 over different epochs . we further measure and analyze properties of the intervening absorbing clouds traced by these detected features in section [ s : sig_features ] . baselines of 0.16 , 0.75 , 0.91 , and 1.08 years have an upper - limit in @xmath15 of 0.012 - 0.018 at this velocity . frail et al . noticed variations of up to @xmath128 for this pulsar , with the largest fluctuations being at a velocity of @xmath129 , where we observe no significant variations . as this is the only pulsar in our sample to exhibit significant absorption variations , we have scrutinized the data acquisition and analysis procedure with extra care . for example , a scaling error of approximately @xmath130 in pulsar strength in a single epoch could result in an erroneous rescaling of optical depths by a similar amount , which would lead to a spurious detection of variation when there is actually none . for example , in stanimirovic et al ( 2003 ) , we pointed out an apparent case of this type of problem in frail et al s spectra of psr b2016 + 28 . however , it appears that the more advanced cbr spectrometer is indeed able to measure spectra more accurately , as we anticipated . for example , we see no evidence of variations in our experiment with psr b2016 + 28 nor with b1133 + 16 , the other particularly bright pulsars in our sample . ( calibration errors would be expected to be most serious in the brighter pulsars , which can easily increase the system temperature by factors of more than two . ) it is notable that b1929 + 10 is the only pulsar in our sample from which we removed a ghost . then an obvious question is whether our detected variability could be an artefact of the ghost fitting procedure . we have scrutinized our reduction procedures and remain confident that the variability is real . spectra for b1929 + 10 were reduced in the same way as for all other sources . even the ghost fitting procedure was done in exactly the same way ( using exactly the same idl routines ) as the polynomial baseline fitting for other pulsars . the derived error envelopes reflect additional uncertainties due to this fitting process . in figure [ f:1929_4plot ] we show all four absorption profiles for this pulsar before ( top ) and after ( bottom ) ghost removal . only the deepest velocity component at 5 shows variability ( both before and after ghost removal ) , while components at @xmath126 and 8 agree extremely well between epochs . if there were some imperfections related either to a scaling problem or to ghost fitting errors , they would affect _ all _ features . we have also examined the estimated coefficients used in equation ( 2 ) . in the case of the 1st session spectrum , for example , @xmath131 . a smaller value for @xmath55 would result in a larger correction and would bring the 1st session absorption spectrum closer to spectra from other sessions . however , even by taking values within @xmath132 of the best - fit estimate , the observed variations in the difference spectra would still stay significant ( above 3-@xmath18 ) . the reason for this is that the hi emission at 5 is not exceptionally high , resulting in a small ghost correction . after all these tests , we remain confident that the detected variability in the case of b1929 + 10 is real . in figure [ f:2016_6plot ] we compare all four absorption profiles for this pulsar . this is the slowest - moving pulsar in our sample , so the range of spatial scales covered with our observations is only 110 aus . no significant variations were found . our 2-@xmath18 sensitivity is @xmath133 for all baselines . frail et al . ( 1994 ) found @xmath134 ( from their fig . 3 ) over periods of 0.6 and 1.7 yr . as first noted in stanimirovic et al . ( 2003 ) , we compared one of our absorption profiles for this pulsar with data from frail et al . and concluded that slight calibration problems may have affected frail et al.s data . two recent pulsar experiments found variability at this level but at much larger scales ( @xmath135 au ; johnston et al . 2003 and weisberg et al . 2008 ) . @xcite presented preliminary results of this experiment , and noted that the unsmoothed hi absorption spectra exhibited several apparently significant optical depth variations , mostly in the form of single - channel fluctuations . the most prominent ones were in the direction of b1929 + 10 , but there were a few additional variable features toward other pulsars . for the current study , however , where we have applied hanning smoothing to ensure full independence between velocity channels , only the detections in the direction of b1929 + 10 survive . could the additional variations in the unsmoothed spectra be real ? the minimum expected temperature for the cnm in thermal equilibrium is 15k @xcite . to achieve such a low temperature , dust grain heating of the hi would need to be absent . at this temperature , the full - width , half - maximum ( fwhm ) of an hi feature will be 0.8 km s@xmath0 even in the absence of turbulent broadening ; for a less extreme 30k the corresponding width is 1.2 km s@xmath0 . using a large sample of cnm clouds , @xcite deduced a relationship between @xmath136 , the temperature inferred from the measured linewidth , and the spin temperature @xmath137 , where @xmath138 is the sonic mach number . for 196 out of 202 cnm components they found that @xmath139 , implying then that @xmath140 . therefore , even with the minimum expected turbulence , the hi absorption line fwhm would be @xmath7 0.9 km s@xmath0 at @xmath141 k , or 1.3 km s@xmath0 at @xmath142 k. so while we have rejected most of @xcite s variations because unsmoothed data can be misleading , there are also physical reasons to doubt reality of features with fwhm@xmath143 . in contrast , our hanning smoothed resolution appears to match the narrowest plausible tsas features . in this section we derive the spin temperature and the cnm column density for gas seen in absorption in the direction of our pulsars . we employ the technique developed by @xcite , which is based on fitting gaussian components to both hi absorption and emission spectra . the purpose of this section is twofold : ( i ) our further data analysis requires temperatures and column densities of associated cnm clouds ; and ( ii ) since the cnm absorption profiles are often superpositions of several components , our original hope was to search for temporal changes in both the peak optical depth and velocity centroids of all individual velocity components . unfortunately , since the gaussian decomposition of absorption and emission profiles is not unique ( see discussion in @xcite ) , our second goal was not achievable . to analyze each pulsar , we first selected the epoch with the lowest noise level . the hi absorption and emission spectra were then decomposed into several gaussian profiles . we note that three of our six pulsars are at high galactic latitudes and have simple line profiles , easily fitted with one or two gaussian functions . optical depth profiles for low latitude pulsars b0540 + 23 , b1929 + 10 , and b2016 + 28 , ( at @xmath144 , @xmath41 , and @xmath42 , respectively ) , are more complex but still with several obvious velocity components . we summarize here very briefly the basic definitions and the procedure , see heiles & troland ( 2003a ) for further details . for each pulsar we have the hi emission profile , @xmath146 ( note that this is the emission spectrum that would be observed in the absence of the pulsar ) and the hi optical depth spectrum , @xmath147 . both spectra in general have multiple velocity components . we first fit @xmath148 with a set of @xmath149 gaussian functions using a least - squares technique : @xmath150 ^ 2 } \label{eqn : tau}\ ] ] where @xmath151 is the peak optical depth , @xmath152 is the central velocity , and @xmath153 is the 1/e width of component @xmath154 . @xmath149 is the minimum number of components necessary to make the residuals of the fit smaller or comparable to the estimated noise level of @xmath54 . while the optical depth spectrum predominantly reflects the cnm , both the cold _ and _ warm neutral media ( wnm ) contribute to the hi emission spectrum . consequently we can write the amplitude of the emission spectrum , @xmath146 , as follows : @xmath155 we now write each of the two terms of eq . [ eqn : tb ] in detail , using the optical depth profile derived from eq . [ eqn : tau ] . first , @xmath156 , the hi emission originating from @xmath149 cnm components , can be written as : @xmath157 where @xmath158 is the spin temperature of cloud @xmath154 , and the subscript @xmath159 represents one of the @xmath138 cnm clouds that lie in front of cloud @xmath154 . next , @xmath160 , the hi emission originating from the wnm , is represented with a set of @xmath161 gaussian functions . the complicating factor here is that a certain fraction @xmath162 of the wnm is located in front of the cnm , while a fraction @xmath163 of the wnm is beyond the cnm with its emissions being absorbed by cnm clouds : @xmath164 \times t_{0,k}e^{-[(v - v_{0,k})/\delta v_{k}]^2 } , \ ] ] where the subscript @xmath165 corresponds to each of the wnm components and a fraction @xmath166 of the wnm cloud @xmath165 lies in front of all cnm components , while a fraction @xmath167 is being absorbed by the cnm clouds . for a given order of cnm clouds along the line of sight and a given set of @xmath166 values , the @xmath146 profile is simultaneously fitted for gaussian parameters of the wnm components and the spin temperature of individual cnm clouds . for each pulsar , we vary the order of gaussian functions along the line - of - sight ( for @xmath149 cnm components there are @xmath168 possible orderings ) and perform the @xmath146 fit . we then choose the ordering of cnm components that gives the smallest residuals in the least - squares fit . unfortunately , the difference in the fit residuals is often not sufficiently statistically significant to distinguish between different values of @xmath166 . however , @xmath166 has a large effect on the derived spin temperatures . hence we follow the heiles & troland ( 2003a ) suggestion and estimate the final spin temperatures by assigning characteristic values of 0 , 0.5 , or 1 to each @xmath166 ( among the extreme possible values of 0 and 1 ) , and repeating this for all possible combinations of wnm clouds . the final spin temperatures are then derived as a weighted average over all trials . figure [ f:1737_wnm_cnm ] shows results from the gaussian decomposition and fitting of hi emission and absorption spectra for two example pulsars , b1737 + 13 and b1929 + 10 . this figure has three panels : the _ top panel _ , displaying the emission spectrum , @xmath169 , and its decomposition into separate cnm [ @xmath156 , shown as a dotted line ] and wnm [ @xmath160 , shown as a dashed line ] components ; the _ middle panel _ giving the optical depth profile and its individual gaussian - fitted components ( shown as dotted lines ) ; and the _ bottom panel _ showing the difference between the observed optical depth profile and its gaussian component fit ( dashed lines on this plot show @xmath170 noise levels ) . in table [ t : table_wnm_cnm ] we list the fitted cnm properties . with the exception of the b1929 + 10 line of sight , all derived spin temperatures listed in table [ t : table_wnm_cnm ] , are in agreement with what is typically found for cnm clouds @xcite . for example , @xcite show histograms of wnm and cnm temperatures from their hi emission / absorption survey of 79 continuum sources . most cnm clouds have a spin temperature in the range 2070 k , while cnm clouds with a temperature in the range 150200 k appear to be rare . however , spin temperatures we derived for cnm clouds in the direction of b1929 + 10 are high , @xmath171 and @xmath172 k. this is the first indication that this particular line of sight could be sampling atypical ism conditions . in table [ t : hi_column_density_for_psrs ] we provide the hi column density in wnm and cnm phases in the direction of each pulsar . the first three pulsars have the cnm / total column density of about 15 - 20% , but for b1929 + 10 this ratio is only 7% , and for b2016 + 28 the ratio is about 40% . the very low cnm abundance in the direction of b1929 + 10 is unusual . for comparison , @xcite found that all except one of their sources with detected cnm have a cnm fraction @xmath173% , with a majority of sightlines having @xmath174% . however , the same study also noted a separate class of objects not exhibiting evidence for the cnm along their lines of sight . they suggested that at least some of these regions may be affected by supershells . the cnm fraction in the direction of b1929 + 10 is similar to several directions with a cnm fraction of only 24% observed by @xcite and @xcite . clearly , properties of the cnm in the direction of b1929 + 10 stand out as unusual . as we will see in [ s:1929_section ] , a significant fraction of this line of sight is located inside the local bubble . [ t : table_wnm_cnm ] lccccc psr & @xmath175 & @xmath176 ( lsr ) & @xmath177 & @xmath178 & n(hi)@xmath179 + & & ( ) & ( ) & ( k ) & ( @xmath180 @xmath29 ) + b0823 + 26 & & & & + & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + + b1133 + 16 & & & & + & @xmath186 & @xmath187 & @xmath188 & @xmath189 & @xmath190 + & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 + + b1737 + 13 & & & & + & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 + & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 + & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 + + b1929 + 10 & & & & + & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + & @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 + & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + + b2016 + 28 & & & & + & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 + & @xmath231 & @xmath232 & @xmath233 & @xmath229 & @xmath234 + & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 + & @xmath240 & @xmath241 & @xmath242 & @xmath229 & @xmath243 + & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 + [ t : hi_column_density_for_psrs ] ccccc psr & @xmath249 & @xmath250 & @xmath251 & @xmath252 + & ( 10@xmath253 @xmath29 ) & ( 10@xmath253 @xmath29 ) & ( 10@xmath253 @xmath29 ) & + b0823 + 26 & 4.3 & 0.7 & 5.0 & 0.14 + b1133 + 16 & 3.5 & 0.7 & 4.2 & 0.16 + b1737 + 13 & 6.3 & 1.7 & 8.0 & 0.21 + b1929 + 10 & 34.7 & 2.5 & 37.2&0.07 + b2016 + 28 & 28.9 & 18.7 & 47.7&0.39 + we have examined the lines of sight of all pulsars in our sample . while b1737 + 13 and b2016 + 28 are at a significant distance from the sun and their lines of sight probably average over many interstellar clouds , b0823 + 26 , b1133 + 16 and b1929 + 10 are close with significant portions of their lines of sight probing the very local ism . we detail here some interesting properties of the latter three pulsars lines of sight . psr b1929 + 10 is the closest pulsar in our sample , @xmath254 pc @xcite , and this relatively close distance raises the question of whether the observed fluctuations of hi optical depth are somehow related to the pulsar s proximity to the solar system . for example , we are embedded near the middle of a low - density , high - temperature region called the local cavity which extends 1 - 200 pc before being bounded by a high density wall in most directions , so a significant portion of the line of sight lies inside it . in what follows , the `` local bubble ( lb ) '' consists of the local cavity plus the wall . @xcite mapped the dense neutral gas wall bounding the local cavity via interstellar absorption measurements of stars of known distance . the updated map of the neutral gas surrounding the local cavity by @xcite confirms that the line of sight from b1929 + 10 enters the wall of the lb at a distance of @xmath255 pc and then pierces the wall for @xmath256 pc . a lower - resolution interstellar reddening map from @xcite also indicates that the lb boundary is distended significantly toward the pulsar . furthermore , the b1920 + 10 line of sight appears to be in or near a long extended neutral finger of the lb ( possibly an interstellar tunnel gouged by a supernova ) and may therefore lie in the finger or along its wall for some distance . the dense finger is most likely associated with the ophiuchus molecular cloud and @xcite found that the co distribution in this region is filamentary and has radial velocities in the range 4.2 - 5.3 ( note that our varying hi component is centered at @xmath217 ) . in addition to the above , we found that two stars within 3 degrees from b1929 + 10 have been observed in nai : hd178125 is at a distance of 173 pc , while hd180555 is at a distance of only 106 pc . based on the data presented in @xcite , these stars show nai absorption at an lsr velocity of @xmath257 and @xmath258 , respectively ( velocity uncertainty for these measurements is 1 - 1.5 ) . as our varying pulsar absorption component is centered at @xmath259 ( with a fwhm of 3 ) , it is most likely tracing the same interstellar cloud seen in the direction of the two hd stars . this would imply that the absorbing cloud is at a distance of @xmath260 pc , as it is seen against hd180555 , and therefore embedded in the local cavity or the lb wall . the local cavity itself is a region of low scattering ; while most investigations have concluded that there is enhanced scattering at its boundary , implying enhanced turbulence there @xcite . based on modeling of many nai absorption features , @xcite found that the @xmath257 and @xmath258 absorption components in the direction of hd178125 and hd180555 belong to a plane - parallel flow immersed in the local cavity , the so - called `` flow c '' . they suggested that flow c could represent gas condensing out of the lb wall . generally , many diffuse interstellar clouds ( so - called `` local fluff '' ) have been found inside the local cavity itself and have been interpreted as signatures of the development of rayleigh - taylor instabilities in the wall , whereby gas is condensing out of the leading edge of the lb wall and flowing toward the sun @xcite . to examine lines of sight toward b0823 + 26 and b1133 + 16 , we turn to figure 7 of @xcite which shows neutral gas in vertical planes at fixed galactic longitudes . in the case of b0823 + 26 ( @xmath261 ) , about one - half of the line of sight appears to be inside the local cavity and traverses the lb wall over only a short distance . in the case of b1133 + 16 ( @xmath262 ) , the line of sight appears to exit the lb through an opening or a chimney of low density material rather than a bounding wall or any other substantially dense neutral material . in summary , since almost one - half of the line of sight toward b1929 + 10 is associated with the lb and the absorbing gas is likely to be within 106 pc , we hypothesize that our detected tsas clouds along this line of sight are formed or influenced by the lb . based on the stellar nai absorption lines at similar velocities to those we find in hi toward b1929 + 10 , we conclude that our tsas fluctuations are likely to be associated with the lb wall . the high spin temperature of the absorbing clouds , as shown in the previous section , may result from the thermally unstable gas due to wall fragmentation ( see also section 9.4 ) . it is then not surprising that we did not detect tsas in the direction of b0823 + 26 and b1133 + 16 , despite equally good sensitivity , because these lines of sight do not traverse much or any of the wall . in sections [ s : ew ] and [ s : results ] we have identified several significant variations in absorption profiles that could be interpreted as tsas . all of these features are in the direction of b1929 + 10 . we summarize their observed properties in table [ t : tsas_sig_features ] . we measure tsas properties from ew variations , as this is more robust than using difference spectra which have a lower s / n . we estimate tsas properties from ew variations with @xmath263-@xmath18 . table [ t : tsas_sig_features ] lists the following quantities : the time baseline over which the change in ew is observed , the corresponding transverse spatial scale , the absolute value of @xmath95 , the significance of the fluctuation , the hi column density of the tsas ( @xmath264 ) , the inferred hi volume density of tsas ( assuming spherical hi blobs whose radius is equal @xmath89 ) , and the inferred thermal pressure . some comments are necessary for interpreting these tables . ( i ) our observations do not constrain the exact _ size _ of tsas , but rather its characteristic _ scale_. in measuring the tranverse distance ( @xmath89 ) traveled by the pulsar over a time span exhibiting a variation in hi absorption spectra , we are sampling a spatial scale over which we have sampled the optical depth function . therefore , even in the case that significant fluctuations are due to discrete blobs of gas , @xmath89 does not necessarily correspond to a physical size of these blobs . this point has been emphasized by deshpande ( 2000 ) . \(ii ) the temperature of tsas , needed to estimate @xmath264 , is also not constrained directly by observations . we have estimated in section 6 the temperature of the parent cnm clouds with which tsas is associated . in the scenario where tsas results from discrete , elongated objects , such as in @xcite and @xcite , cnm clouds are constituted from tsas and an inter - tsas medium . the derived spin temperature in this case corresponds to the inter - tsas medium , while tsas could have a much colder temperature of @xmath265 k ( heiles 1997 ) . there are several interesting results : 1 . significant fluctuations ( table [ t : tsas_sig_features ] ) are found on a variety of spatial scales @xmath266 au ( with @xmath267 being a tsas distance ) , suggesting that the cnm in the direction of b1929 + 10 has structure on spatial scales of 550 aus . our characteristic scales for tsas are similar to results from vlba experiments : brogan et al . ( 2005 ) found a typical size of 25 au and lazio et al ( 2009 ) found a typical size of 10 au . if we assume that tsas has the same spin temperature as its parent cnm cloud ( @xmath268 k in the case of b1929 + 10 ) , we estimate a hi column density of @xmath269 @xmath29 . this is slightly lower than the typical cnm column density of @xmath270 @xmath29 found for the cnm in the heiles & troland ( 2003b ) survey . all tsas features contribute 10 - 30 percent to the hi column density of their ` parent ' cnm clouds . 4 . if we interpret tsas as being in the form of spherically symmetric blobs , its observationally inferred hi volume density and thermal pressure are @xmath271 @xmath5 and @xmath272 @xmath5 k , respectively . these values are more than two orders of magnitude larger than what is typically expected for the ism , in agreement with previous tsas measurements ( see section 2 ) . 5 . in terms of hi mass , tsas contains only @xmath273 earth masses . this is many orders of magnitude below what is required for gravitationally confined clouds . 6 . in the shaped tsas model ( heiles et al . 1997 ) tsas could be much colder with a temperature up to a factor of 10 lower than what we have used . this would reduce the derived @xmath264 and @xmath274 by a factor of 10 , and @xmath275 by a factor of 100 . if an elongation factor @xmath276 is further introduced , the derived thermal pressure can be brought down to @xmath277 @xmath5 k which is close to the traditional hydrostatic pressure ( with some turbulent contribution ) . however , this could work only for the case of _ both _ low tsas temperature and significant line - of - sight elongation . in table [ t : tsas_upper ] we provide upper limits on tsas properties based on non - detections in the directions of four other pulsars . in the heiles ( 1997 ) scenario , a typical cnm cloud consists of an inter - tsas medium and many discrete tsas features in the form of cylinders or disks , which are homogeneously and isotropically distributed within the cloud . an arbitrary line of sight samples many cylinders or disks ; however only cylinders seen end - on or disks seen edge - on have sufficient column densities to produce absorption features recognized as tsas . meanwhile , non - end - on cylinders or non - edge - on disks have very little column density and blend to form a smooth background which contributes very little to the measured hi absorption profile . the model assumed the following basic parameters for tsas : a cloud size perpendicular to the line of sight @xmath278 au , @xmath279 k , @xmath280 @xmath29 , and @xmath281 for the number of end - on cylinders or edge - on disks along every line of sight . two values were considered for @xmath282 , the ratio of the cnm to tsas column density : @xmath283 and 100 . the model predicts the elongation factor @xmath284 and the volume filling factor @xmath285 of tsas required to ameliorate the over - pressure problem , under the assumption that the asymmetric tsas features are uniformly and isotropically distributed throughout cnm clouds . if tsas is in the form of edge - on disks then @xmath286 and they fill about 3.3% of the cnm volume ; whereas if tsas is in the form of end - on cylinders then @xmath287 and they fill about 3.6% of the cnm volume . however , the volume filling factor of tsas is directly proportional to the number of end - on cylinders or edge - on disks @xmath288 along any line of sight . if we assume @xmath289 k for our tsas detections , a geometric factor @xmath290 is required to explain the observed tsas volume density and pressure within the context of the heiles model . this is in the range proposed by heiles ( 1997 ) . in addition , the tsas hi column density as well as the @xmath282 ratio are similar to values explored in the model . it is therefore reasonable to compare our estimated volume filling factor of tsas with the model predicted values . in the case of b1929 + 10 , as the pulsar moves through the ism our hi absorption spectra have essentially sampled a one - dimensional cut across the absorbing cnm cloud ( about 50 au long ) and detected at least two tsas features . this means that a one - dimensional covering fraction of tsas in this direction is at least 40% , if we assume a 10au plane - of - sky tsas size . we can now estimate the fraction of the line of sight occupied by tsas . we first estimate the line - of - sight size of the parent cnm cloud . as this cloud has @xmath291 @xmath29 ( from table [ t : table_wnm_cnm ] ) , and if we assume a cnm volume density of 100 @xmath5 , this results in a length of about @xmath292 au . assuming a cylindrical tsas with a 10au plane - of - sky size and a geometrical elongation @xmath293 , results in the volume filling factor of @xmath294% in the direction of b1929 + 10 . the volume filling factor in the case of other pulsars in our experiment is zero . we note that our sensitivity for b0823 + 26 and b1133 + 16 is equally good as for b1929 + 10 , and all three pulsars probe interstellar environments with a similar optical depth ( peak of @xmath295 , see figures 1 and 2 ) and a similar line - of - sight length . in a similarly sensitive experiment , @xcite had 18 observing epochs for one pulsar , about 150 trials , and _ no _ detections . the observationally inferred volume filling factor of tsas in the direction of our pulsars ( 0 or @xmath294% ) is lower than the model prediction ( @xmath296% ) . more importantly , we observe significant inhomogeneity in the galactic distribution of tsas : no detections toward four pulsars , and a high detection rate toward b1929 + 10 . for the heiles ( 1997 ) model to remain viable , our observations require inclusion of the possibility of different tsas properties ( specifically @xmath91 , @xmath284 and @xmath288 ) along different lines of sight , and/or additional local physical processes which can modify the tsas production rate . for example , a highly elongated tsas with @xmath91 a million times smaller than the size of its parent cnm cloud would theoretically result in a volume filling factor close to zero . while in principle this scenario could work , we are still left with the puzzle of what physical processes can produce such tiny clouds and at the same time greatly vary cloud properties across the ism . [ t : tsas_sig_features ] ccccccc @xmath88 & @xmath89 & @xmath297 & signif . & @xmath298 & @xmath299 & @xmath275 + ( yr ) & ( au ) & ( ) & ( @xmath18 ) & ( 10@xmath300 @xmath29 ) & ( 10@xmath301 @xmath5 ) & ( 10@xmath302 @xmath5 k ) + 0.16 & 6 & 0.046 & 2.8 & 1.4 & 16 & 27 + 0.33 & 12 & 0.083 & 3.3 & 2.6 & 14 & 24 + 0.75 & 28 & 0.040 & 2.5 & 1.2 & 3 & 5 + 1.24 & 46 & 0.089 & 3.5 & 2.8 & 4 & 7 + @xmath303 we have assumed @xmath304 k based on table [ t : table_wnm_cnm ] . + @xmath305 calculated assuming a spherical geometry . + [ t : tsas_upper ] cccccc psr & @xmath89 & @xmath297 & @xmath298 & @xmath299 & @xmath275 + & ( au ) & ( ) & ( 10@xmath306 @xmath29 ) & ( 10@xmath301 @xmath5 ) & ( 10@xmath302 @xmath5 k ) + b0823 + 26 & 5 - 50 & 0.05 & @xmath307 & @xmath308 & @xmath309 + b1133 + 16 & 20 - 170 & 0.06 & @xmath309 & @xmath310 & @xmath143 + b1737 + 13 & 20 - 180 & 0.35 & @xmath311 & @xmath308 & @xmath309 + b2016 + 28 & 1 - 10 & 0.45 & @xmath312 & @xmath313 & @xmath314 + @xmath303 we have assumed @xmath315 k. + @xmath305 calculated assuming a spherical geometry and the largest probed size . + several recent numerical simulations of the small - scale structure in the ism have addressed the tsas phenomenon . in particular , @xcite performed relatively high resolution numerical simulations of colliding wnm flows and the formation of cnm clouds . a collision of incoming wnm streams creates a thermally unstable region of higher density and pressure but lower temperature , which further fragments into cold structures . the thermally unstable gas has filamentary morphology and its fragmentation into cold clouds is promoted and controlled by turbulence . while typical cnm clouds are formed in these simulations within @xmath11 myr , transient shocked cnm regions , produced by supersonic collisions between cnm fragments , are also found , with properties similar to those of tsas : @xmath316 @xmath5 , @xmath317 k , @xmath318 k @xmath5 , and a slightly larger size of @xmath319 au . @xcite found that 10% of lines of sight cross dense cnm gas with @xmath320 @xmath5 , and only 0.1% of lines of sight cross gas denser than @xmath321 @xmath5 , suggesting that extremely dense tsas should rarely be detected . while this was a 2d simulation of a coherent region only @xmath322 pc large , basic observed properties of the cnm / wnm were surprisingly well captured . for example , the cnm column density peak , and the hi column density distribution function appear close to observed properties based on the heiles & troland ( 2003 ) survey . considering that the simulation has a resolution of 400 au ( cruder than the scales generally sampled in pulsar and interferometer experiments ) and probes regions only 20 pc across ( much smaller than our lines of sight ) , direct comparison with our observations is not straightforward . nevertheless , the predicted properties of the shocked cnm clouds , especially cloud abundance , are encouraging . considering that the cnm / tsas production in converging flows is controlled by the level of turbulence in colliding flows , local turbulent enhancements may be able to provide the necessary tuning and therefore explain the inhomogeneity in the spatial distribution of tsas observed in the direction of b1929 + 10 . in an attempt to compare various observations , we have investigated hi tsas detections and upper limits from all published and recent pulsar and interferometric experiments as a function of their spatial scale . the results are summarized in figures [ f : us_desh1 ] and [ f : us_desh2 ] . in figure [ f : us_desh1 ] we first show the maximum variation in hi optical depth ( @xmath15 ) as a function of peak optical depth along the line of sight ( @xmath323 ) . this plot separates all sources cleanly into two groups : `` low optical depth '' with @xmath324 ( b1929 + 10 , b0823 + 16 , b1133 + 16 , and b0301 + 19 ) , and `` high optical depth '' with @xmath325 ( 3 interferometric detections , pulsars b1737 + 13 b2016 + 28 from this study , highly sampled pulsar b0329 + 54 from minter et al . 2005 , and b0736 - 40 , b1451 - 68 , b1557 - 50 from johnston et al . 2003 ) . the only exception in this figure is the 2.6-@xmath18 detection by weisberg et al . ( 2008 ) in the case of b0301 + 19 which probes a low optical depth cnm but has @xmath326 . this plot suggests that regions with high optical depth ( and therefore high cnm column density ) have higher level of @xmath15 variations than regions with low cnm column density . in other words , this implies that the the level of @xmath53 fluctuations tracing tsas is intimately associated with the larger scale structure of the parental cnm . figure [ f : us_desh2 ] shows the detected level of optical depth variations , or upper limits , as a function of tsas spatial scale @xmath89 . no obvious correlation between the two quantities is noticeable . the kendall s tau correlation test for censored data @xcite gives a probability of 62% that the correlation is not present . the low and high @xmath323 groups of sources nicely separate in this plot around @xmath327 : the high-@xmath323 sources are above this line , while all low-@xmath323 sources except b0301 are below this line . we investigated whether a correlation between tsas scale @xmath89 and @xmath15 exists for the two groups separately . for both groups the kendall s tau correlation test for censored data gives a probability of @xmath328% that the correlation is not present . we therefore conclude that the detected level of optical depth variations , as well as upper limits , do not show evidence for a correlation between @xmath89 and @xmath15 . such a correlation , with @xmath329 , is expected in the deshpande ( 2000 ) power law model , and would represent the tail - end of the turbulent spectrum of the hi optical depth on larger scales ( 0.02 to 4 pc , deshpande et al . we overplot this predicted level of variations in figure [ f : us_desh2 ] as a sloping line . our compilation of recent interferometric and pulsar observations of tsas does not provide evidence for the power - law turbulent spectrum of @xmath15 on scales @xmath143 to 1000 au . while various sources probe separate lines of sight , we note that b0823 + 26 , b1133 + 16 and b1929 + 10 all probe the local ism ( within 400 pc ) , with similar peak optical depth , and with very similar sensitivity . this essentially provides an almost uniformly sampled region on this graph from @xmath330 to 200 au . similar arguments apply to the higher optical depth portion of this diagram . we are left with the conclusion that tsas is a sporadic phenomenon . it is interesting to note that the smallest spatial scale of tsas detections in both interferometric and b1929 + 10 studies is @xmath330 au . two pulsars ( b2016 + 28 and b0329 + 54 ) have probed scales @xmath331 au , however tsas was not detected . the spatial scales of @xmath330 au for tsas in the cnm are especially interesting , since in the case of hydrodynamic turbulence the cut - off scale of cnm turbulence corresponds to the viscous damping scale of @xmath332 au @xcite . however , in the case of magnetohydrodynamic ( mhd ) turbulence , the turbulent spectrum can cascade below the viscous cut - off scale down to the scale set by magnetic diffusion @xcite . in fact , magnetic and kinetic spectra can be greatly decoupled , with the magnetic spectrum being much shallower than the kolmogorov spectrum . our pulsar experiment shows only a few significant tsas detections , all in the direction of one pulsar b1929 + 10 . this result strongly suggests inhomogeneity in the galactic distribution of tsas . we have examined two proposed origins of tsas ( i ) discrete hi blobs elongated along the line of sight ; and ( ii ) fluctuations at the tail - end of the turbulent cnm spectrum ; but our data do not provide support for either of these models . the principal problem is that we do not find these features to be homogeneously distributed . meanwhile , recent numerical simulations do generate cnm clouds with a wide range of physical properties , including rare examples of very dense and over - pressured tsas - like clouds ; but the direct comparison between observations and simulated data is not straightforward since simulations probe much smaller ism regions on scales somewhat larger than these experiments . the most striking result from our examination of specific lines of sight is a possible correlation between the b1929 + 10 fluctuations and nai interstellar clouds inside the local cavity or at the lb wall . as about one - half of b1929 + 10 s line of sight is associated with the lb , and a significant fraction of this length appears to be traversing the lb wall , we suggest that the observed hi absorption fluctuations are most likely sampling small - scale structure in the lb . numerous observations have indicated that the lb wall is highly inhomogeneous . many interstellar cloudlets found inside the local cavity are thought to be caused by hydrodynamic instabilities fragmenting the lb wall . through various studies , it has been shown that these cloudlets have velocities of up to 25 relative to the sun @xcite . this translates to @xmath333 au yr@xmath0 . coupled with the transverse speed of b1929 + 10 of 37 au yr@xmath0 , this suggests that our observed tsas could result from the pulsar s line of sight crossing very local cloudlets . to investigate the origin of interstellar clouds inside the local cavity , we look to 2d hydrodynamical simulations of the wnm compression by supernova explosions and subsequent fragmentation into small clouds @xcite . they showed that within 0.3 myr a thermally collapsing layer is formed behind the ( supernova ) shock front where cooling dominates heating and temperature monotonically decreases . as the gas cools it becomes thermally unstable , with a temperature in the range of 300 - 6000 k. within @xmath11 myr this layer starts to fragment into many small ( @xmath334 pc ) cloudlets . finally , these cloudlets can even reach thermal equilibrium at a temperature of 20 k and density of 2000 @xmath5 . as the ism is frequently compressed by supernova explosions , fragmentation of the shock - compressed medium can provide a mechanism for the formation of many small cloudlets . in the @xcite simulation , the cloudlets become embedded in the hot high - pressure gas as the shocked layer fragments , leading to a situation similar to what is observed in the lb . the exact temperature inside the local cavity has been uncertain ( see @xcite for references ) , however it is clear that the cavity is occupied by a warm / hot ionized medium . the cold cloudlets will eventually evaporate in the surrounding warm / hot medium . this process has been studied by many authors . the mass - loss rate in the case of classic evaporation @xcite is slightly lower when cooling is included @xcite . the mass - loss rate of @xcite who considered both flux saturation effects and radiative cooling is about 100 times lower than the classic evaporation rate . we continue by assuming the lowest suggested mass - loss rate , which will have the gentlest effect on the cold clouds . the evaporation timescale depends greatly on the cloud size and the temperature of the surrounding medium . for a cloud size of 30 au , @xmath335 - @xmath336 yr , depending upon whether the cloud is surrounded by the hot ionized medium ( him ) or wnm . for a larger cloud with a size of @xmath337 au , @xmath338 ( him ) or 10@xmath339 ( wim ) yr . clearly larger clouds survive longer and can even coalesce to build larger structures , even if surrounded by a hot medium . in about 1 myr larger clouds could cross a large distance , @xmath332 pc , and may even escape outside of the lb into the ism through various chimneys and tunnels resulting from stellar winds and supernovae , and frequently observed in expanding shells . this scenario may be able to explain the existence of hi clouds on scales of thousands of au or more in the ism , as observed recently by @xcite . on the contrary , very small clouds will evaporate quickly and therefore may not be common in the ism . their short inferred lifetimes indicate that they would usually be found close to their formation sites . we have used the arecibo telescope to obtain multi - epoch hi absorption spectra in the direction of six bright pulsars , with the goal of searching for structure on au scales in the cnm . our target sources were observed previously by frail et al . ( 1994 ) , who found pervasive tiny - scale hi structure . we used an advanced spectrometer capable of more accurate measurements in the face of wildly fluctuating pulsar signals , and carefully analyzed noise statistics . to search for variability of hi absorption profiles , which signifies inhomogeneities on au scales in the cnm , we have searched for time - variability of the absorption line equivalent widths and of the absorption spectra themselves . while we had excellent sensitivity , significant variations in both ew and absorption spectra have been found only in the case of one pulsar : b1929 + 10 . these variations imply tsas on spatial scales @xmath340 au ( where @xmath267 is the distance to the tsas feature ) , with an inferred volume density of @xmath341 @xmath5 and thermal pressure of @xmath342 @xmath5 k. our study clearly shows the inhomogeneity in the galactic distribution of tsas : the detection rate toward four of our pulsars is zero , while the detection rate toward b1929 + 10 is high . such inhomogeneity is hard to explain with the model of discrete elongated disks or cylinders being a _ general _ property of cnm clouds ( heiles 1997 ) ; instead large spatial variations of tsas physical properties or abundances must be invoked . while the heiles ( 1997 ) model can help in bringing down the volume density and pressure of tsas , the puzzles regarding its distribution and origin still remain . our examination of all pulsar and interferometric tsas detections and upper limits does not show a correlation between the level of optical depth fluctuations and the tsas spatial scale , as would be expected if the turbulent spectrum on much larger scales is extrapolated to au - scales . the detections and non - detections probe an almost continuous range of spatial scales from @xmath343 to 1000 au . therefore , the large number of non - detections of tsas suggests that the cnm clouds on scales @xmath344 to @xmath336 au are not a pervasive property of the ism . the sporadic tsas detections on scales of tens of au may indicate occasional local bursts of turbulent energy dissipation , rather than end - points of the turbulent spectrum . another striking result from our study is a possible correlation between b1929 + 10 s tsas and interstellar clouds observed in nai absorption inside the local bubble . we find evidence that the tsas is likely to be within 106 pc of the sun , and is sampling the small - scale structure of the lb caused by hydrodynamic instabilities fragmenting the lb wall . we propose that the line of sight of b1929 + 10 is revealing this recently formed small - scale structure . similar bubbles and their walls are found throughout the milky way , but the lifetime of a tsas cloud created from them depends strongly on the its size and the temperature of the surrounding medium . larger fragments ( size @xmath345 au ) survive longer and can travel large ism distances , becoming a more general ism property . on the other hand , the smallest clouds ( size @xmath13 au ) evaporate quickly close to their formation site , and are therefore not very commonly observed in the ism . additional processes , such as stellar mass - loss and collisions of interstellar clouds / filaments , probably contribute to the cnm structure formation on somewhat larger ( sub - pc ) scales . we are grateful to caltech s center for advance computation and research for the use of their facilities for data storage and processing . we express our thanks to stuart anderson , rick jenet , and kathryn devine for help with observations and data processing . we acknowledge the use of the atnf pulsar catalog . we thank carl heiles , j .- macquart , miller goss , jay gallagher , and shu - ichiro inutsuka for stimulating discussions . we would also like to thank the anonymous referee , whose comments improved greatly the clarity of the paper . ss acknowledges support from nsf grants ast-0097417 and ast-9981308 , and the research corporation . she also thanks astron and the helena kluyver visiting program for their hospitality during the final stage of the manuscript preparation . jmw , zp , kt , and jtg were supported by nsf grants ast-0406832 and ast-0807556 . , c. l. , goss , w. m. , lazio , t. j. w. , & faison , m. d. 2007 , in astronomical society of the pacific conference series , vol . 365 , sins - small ionized and neutral structures in the diffuse interstellar medium , ed . m. haverkorn & w. m. goss , 12 , c. r. , reynolds , j. e. , & wilson , w. w. 2007 , in astronomical society of the pacific conference series , vol . 365 , sins - small ionized and neutral structures in the diffuse interstellar medium , ed . m. haverkorn & w. m. goss , 90 , s. , heiles , c. , & kanekar , n. 2007 , in astronomical society of the pacific conference series , vol . 365 , sins - small ionized and neutral structures in the diffuse interstellar medium , ed . m. haverkorn & w. m. goss , 22 , j. m. & stanimirovi , s. 2007 , in astronomical society of the pacific conference series , vol . 365 , sins - small ionized and neutral structures in the diffuse interstellar medium , ed . m. haverkorn & w. m. goss , 28	we present results from multi - epoch neutral hydrogen ( hi ) absorption observations of six bright pulsars with the arecibo telescope . moving through the interstellar medium ( ism ) with transverse velocities of 10150 au yr@xmath0 , these pulsars have swept across 1200 au over the course of our experiment , allowing us to probe the existence and properties of the tiny scale atomic structure ( tsas ) in the cold neutral medium ( cnm ) . while most of the observed pulsars show no significant change in their hi absorption spectra , we have identified at least two clear tsas - induced opacity variations in the direction of b1929 + 10 . these observations require strong spatial inhomogeneities in either the tsas clouds physical properties themselves or else in the clouds galactic distribution . while tsas is occasionally detected on spatial scales down to 10 au , it is too rare to be characterized by a spectrum of turbulent cnm fluctuations on scales of @xmath1 au , as previously suggested by some work . in the direction of b1929 + 10 , an apparent correlation between tsas and interstellar clouds inside the warm local bubble ( lb ) indicates that tsas may be tracing the fragmentation of the lb wall via hydrodynamic instabilities . while similar fragmentation events occur frequently throughout the ism , the warm medium surrounding these cold cloudlets induces a natural selection effect wherein small tsas clouds evaporate quickly and are rare , while large clouds survive longer and become a general property of the ism .
this supporting information contains : * details of the models described in the main text , and the methods used to identify locally - favored structures . * illustrative results of mutual information between propensity and particle type * discussion of the @xmath40-dependence of the results shown in fig . 1(b , c ) * analysis of the numerical method that we use when estimating mutual information . the ka mixture is defined as in [ 25 ] . the system consists of @xmath79 particles of which @xmath25 are of type a and @xmath80 of type b. the particles interact by lennard - jones potentials with parameters @xmath81 and @xmath82 . temperatures are quoted in units of @xmath83 , with boltzmann s constant @xmath84 . the onset temperature of glassy dynamics is @xmath85 and the mode - coupling temperature has been estimated [ 25 ] to be @xmath86 . the total number density of particles is @xmath87 , and we use a cubic simulation box with periodic boundaries . the system evolves by monte carlo dynamics as in [ 27 ] , with trial displacements drawn from a cube of side @xmath88 centred at the origin . the mean - square displacement per trial move is @xmath89 and we define the fundamental time unit @xmath90 where @xmath91 is the diffusion constant of free particle . the result is that @xmath4 corresponds to @xmath92 proposed mc moves per particle . the structural relaxation time @xmath39 discussed in the main text is defined as @xmath93 where @xmath94}\rangle$ ] is evaluated by an average over particles of type a , and @xmath95 . the polydisperse hard sphere system consists of an equimolar mix of five particle species with diameters @xmath96 , all with equal masses @xmath97 . the particles interact as hard spheres and the system evolves by event - driven molecular dynamics ( implemented by dynamo [ 28 ] ) . the system comprises @xmath98 particles and the simulation box is cubic with periodic boundary conditions . the time unit in the system is @xmath99 . the structural relaxation time is evaluated at @xmath100 . here , we briefly describe the structural measurements @xmath60 and @xmath62 that we use to identify locally - favoured structures in these systems . these measurements are based on voronoi analyses of the system . in the ka model , we perform this analysis after quenching the system to its nearest energy minimum ( inherent structure ) . we follow [ 7 ] in using a voronoi analysis where faces between a and b particles are located closer to the b particles , consistent with their smaller size . in the hs system , we use a regular voronoi analysis , in which faces are midway between neighbouring particles . we identify @xmath57 voronoi polyhedra in the ka system as those with ten faces , of which exactly two have four edges , and eight have five edges . this particle and its ten voronoi neighbours form a cluster ( 11a in the topological cluster classification [ s1 ] ) , and we set @xmath58 for all particles in these clusters . to identify the pentagonal bipyramids in which particle @xmath6 participates ( in both hs and ka systems ) , we identify @xmath56 as the number of pentagonal faces on the voronoi cell of that particle . this gives the number of neighbours of particle @xmath6 that share exactly five mutual neighbours with particle @xmath6 . the procedure is equivalent to counting the number of ` 1551 ' bonds in the common neighbour analysis ( cna ) [ 34,35 ] , and is similar to the identification of ` 7a ' clusters in the topological cluster classification [ s1 ] . to demonstrate the physical meaning of mi , fig . [ fig : mi - type](a ) shows @xmath24 for the ka system , where the structural measurement @xmath12 is taken to be the particle type @xmath101 . the b - particles are more mobile in this system , and as shown in the inset , at large times @xmath102 , the propensity distributions for the two kinds of particle have almost zero overlap . thus , for these very long times , measuring the particle type splits the propensity distribution into two distinct components : this provides @xmath103 bits of information ( similar to a mixing entropy ) , where @xmath104 and @xmath105 are the fractions of particles in each component . if the components were equal in size , the mi would be exactly 1 bit : here the b particles are less numerous ( @xmath106 ) so the mi is less , approximately @xmath107 bits . for times @xmath9 close to the structural relaxation time @xmath39 , fig . [ fig : mi - type](b ) shows that the propensity distributions of the two types differ from each other , but there is a region of significant overlap . in this case , measuring the particle type provides @xmath108 bits of information about @xmath21 . and the particle type @xmath27 in the ka system . ( inset ) mean square displacement of all particles , with arrows indicating the structural relaxation time @xmath109 and the lag time @xmath110 . ( b ) distributions of the propensity at @xmath111 and @xmath112 ( inset ) : see text for discussion . , title="fig:",width=264 ] and the particle type @xmath27 in the ka system . ( inset ) mean square displacement of all particles , with arrows indicating the structural relaxation time @xmath109 and the lag time @xmath110 . ( b ) distributions of the propensity at @xmath111 and @xmath112 ( inset ) : see text for discussion . , title="fig:",width=264 ] [ [ predictive - power - of - local - energydensity - dependence - on - length - scale - ell ] ] predictive power of local energy / density : dependence on length scale @xmath40 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the coarse - grained measurements of local energy and density @xmath115 and @xmath116 discussed in the main text depend ( by definition ) on a length scale @xmath40 that indicates the size of the local coarse - graining region . the effect of varying this length scale is illustrated in figure [ fig : supp - mi - ell ] , where the mi between propensity and @xmath117 is plotted , for various @xmath40 . for short lengths ( here , @xmath118 ) , the local density is strongly predictive on short times , presumably due to the influence of free volume on the vibrational motion of particles within a single metastable state . for longer times , the mi increases with @xmath40 before saturating at @xmath119 . in the long time regime , and for all cases considered , the mi for @xmath48 is almost always larger than for smaller @xmath40-values , and increasing @xmath40 above @xmath120 does not significantly increase the mi ( as in fig . [ fig : supp - mi - ell ] ) . this observation motivated our choice of @xmath48 for the data shown in figs . 2(c , f ) . the mi between coarse - grained energy and propensity does not show the early - time signal found for @xmath121 at @xmath118 but otherwise behaves similarly to @xmath121 . calculating mutual information ( mi ) from numerical data requires some care , since estimators are vulnerable to systematic errors if sample sizes are not sufficiently large . a variety of estimators have been developed ( see for example [ s1-s6 ] ) , many of which use bayesian methods , exploiting prior knowledge ( or assumptions ) about the form of the underlying distributions in order to better estimate either entropies or mutual informations @xcite . in this work , we use a simple method that we have tailored to the problem of interest here , based on the method of @xcite . in all measurements , we discretise the propensity , forming a histogram with bins of width @xmath122 , where @xmath123 is the mean propensity . the width of the bins is comparable with the numerical uncertainties in our estimates of the @xmath124 , which are obtained from between 100 and 250 independent trajectories . for this reason , storing the propensities to greater accuracy than the bin width would not make our measurements of mi any more accurate the binning does not introduce numerical artefacts , and is convenient in what follows . further , since the same binning is used for all mi measurements , we are able to make a fair comparison between the different structural measurements shown in figures 1 and 2 of main text . in the following , we use @xmath125 as an integer - valued label for the bin in which the propensity @xmath124 is located ( for example , one may take @xmath126 , the largest integer that is less than or equal to @xmath127 ) . we first describe the estimator that we use for calculating mi between two discrete - valued variables . we have in mind that @xmath12 is a structural observable with a discrete set of possible values , while @xmath125 is the propensity bin - index as described above . however , the discussion is general for joint distributions of discrete random variables . for each particle , suppose that we measure two integers @xmath125 and @xmath12 . then given data for @xmath69 particles , let @xmath128 be the number of particles with @xmath129 ; also let @xmath130 be the number of particles with @xmath131 , and similarly @xmath132 . the simplest mi estimate based on these data is the `` plugin estimator '' : @xmath133 where the sum runs over all pairs @xmath134 for which @xmath135 . given sufficient data , @xmath136 converges to the mutual information @xmath137 , however , this convergence is often quite slow , requiring very large @xmath69 for an accurate estimate . in particular , even if the data set is constructed so that @xmath125 and @xmath12 are independent , one typically finds @xmath138 , recovering @xmath139 only as @xmath140 . the key point is that for large enough data sets ( @xmath140 ) one has @xmath145 by the law of large numbers , so that @xmath146 is an entropy estimator for @xmath147 . similarly , @xmath148 converges to a weighted sum of conditional entropies of the form @xmath149 , as long as @xmath150 for all @xmath18 . the difficulty is that the convergence of @xmath146 and @xmath148 to their respective limits are ruled by different large parameters ( @xmath69 and @xmath151 ) , and there are systematic errors associated with this convergence if these parameters are not large enough . in general , @xmath146 and @xmath148 both underestimate the relevant entropies , but the error on @xmath146 is smaller , resulting in a positive systematic error for @xmath152 . and the propensity , using the estimator @xmath153 . we show one estimate based on a sample of @xmath154 type - a particles ( from 8 independent initial conditions ) , and four estimates each based on @xmath155 particles ( each estimate based on just one initial condition ) . the error bars are conservative estimates of numerical uncertainty ( see text ) and are reduced on including more data . the mean estimate has only a weak drift with sample size indicating that systematic errors arising from small samples are small . ( b ) estimates of the same mi based on the same data but using the estimator @xmath136 . the estimation is poor in this case : there is a strong systematic error coming from the finite sample size . the error decreases as more data is included but the largest sample used is not large enough to saturate the limit @xmath140 and does not provide a reliable estimate of the mi . , title="fig:",width=226 ] and the propensity , using the estimator @xmath153 . we show one estimate based on a sample of @xmath154 type - a particles ( from 8 independent initial conditions ) , and four estimates each based on @xmath155 particles ( each estimate based on just one initial condition ) . the error bars are conservative estimates of numerical uncertainty ( see text ) and are reduced on including more data . the mean estimate has only a weak drift with sample size indicating that systematic errors arising from small samples are small . ( b ) estimates of the same mi based on the same data but using the estimator @xmath136 . the estimation is poor in this case : there is a strong systematic error coming from the finite sample size . the error decreases as more data is included but the largest sample used is not large enough to saturate the limit @xmath140 and does not provide a reliable estimate of the mi . , title="fig:",width=226 ] to reduce this effect , we define an alternative estimator @xmath156 \label{equ : i1}\ ] ] here , the @xmath157 are obtained from a random null data set , as follows . for each @xmath18 , we draw a random sample of size @xmath151 ( without replacement ) from the original data set , and @xmath157 is defined as the number of particles in that random sample that have @xmath131 . for a large enough sample , @xmath158 . however , the advantage of the method is that the convergence of @xmath159 to @xmath160 and @xmath161 to @xmath162 are now ruled by the same parameter @xmath151 , and the result is that the systematic errors arising from the two terms in ( [ equ : i1 ] ) tend to cancel each other . we note that if instead of drawing a random sample we simply set @xmath163 , independent of @xmath18 , then we recover the original estimator @xmath136 . it is also notable that if @xmath97 and @xmath18 are independent then the conditional distribution @xmath159 should be statistically equivalent to the distribution obtained in the random sample @xmath157 . this means that @xmath153 is free from systematic error in the case where @xmath18 and @xmath97 are independent . this is the most important case for the calculations of this paper , because the mi values found are typically quite small , and systematic errors when @xmath164 correspond to false - positive signals of correlation between structure and dynamics , which can be misleading . for each estimate of mi , we compute @xmath153 using several random null data sets ( typically 100 realisations are sufficient ) . the average value of @xmath153 over the realisations provides our estimate of @xmath165 while the standard deviation among the values of @xmath153 gives an estimate on the uncertainty of this estimate . we therefore use this standard deviation as the error bar for the estimate of @xmath165 . we emphasise that the @xmath159 are determined by the original data and are the same for every realisation of the null data : it is the finite size of this original data set that introduces a finite uncertainty on estimates of @xmath165 . this uncertainty is not reduced by repeated sampling over different null data sets , so it is the standard deviation of @xmath153 that gives the relevant error estimate , not the standard error . [ fig : cna - estim ] shows estimates for the mi between @xmath60 and the propensity , obtained by the estimators @xmath153 and @xmath136 , for data sets of two different sizes . it can be seen that @xmath136 is not sufficient for the purposes used here , even for the larger data set , while @xmath153 gives a consistent estimate of the mi for data sets of both sizes considered . the estimate of the uncertainty based on @xmath153 is also self - consistent , in that error bars from independent estimates typically overlap with each other . an alternative to ( [ equ : i1 ] ) can be obtained by interchanging @xmath18 and @xmath97 , since the mi is symmetric : @xmath166 \label{equ : i1-prime}\ ] ] we use @xmath153 for all calculations of mi between discrete variables , but it is useful to define @xmath167 in preparation for later sections . and the propensity , using the estimator @xmath168 . we show one estimate based on a sample of @xmath154 type - a particles ( from 8 independent initial conditions ) , and four estimates each based on @xmath155 particles ( from independent initial conditions ) . the error bars are conservative estimates of numerical uncertainty ( see text ) and are reduced on including more data . systematic errors from small samples are weak . , width=226 ] we now turn to the case where the structural variable of interest takes continuous values . the analogue of the estimator @xmath167 in ( [ equ : i1-prime ] ) is @xmath169 where @xmath170 is an estimator for @xmath171 $ ] . to define @xmath170 , we again take a random sample of size @xmath172 from the original data , to ensure that systematic errors on estimates for @xmath173 and @xmath174 should cancel as far as possible . then , if @xmath175 is an ordered list of the values of @xmath12 for those particles with @xmath131 , and @xmath176 is an ordered list of the values of @xmath12 in the random sample , we define @xmath177\ln2}\sum_{i=1}^{n(m)-1 } [ \ln(s^{i+1}-s^i ) - \ln(\tilde{s}^{i+1}-\tilde{s}^i ) ] .\ ] ] this converges to the required result as @xmath178 because if @xmath179 is an ordered list of independent random samples from @xmath180 then , as @xmath178 , one has @xmath181 where @xmath182 is the digamma function , which satisfies @xmath183 and @xmath184 where @xmath185 is euler s constant @xcite . [ fig : dw - estim ] shows results using the estimator @xmath168 . as with @xmath153 , the uncertainty on the estimate of mi is reduced on including more data , and the systematic variation on increasing @xmath69 is weak , indicating that the estimator is reliable .	we analyse the connections between structure and dynamics in two model glass - formers , using the mutual information between an initial configuration and the ensuing dynamics to compare the predictive value of different structural observables . we consider the predictive power of normal modes , locally favoured structures , and coarse - grained measurements of local energy and density . the mutual information allows the influence of the liquid structure on the dynamics to be analysed quantitatively as a function of time , showing that normal modes give the most useful predictions on short time scales while local energy and density are most strongly predictive at long times . as supercooled liquids approach their glass transitions , structural relaxation slows down dramatically , but molecular configurations remain disordered and apparently random @xcite . however , computer simulations @xcite and experiments @xcite show that liquid structure and dynamical relaxation are correlated in these systems , as predicted ( or assumed ) in several theories @xcite . however , correlations between structure and dynamics do not by themselves imply a causal relationship @xcite : other theories @xcite assume that local structure plays only a peripheral role in dynamical relaxation . correlations between structure and dynamics can be demonstrated at a microscopic level @xcite , by exploiting the dynamically heterogeneous nature of glassy relaxation @xcite . that is , individual particles have different propensities for motion @xcite , depending on local structure . here , we use _ information theory _ @xcite to analyze the strength of these correlations , by measuring the extent to which structural measurements can be used to _ predict _ particle dynamics at subsequent times . this quantitative analysis provides a stringent test of proposed causal links between structural features and slow dynamics , in contrast to previous analyses based on restricted subsets of particles or snapshots of the system . in two model glass - formers , we find that coarse - grained measurements of energy and density @xcite give the most predictive information for long times . in one of the models , we also find that vibrational modes @xcite are strongly correlated with motion on relatively short time scales . compared to these effects , the correlation between dynamics and low energy ( or low enthalpy ) local structures is relatively weak . we present results for the kob - andersen ( ka ) mixture of lennard - jones particles @xcite , and an equimolar five - component hard sphere ( hs ) mixture , which mimics colloidal suspensions @xcite . both systems contain particles of different sizes , with the diameter of the largest particles being @xmath0 ( which sets the unit of length ) . the ka system evolves with overdamped ( monte carlo ) dynamics as in @xcite ; we focus on a temperature @xmath1 . the hs system evolves by event - driven molecular dynamics @xcite ; we consider volume fractions @xmath2 in the range @xmath3 . in both systems , we use @xmath4 to indicate the fundamental unit of time . the relaxation at the state points that we consider is up to @xmath5 decades slower than relaxation at the onset of glassy dynamics , where it is of order @xmath4 ( in both systems ) . further system details are given in the supporting information ( si ) @xcite . to characterize particle dynamics in these systems , we define the dynamical propensity @xcite of particle @xmath6 as @xmath7 , where @xmath8 is the particle position at time @xmath9 , and the isoconfigurational average is calculated over many independent dynamical simulations , all with the same initial particle positions but with independent random initial velocities ( and independent stochastic dynamics in the ka system ) . the role of the `` lag time '' @xmath10 is discussed in si @xcite : we take @xmath11 . we use @xmath12 to denote a structural measurement at time @xmath13 , which depends in general on particle @xmath6 and all particles in its vicinity . to quantify the strength of the correlation between @xmath12 and the dynamical propensity @xmath14 , we use mutual information ( mi ) measurements @xcite . the mi is defined as @xmath15 where @xmath16 is the joint probability distribution of @xmath17 and @xmath18 , while @xmath19 and @xmath20 are its marginal distributions . we assumed here that @xmath12 takes discrete values : for continuous attributes @xmath12 , the sum over @xmath18 replaced by an integral . the mi gives `` the average amount of information about the propensity @xmath21 that is provided by a measurement of @xmath12 '' . since @xmath12 depends only on the initial condition , the mi measures _ predictive _ information . the mi may be evaluated for any structural observable @xmath12 , and it makes no assumptions on the nature of the correlation between @xmath14 and @xmath12 . as such , it represents a generally - applicable figure - of - merit for comparing the influence on dynamics of different structural measures , going beyond previous comparisons of snapshots @xcite or analyses of selected subsets of particles @xcite . this use of ( [ equ : mi - mu ] ) as a quantitative measure of information @xcite is similar to the use of entropy as a measure of disorder in statistical mechanics , with the role of disorder being taken by the variation in propensity between different particles . particles with the same value of @xmath12 typically have less variation in their propensity , so specifying @xmath12 reduces the variation in @xmath21 , just as introducing a constraint in statistical mechanics reduces the entropy @xcite . the mi is equivalent to this entropy reduction . information is conventionally measured in bits , with one bit corresponding to a reduction in entropy of @xmath22 . our procedure for estimating mi is described in the si : the method ensures as far as possible that we obtain @xmath23 if @xmath17 and @xmath18 are independent ; it also provides an estimate of the numerical uncertainty in the mi . to illustrate the use of mi , let @xmath12 be the type ( a or b ) of particle @xmath6 in the ka system . the different types have different dynamical relaxation so measuring the particle type provides predictive information about particle dynamics . in si @xcite , we show that measuring the type of particle @xmath6 provides between 0.1 and 0.7 bits of information about the propensity @xmath14 , depending on the time @xmath9 . this value is a useful baseline in interpreting the results that follow : if a structural measurement is strongly coupled with dynamics , we argue that @xmath24 should be at least of order 0.1 bit , while mis much less than this are indicative of weak coupling . figure [ fig : mi - modes - hydro ] shows mi measurements between particle propensities and several aspects of liquid structure , for both ka and hs systems . since the influence of particle type on dynamics is not directly related to glassy behaviour , we measure mutual information where the predictability based on particle type has already been taken into account . that is , for several different @xmath12 , we measure `` the information about @xmath14 that is provided by a measurement of @xmath12 , for a particle whose type is already known '' . in the ka system , we achieve this by restricting the distributions in ( [ equ : mi - mu ] ) to particles of type a , which form the majority ( @xmath25 ) of the system . in the hs system , we use a ` conditional mi ' , @xmath26 where @xmath27 indicates the particle type @xcite . our choices of @xmath12 reflect different theoretical pictures of glassy systems : we now discuss the implications of these results for those theories . several links have been proposed between normal modes in glassy systems and their material properties @xcite . low - frequency modes in a supercooled liquid define a set of `` soft directions '' on its potential energy surface ( or energy landscape ) , and both thermal fluctuations and structural relaxation couple significantly to these modes @xcite . these modes also play a central role in the analogy between glassy behaviour and jamming @xcite . we analyze them @xcite by quenching the ka system to its nearest energy minimum ( inherent structure ) , and diagonalizing the hessian matrix of the energy at that minimum . the resulting eigenvectors and eigenvalues are @xmath28 and @xmath29 , for @xmath30 , and one defines a `` local debye - waller ( dw ) factor '' @xmath31 that indicates @xcite the expected size of fluctuations in the position of particle @xmath6 , based on an expansion about the energy minimum . ( here @xmath32 is a vector containing the three components of @xmath28 associated with particle @xmath6 . ) since low frequency modes couple most strongly to structural relaxation @xcite , we also define a generalised dw factor @xmath33 , which is calculated using only the @xmath34 modes with lowest @xmath35 . in hs systems , normal modes can not be defined by reference to a potential energy surface so we do not consider them here , although alternative definitions of normal modes are possible @xcite . figure [ fig : mi - modes - hydro](a ) shows that for relatively short time scales @xmath36 in the ka model , the mutual information between propensity and dw factors is large ( up to 0.5 bits ) , so @xmath37 and @xmath38 are strongly correlated with particle motion . this indicates that the normal modes accurately mimic the fluctuations of the system within its initial metastable state . on longer time scales , the information provided by these measurements decreases strongly , but @xmath38 still provides more than 0.1 bits at the structural relaxation time @xmath39 , confirming that the low frequency normal modes do have significant predictive power for structural relaxation @xcite . coarse - grained energy and density measurements are also correlated with dynamical fluctuations @xcite . we define a local density , coarse - grained on a scale @xmath40 , as @xmath41 , where the sum runs over all particles @xmath42 and @xmath43 is the distance between particles @xmath6 and @xmath42 @xcite . similarly , the locally - averaged energy is @xmath44 where @xmath45 is the energy of particle @xmath42 . figures [ fig : mi - modes - hydro](b , c ) show that for @xmath46 these coarse - grained quantities have strong predictive power on time scales longer than the structural relaxation time , but the mi is smaller for relaxation times up to and including @xmath39 . the results are broadly similar for both models ( for the hs model , error bars are shown only at @xmath47 , to indicate the @xmath2-dependence of this time scale ) . we show data for @xmath48 since this gives a significant mi throughout this range of data : dependence of the mi on @xmath40 is discussed in si @xcite . throughout the glassy regime , we expect @xmath49 and @xmath50 to have peaks at some time @xmath51 , before decreasing at longer times ( see for example the hs data at @xmath52 ) . however , for the largest volume fractions it is clear that @xmath51 is significantly larger than @xmath39 , and is larger than our sampling window . we attribute this large @xmath51 to hydrodynamic effects that are largely independent of glassy behavior : regions of size @xmath40 with high density or low energy relax on a time scale @xmath53 where @xmath54 is a diffusion constant . one therefore expects relaxation in such regions to be predictably slower than average up to times @xmath55 , which is significantly larger than @xmath39 . our focus here is on predictability on time scales of order @xmath39 , where the system is has significant dynamical heterogeneity and the motion is complex and co - operative . for this reason , we have not explored the large - time hydrodynamic behaviour in detail . we do note that for the hs system , the mi at @xmath39 increases at large @xmath2 , indicating that the coupling of dynamics to local density is increasing as the glass transition is approached , consistent with @xcite . however , even for the largest @xmath2 , the mi is less than 0.1 bit at @xmath39 , although it does grow rapidly for larger times . , and all are less than 0.1 bit . ( a ) ka system , using lfs with ` 155 ' and ( 0,2,8 ) signatures , as described in the main text and illustrated in the figure . ( b ) hs system , using lfs with cna-155 signatures . representative error bars are shown in ( a ) , while in ( b ) we show error bars only at @xmath47.,width=283 ] an alternative picture of glassy relaxation is based around locally - favoured structures ( lfs ) : atomic or molecular packings that have low energy ( or enthalpy ) @xcite . particles in lfs typically have slower than average dynamics in glassy systems @xcite . for both ka and hs models , we consider an lfs based on a pentagonal bipyramid , identified by the signature 1551 in the analysis of @xcite . these structures are indicative of local fivefold symmetry @xcite . let @xmath56 be the number of pentagonal bipyramids in which particle @xmath6 participates @xcite : we expect larger @xmath56 to be associated with lower propensity for motion . in the ka model , we also consider an lfs ( bicapped square antiprism ) that has been found to be correlated with slow dynamics @xcite . these lfs are associated with voronoi polyhedra whose signature is @xmath57 in the notation of @xcite . we define @xmath58 if particle @xmath6 participates in such an lfs , with @xmath59 otherwise . for the ka model , we calculate @xmath60 and @xmath61 using the inherent structure of the system . figure [ fig : mi - lfs](a ) shows results for the ka model , indicating that @xmath60 and @xmath62 are correlated with particle motion @xcite . as with the low - frequency normal modes , the signal is largest on time scales @xmath63 indicative of @xmath64-relaxation , but there is still some correlation at the structural relaxation time . however , the strength of the correlation is smaller for the lfs than for the normal modes , less than 0.1 bit in all cases . figure [ fig : mi - lfs](b ) shows similar results for the hs system . the mi values are larger than those of the ka system , indicating that lfs have more predictive power for dynamics . at short times , the mi increases with increasing volume fraction ; however the mi at @xmath39 ( indicated by the error bars ) depends more weakly on @xmath2 . we argue that the small mi values at @xmath39 and longer times , and the absence of an increase of the values at @xmath39 with volume fraction , both indicate that the lfs identified here are more weakly coupled to the dynamics than the normal modes , at least for the degree of supercooling accessed here . to summarize our findings so far , figure [ fig : mi - modes - hydro ] shows that debye - waller factors and coarse - grained measurements of energy and density have significant coupling to dynamics , providing predictive information comparable with measurements of particle type in the ka model . however , the information available from the different measurements has very different time - dependences . for short times , the normal mode analysis captures fast vibrational motion accurately , but the predictive power of this analysis decreases strongly with time . this indicates that as structural relaxation starts to take place , the ` soft directions ' for further relaxation quickly diverge from those that were present at @xmath13 . on the other hand , coarse - grained energy and density measurements have almost no predictive value at short times , but the slow decay of large - scale hydrodynamic fluctuations means that they can influence particle dynamics quite strongly even on time scales much longer than @xmath39 : on these time scales , almost all memory of the initial structure has been lost , leaving only the hydrodynamic fluctuations in energy or density . for the state points considered here , figure [ fig : mi - lfs ] shows that lfs measurements have less predictive power for dynamics in these models , and that this predictive power is largest on relatively short time scales associated with @xmath64-relaxation . our interpretation is that since the lifetimes of most lfs are less than @xmath39 @xcite , the influence of lfs on dynamics is ( in most cases ) similarly short - lived , limiting the predictive power of such measurements for dynamics . . ( a ) ka system for @xmath1 and @xmath65 ; ( b ) hs system at @xmath66 . see text for discussion . , title="fig:",width=158 ] . ( a ) ka system for @xmath1 and @xmath65 ; ( b ) hs system at @xmath66 . see text for discussion . , title="fig:",width=158 ] finally , in contrast to the measurements so far , we show how information theory can also be used to analyze how predictable particle motion is in these models , independent of any specific structural observable . let @xmath67 be the ( isoconfigurational ) distribution of particle displacements @xmath68 . given data for @xmath69 particles ( which may be obtained in general from many initial configurations ) , we define @xmath70 which is the `` average amount of information about a particle s motion that is provided by specifying its initial environment '' . since a particle s initial environment encodes all predictable aspects of its future motion , @xmath71 indicates how predictable ( or reproducible ) particle motion is within the system @xcite . figure [ fig : mi - dyn ] shows that @xmath71 is much larger at low temperatures in the ka model than at high temperatures , indicating that structure is more strongly coupled to dynamics at low temperatures . it is useful to compare @xmath71 with `` the average amount of information about a particle s dynamics that is provided by specifying its propensity '' , which is @xmath72 where @xmath73 is the joint distribution of displacement @xmath74 and propensity @xmath17 , and @xmath75 is the marginal distribution of the displacement . since fixing a particle s initial environment necessarily fixes its propensity , one has @xmath76 from fig . [ fig : mi - dyn ] , the two quantities in ( [ equ : mu - i ] ) are almost equal for the ka model . eq . ( [ equ : mu - i ] ) is an `` information - processing inequality '' @xcite , so this result indicates that the propensity captures almost all predictable information about single - particle displacements . for the hs system , we use a conditional mi between @xmath74 and @xmath17 , to account for particle type , as above . the two mis in ( [ equ : mu - i ] ) differ somewhat more strongly than they do in the ka model : this situation might arise ( for example ) if some particles have finite average displacements @xmath77 that are only weakly correlated with their propensities . nevertheless , we have @xmath78 in both models , indicating that the propensity captures all predictable ( reproducible ) aspects of the single - particle dynamics @xcite . this further validates the use of the mutual information as a general figure - of - merit for evaluating proposed connections between structure and dynamics . given the implications of figs . [ fig : mi - modes - hydro ] and [ fig : mi - lfs ] for the strength and time - dependence of the coupling between structure and dynamics , we hope that future studies will exploit these information - theoretic measurements to further elucidate which ( if any ) structural features are responsible for the strong dynamical slowing in supercooled liquids . we thank peter harrowell , peter sollich , gilles tarjus , and karoline wiesner for helpful discussions . rlj and ajd were supported by the epsrc through grants ep / i003797/1 and ep / e501214/1 respectively . cpr gratefully acknowledges the royal society for financial support . 99 m. d. ediger , c. a. angell and s. r. nagel , j. phys . chem . * 100 * , 13200 ( 1996 ) . p. g. debenedetti and f. h. stillinger , nature * 410 * , 259 ( 2001 ) . a. widmer - cooper , p. harrowell and h. fynewever , phys , rev . lett . * 93 * 135701 ( 2004 ) ; a. widmer - cooper and p. harrowell , j. chem . phys , * 126 * 154503 ( 2007 ) . d. coslovich and g. pastore , europhys . lett . * 75 * , 7840 ( 2006 ) . c. brito and m. wyart , j. stat . mech . ( 2007 ) l08003 . a. widmer - cooper , h. perry , p. harrowell and d. r. reichman , nature physics * 4 * , 711 ( 2008 ) . d. coslovich and g. pastore , j. chem . phys . * 127 * , 124504 ( 2007 ) . a. malins , j. eggers , h. tanaka and c. p. royall , faraday . discuss . * 167 * , in press ( 2013 ) . m. leocmach and h. tanaka , nature communications * 3 * , 974 ( 2012 ) . r. candelier , a. widmer - cooper , j. k. kummerfield , o. dauchot , g. biroli and d. r. reichman , phys . rev . lett . * 105 * , 135702 ( 2010 ) . v. lubchenko and p. g. wolynes , ann . rev . phys . chem . * 58 * , 235 ( 2007 ) j .- p . bouchaud and g. biroli , j. chem . phys . * 121 * , 7347 ( 2004 ) . w. gtze and l. sjgren , rep . . phys . * 55 * , 241 ( 1995 ) . g. tarjus , s. a. kivelson , z. nussinov and p. viot , j. phys . : cond . matt . * 17 * , r1143 ( 2005 ) . t. kawasaki , t. araki and h. tanaka , phys . rev . lett . * 99 * , 215701 ( 2007 ) m. l. manning and a. j. liu , phys . rev . lett . * 107 * , 108302 ( 2011 ) . g. biroli , j.p . bouchaud , k. miyazaki , and d. r. reichman , phys . rev . lett . * 97 * , 195701 ( 2006 ) . d. j. ashton and j. p. garrahan , eur j. phys . e * 30 * , 30 ( 2009 ) . j. p. garrahan and d. chandler , proc . nat . acad . sci . usa * 100 * , 9710 ( 2003 ) ; d. chandler and j. p. garrahan , ann . rev . phys . chem . * 61 * , 191 ( 2010 ) . m. d. ediger , annu . rev . phys . chem . * 51 * , 99 ( 2000 ) . t. m. cover and j. a. thomas , _ elements of information theory _ ( wiley , new york , 1991 ) . g. s. matharoo , m. s. g. razul , and p. h. poole , phys . rev . e * 74 * , 050502 ( 2006 ) . l. berthier , g. biroli , j .- p . bouchaud , w. kob , k. miyazaki and d. r. reichman , j. chem . phys * 126 * 184503 ( 2007 ) . l. berthier and r. l. jack , phys . rev . e * 76 * , 041509 ( 2007 ) . w. kob and h. c. andersen , phys . rev . e * 51 * , 4626 ( 1995 ) ; * 52 * , 4134 ( 1995 ) . c. p. royall , a. malins , a. j. dunleavy , and r. pinney , in _ fragility of glass - forming liquids _ , editors : k. a. kelton , l. grier , s. sastry ( hindustan book agency , new delhi , 2014 ) . l. berthier and w. kob , j. phys : cond . matt * 19 * , 205130 ( 2007 ) . m. n. bannerman , r. sargant and l. lue , j. comp . chem . * 32 * , 3329 ( 2011 ) . d. chandler , _ an introduction to modern statistical mechanics _ , ( oup , oxford , 1987 ) . a. widmer - cooper and p. harrowell , phys . rev . lett . * 96 * , 185701 ( 2006 ) . l. berthier , g. biroli , j .- p . bouchaud , l. cipelletti , d. el masri , d. lhte , f. ladieu and m. pierno , science * 310 * , 1797 ( 2005 ) . c. p. royall , s. r. williams , t. ohtsuka and h. tanaka , nature materials * 7 * , 556 ( 2008 ) . a. malins , j. eggers , c. p. royall , s. r. williams and h. tanaka , j. chem . phys . * 138 * , 12a535 ( 2013 ) j. d. honeycutt and h. c. andersen , j. phys . chem * 91 * , 4950 ( 1987 ) . h. jonsson and h. c. andersen , phys . rev . lett . * 60 * , 2295 ( 1988 ) . see supporting information .
there have been many generalizations of conway s `` game of life '' ( gol ) since its invention in 1970 @xcite . almost all attributes of the gol can be altered : the number of states , the grid , the number of neighbors , the rules . one feature of the original gol is the glider , a stable structure that moves diagonally on the underlying square grid . there are also `` spaceships '' , similar structures that move horizontally or vertically . attempts to construct gliders ( as we will call all such structures in the following ) , that move neither diagonally nor straight , have led to huge man - made constructions in the original gol . an other possibility to achieve this has been investigated by evans @xcite , namely the enlargement of the neighborhood . it has been called `` larger than life '' ( ltl ) . instead of 8 neighbors the neighborhood is now best described by a radius @xmath0 , and a cell having @xmath1 neighbors . the rules can be arbitrarily complex , but for the start it is sensible to consider only such rules that can be described by two intervals . they are called `` birth '' and `` death '' intervals and are determined by two values each . these values can be given explicitly as the number of neighbors or by a filling , a real number between 0 and 1 . in the first case , the radius has to be given , too , in the last case , this can be omitted . the natural extension of evans model is to let the radius of the neighborhood tend to infinity and call this the continuum limit . the cell itself becomes an infinitesimal point in this case . this has been done by pivato @xcite and investigated mathematically . he has called this model `` reallife '' and has given a set of `` still lives '' , structures that do not evolve with time . we take a slightly different approach and let the cell not be infinitesimal but of a finite size . let the form of the cell be a circle ( disk ) in the following , although it could be any other closed set . then , the `` dead or alive '' state of the cell is not determined by the function value at a point @xmath2 , but by the filling of the circle around that point . similarly , the filling of the neighborhood is considered . let the neighborhood be ring shaped , then with @xmath3 our state function at time @xmath4 we can determine the filling of the cell or `` inner filling '' @xmath5 by the integral @xmath6 and the neighborhood or `` outer filling '' @xmath7 by the integral @xmath8 where @xmath9 and @xmath10 are normalization factors such that the filling is between 0 and 1 . because the function values of @xmath11 lie also between 0 and 1 the factors simply consist of the respective areas of disk and ring . the radius of the disk or `` inner radius '' is given by @xmath12 which is also the inner radius of the ring . the outer radius of the ring is given by @xmath13 . in the original gol the state of a cell for the next time - step is determined by two numbers : the live - state of the cell itself , which is 0 or 1 , and the number of live neighbors , which can be between 0 and 8 . one could model all general rules possible by a @xmath14 matrix containing the new states for the respective combinations . it could be called the transition matrix . now in our case this translates to the new state of the point @xmath2 being determined by the two numbers @xmath5 and @xmath7 . the new state is given by a function @xmath15 . let us call it the transition function . it is defined on the interval @xmath16 \times [ 0,1]$ ] and has values in the range @xmath16 $ ] . to resemble the corresponding situation in gol , typically @xmath17 is chosen ( the diameter of the neighborhood is 3 cells wide ) . as simple as the theoretical model is , it is not immediately obvious , how to implement it on a computer , as a computer can not handle infinitesimal values , continuous domains , etc . but it can handle real numbers in the form of floating point math , and as it turns out , this is sufficient . we also can model the continuous domain by a square grid , the ideal data structure for computation . so we will be able to implement our function @xmath3 as a @xmath18 array . when implementing the circularly shaped integrals we run into a problem . pixelated circles typically have jagged rims . so either we let the radius of the circle be so huge , that the pixelation due to our underlying square grid is negligible . then the computation time will be enormous . or we use another solution used in many similar situations : anti - aliasing . consider for example the integration of the inner region . for the cell @xmath2 function values are taken at locations @xmath19 . let us define @xmath20 . with an anti - aliasing zone around the rim of width @xmath21 we take the function value as it is , when @xmath22 . in the case when @xmath23 we take 0 . in between we multiply the function value by @xmath24 . similarly for the inner rim of the ring and the outer rim . in this way the information on how far the nearest grid point is away from the true circle , is retained . typically , @xmath25 is chosen . we also have to construct the transition function @xmath15 explicitly . luckily we can restrict ourselves like ltl , for the beginning , to four parameters : the boundaries of the birth and death intervals . to make things smooth and to stay in the spirit of the above described anti - aliasing we use smooth step functions instead of hard steps . we call them sigmoid functions to emphasize this smoothness . for example we could define @xmath26 @xmath27 @xmath28 then we can define the transition function as @xmath29 where birth and death intervals are given by @xmath30 $ ] and @xmath31 $ ] respectively . the width of the step is given by @xmath32 . as we have two different types of steps we have an @xmath33 and an @xmath34 . note that neither the anti - aliasing nor the smoothness of the transition function are necessary for the computer simulation to work . they just make things smoother and allow one to choose smaller radii for neighborhood and inner region and so achieve faster computation times for the time - steps . so far we have made everything smooth and continuous but one thing : the time - steps are still discrete . at time @xmath4 the function @xmath15 is calculated for every @xmath2 and this gives the new value @xmath35 at time @xmath36 . if we think of the application of @xmath15 as a nonlinear operator @xmath37 we can write @xmath38 \ , f(\vec x , t)\ ] ] to give us the ability to obtain arbitrarily small time steps , we introduce an infinitesimal time @xmath39 and reinterpret the transition function as a rate of change of the function @xmath3 instead of the new function value . then we can write @xmath40 \ , f(\vec x , t)\ ] ] where we have defined a new @xmath15 , that has values in the range @xmath41 $ ] instead of @xmath16 $ ] . if the transition function in the discrete time - stepping scheme was @xmath42 then the smooth one is @xmath43 . the formula above is also the most trivial integration scheme for the integro - differential equation @xmath44 \ , f(\vec x , t)\ ] ] this equation however leads to a different form of life . the same generic gliders can not be found at the same birth / death values as in the version with discrete time - stepping , but it also leads to gliders , oscillating and stable structures . we have described a model to generalize conway s `` game of life '' to a continuous range of values and a continuous domain . the @xmath14 transition matrix of the gol has been generalized to the @xmath15 transition function . the 8 pixel neighborhood and 1 pixel cell of gol have been generalized to a ring shaped neighborhood and a disk shaped cell . the rule set has been generalized to four real numbers : the boundaries of the birth and death intervals . the last remaining discrete attribute was the time - stepping . we proposed a method for continuous time - stepping which reinterprets the transition function as the velocity of change . the technique with two radii has been used in other contexts @xcite , but no gliders were described . there has also been a computer implementation of a continuous version of gol , but without the inner radius technique , and no gliders were found at that time @xcite . the goal of finding a glider that can move in arbitrary directions has been achieved . of the original gol it resembles both the glider and the spaceship at the same time . it also resembles similar structures found in ltl . so we think we have found the generic , generalized glider , and call it the `` smooth glider '' .	we present what we argue is the generic generalization of conway s `` game of life '' to a continuous domain . we describe the theoretical model and the explicit implementation on a computer .
the exact solution of dirac s equation for an electron moving in the field of a plane electromagnetic wave was found by d. m. volkov in 1935 @xcite . the electron wave function , compared to the field - free case , changes due to a modification of its spinor structure and the appearance of an additional phase factor . the electron momentum changes to an effective quasi - momentum , and the electron mass becomes an effective `` dressed '' mass . these modifications depend on the dimensionless variable @xmath3 related to the amplitude of the electromagnetic four - potential @xmath4 @xcite @xmath5 where @xmath6 is the absolute value of electron charge ( @xmath7 with @xmath8 ) and @xmath9 is the electron mass . ( for a manifestly gauge invariant formulation cf . @xcite . ) decades later , volkov s solution was applied to compton scattering @xcite and electron - positron pair production @xcite in strong electromagnetic fields . a consistent systematic analysis of these electromagnetic and further weak processes , such as pion and muon decays , @xmath10 emission by an electron in an external field etc.was performed by nikishov and ritus and coworkers in a series of papers @xcite and summarized in the review @xcite . later , some aspects of weak interaction , in particular neutrino emission , by electrons in a strong electromagnetic field were considered in refs . . the twofold extension of qed for strong electromagnetic fields was discussed in the recent paper @xcite . the main result of these previous studies is the conclusion that the quantum processes are modified significantly in strong electromagnetic fields . for instance , an electron can absorb or emit simultaneously a certain number of field photons , depending on the initial conditions of the considered process . this fact , together with the modifications of above mentioned electron properties , results in strong non - linear and non - perturbative effects which can not be described within the usual perturbative quantum electrodynamics ( pqed ) . consider , for example , the emission of a photon with four - momentum @xmath11 by an electron moving in a electromagnetic wave field . the process depends on the invariant variable @xmath12 @xcite , which varies in the range of @xmath13 for the absorption of @xmath14 photons with four - momenta @xmath15 by the electron with four - momenta @xmath16 and @xmath17 prior and after the emission process . one can see that ( i ) the kinematical limit @xmath18 ( phase space ) increases with the number of absorbed photons ( `` cumulative effect '' ) and ( ii ) decreases with increasing field intensity @xmath3 because of the electron mass modification . on the other hand , the contribution of higher harmonics also increases with @xmath3 , where , following @xcite , we use the notion `` harmonics '' for processes with different @xmath14 s . since @xmath3 plays an important role , it seems to be useful to recall the relation between @xmath3 and the electromagnetic ( laser ) field intensity @xmath19 , where the electromagnetic field is considered as a classical background field . for the case of a monochromatic circularly polarized plane wave with four - potential @xmath20 , where @xmath21 , and @xmath22 , i.e. @xmath23 , the average value of @xmath24 is equal to @xmath25 , meaning @xmath26 . on the other hand , the field intensity may be expressed through the electric ( @xmath27 ) and magnetic ( @xmath28 ) field strengths by @xmath29 . taking into account @xmath30 , one gets an expression for the average intensity @xmath19 in terms of the amplitude @xmath31 , @xmath32 , which leads to @xmath33 , where @xmath34 is the wave length of the electromagnetic field @xmath4 . the dependence of @xmath3 on the electromagnetic field intensity @xmath19 for different wavelengths @xmath35 is exhibited in fig . [ fig:1 ] . the wavelength @xmath36 m ( or @xmath37 ev ) corresponds to the widely used titanium - sapphire laser oscillator ( cf . refs . the short wavelength @xmath38 nm ( or @xmath39 kev ) corresponds to the soft x - ray ( sxr ) free electron laser at slac @xcite . the long wavelength @xmath40 m ( or @xmath41 ev ) may be obtained at the free electron laser for infrared experiments ( felix ) @xcite . one can see that @xmath3 varies within a fairly large range , depending on the field intensity and wavelength . in the low - frequency limit , @xmath42 , the intensity parameter becomes large , i.e. @xmath43 at fixed intensity @xmath19 or @xmath44 . this limit was considered in some detail by nikishov and ritus @xcite who pointed out that the invariant variable @xmath45 remains finite and the total probabilities of most of the considered processes depend only on @xmath46 @xcite . here , @xmath47 is the electromagnetic field tensor . such a case of simultaneous limits of @xmath48 and @xmath42 at finite @xmath19 corresponds to the situation of an electron interacting with a constant ( crossed ) electromagnetic field . note that two asymptotic regions of the external field were considered in most of the above quoted papers . one corresponds to the weak - field limit @xmath49 . in this case , only a limited number of harmonics @xmath50 contributes . the opposite case of large intensity @xmath51 with @xmath52 allows for two asymptotic limits : @xmath53 and @xmath54 . of course , such an analysis of limiting cases is interesting and important by its own . however , the rapidly evolving laser technology @xcite can provide conditions where the limit of @xmath55 is achieved at finite @xmath56 , as well as @xmath57 as can be inferred from fig . [ fig:1 ] and by numerical evaluation of eq . ( [ i10 ] ) . therefore , it seems relevant to consider the probabilities of quantum processes without the restrictions imposed in @xcite . the goal of present work is accordingly an analysis of neutrino pair emission off an electron moving in a strong external electromagnetic ( laser ) wave field in a wide region of @xmath58 and @xmath46 . our paper is organized as follows . in sect . ii , we consider the neutrino pair emission . a scheme is presented to overcome convergence problems in the expansion in terms of harmonics . the employed method is similar to the one - photon emission process which is outlined in appendix a to expose these similarities and the important differences . the perturbative neutrino pair emission is recapitulated in appendix b. our conclusions can be found in sect . similar to the emission of a photon by an electron moving in an electromagnetic external ( background ) field ( see appendix [ a.1 ] ) one can evaluate the emission of a neutrino pair @xmath59 of species @xmath60 by the @xmath61 matrix element @xmath62 corresponding to the diagram in fig . [ fig:8a ] . @xmath63 is the invariant mass of the @xmath64 pair : @xmath65 with @xmath66 and @xmath67 as four - momenta of neutrino and antineutrino . ( we employ units with @xmath68 . ) the volkov solution @xmath69 in ( [ sss ] ) ( cf . ( [ ii10 ] ) in appendix [ a.1 ] ) describes the in - state of the electron in an external wave field , while @xmath70 refers to the out - state of the electron also accounting for the external field . the neutral neutrino current @xmath71 couples directly to the electron current @xmath72 with a strength given by fermi s constant @xmath73 gev@xmath74 . the expression ( [ sss ] ) holds in the local limit where all momenta involved in the process are much smaller than the masses of the intermediate vector bosons @xmath75 and @xmath76 . then , one can obtain `` universal '' interactions described by the effective low - energy lagrangian @xcite for the direct current - current interaction @xmath77 \left [ \bar u_e \gamma^\alpha ( c_v^{(i ) } - c_a^{(i)}\gamma_5 ) u_e \right]~ , \label{iii3}\end{aligned}\ ] ] with @xmath78 the average value of @xmath79 is taken from ref . @xcite . in the weak - field approximation , where the interaction with the external wave field is mediated by one photon , the diagram in fig . [ fig:8a ] would be resolved by diagrams exhibited in fig . [ fig:8b ] corresponding to the process @xmath2 which can be dealt with perturbatively ( see appendix [ b ] ) . the effective weak vertices for the @xmath80 interaction , in turn , are resolved in the tree - level approximation within the standard model as exhibited in fig . [ fig:8c ] . the advantage of the diagram in fig . [ fig:8a ] and eq . ( [ sss ] ) is that multi - photon effects are included which become important at high intensities where the external wave behaves more and more as a classical field . as in the case of the emission of one photon ( see appendix [ a ] ) , the differential probability of neutrino pair emission can be represented by the infinite sum of the partial contributions , @xmath81 with @xmath82 where @xmath83 is the azimuthal angle of the outgoing electron and @xmath84 analog to the case of one - photon emission ( see appendix [ a.1 ] ) the label @xmath14 refers to the number of photons absorbed from the external field . for the sake of simplicity , we skip henceforth the index @xmath85 . the electron tensor @xmath86 \label{iii12}\end{aligned}\ ] ] incorporates the electromagnetic wave field via @xmath87~. \nonumber\end{aligned}\ ] ] here , @xmath88 is the unit vector of @xmath89 , and the functions @xmath90 are related to the bessel functions ( cf . refs . @xcite ) : @xmath91 the argument of these bessel functions is @xmath92 where @xmath18 is the kinematical limit of the invariant variable @xmath93 defined by @xmath94 which determines the upper limit of @xmath95 at given @xmath93 by @xmath96 the functions @xmath97 in eq . ( [ iii11 ] ) do not depend on @xmath83 ; the explicit expression reads @xmath98 where @xmath99 is the relative phase ( polarization ) of the amplitudes @xmath100 and @xmath101 , and @xmath102 . our expression coincides with the result of ref . @xcite in a different notation . in fig . [ fig:10 ] we exhibit the differential emission probabilities summed over all neutrino types for a head - on collision of @xmath103 mev electrons with a laser beam characterized by @xmath104 m ( or @xmath105mev ) . we fix the invariant mass of the outgoing neutrino pair by @xmath106 ev . ( the numerical results for the first two harmonics coincide for @xmath107 with the prediction @xcite based on the asymptotic decomposition of @xmath108 for @xmath109 in the limit @xmath110 . ) the prediction of perturbative qed for the reaction @xmath111 ( see fig . [ fig:8b ] and eq . ( [ iii81 ] ) in appendix [ b ] ) is shown by the dashed curves . one can see that for small @xmath3 the result for @xmath112 coincides practically with that of pqed . similar to the case of non - linear compton scattering ( see appendix [ a.1 ] ) , even in this case the contribution of higher harmonics increases the phase space , i.e. the kinematic limit , and modifies the total probability . when @xmath3 increases , the difference between the non - perturbative calculations and pqed is rather large , even for @xmath112 , see middle and right panels of fig . [ fig:10 ] . in fig . [ fig:11 ] we show the total probability normalized as @xmath113 , where @xmath114 is the initial electron density divided by the electron mass in the electron rest frame , @xmath115 for the sum of all types of neutrinos as a function of @xmath3 for the first 25 harmonics . for small @xmath3 , @xmath116 , along with the predominant contribution of the first two harmonics the contribution of the higher harmonics is significant . when @xmath3 increases , e.g. for @xmath117 , the contribution of higher harmonics exceeds the contribution of lowest harmonic by orders of magnitude . the qualitative difference in the relative contribution of higher harmonics to the total probability of the emission of a photon ( cf . [ fig:4 ] in appendix a 1 ) and neutrino pairs is explained by the employed four - fermion structure of the weak - interaction @xmath80 vertex and the three - particle phase space . therefore , the problem of convergence for the total probability in case of neutrino emission becomes severe and deserves special consideration . figure [ fig:12 ] illustrates the convergence problem of the total @xmath118 emission probability with increasing @xmath3 for the case of @xmath119 ev and @xmath120 mev , where the number of included harmonics goes up to @xmath121 . one can see some saturation at @xmath122 . however , for @xmath123 the difference of probabilities with @xmath124 and @xmath121 is more than two orders of magnitude . the problem of convergence may be solved by the same method as proved useful for the one - photon emission ( see appendix [ a.1 ] ) using the transformations in eqs . ( [ ritustransformations_a ] ) and ( [ ritustransformations_b ] ) . the main difference to the one - photon emission ( cf . ( [ ritustransformations2 ] ) ) is a modification of the argument of the airy functions , which emerge from the above bessel functions , as @xmath125 and an additional integration over the invariant mass @xmath126 . also , the interference term in eq . ( [ iii15 ] ) proportional to the rapidly oscillating combination @xmath127 is negligible and can be omitted . the final expression for the probability @xmath128 of @xmath129 emission in the limit of @xmath130 reads @xmath131 where @xmath132 . the emission probability @xmath133 for all three types of neutrinos summed up is shown in fig . [ fig:13 ] for a wide region of @xmath46 and @xmath58 . the stars and crosses correspond to the asymptotic values of @xmath134 for @xmath53 and @xmath135 , respectively @xcite : @xmath136 now , the dependence @xmath133 on @xmath58 is weaker compared with the photon emission ( cf . [ fig:6 ] ) , mainly because of the additional integration over @xmath95 . note that the asymptotic estimates by eq . ( [ iii19 ] ) do not match in the intermediate region @xmath137 . in fig . [ fig:14 ] we show the total probability @xmath138 of neutrino pair emission for all neutrinos calculated for wide initial experimental conditions ranging from @xmath139 ev up to @xmath140 kev as a function of @xmath3 , where @xmath3 varies from 0.1 up to @xmath141 . this interval covers possible experimental conditions illustrated in fig . [ fig:1 ] . at @xmath142 , @xmath138 is evaluated using the asymptotic expression of eq . ( [ w_asympt ] ) . for @xmath143 , the probability might be evaluated as a sum of partial harmonics , in our case up to @xmath121 . one can see that , at large @xmath3 , the difference between the probability calculated as a sum of a large but finite number of harmonics and its asymptotic value which includes an infinite number of harmonics is several orders of magnitude . finally , we would like to note the following . in average , the difference of neutrino pair emission probabilities for @xmath144 ev and 0.5 kev is about 17 orders of magnitude ( cf . left and right panels in fig . [ fig:14 ] ) . this is much greater than the corresponding difference in one - photon emission shown in the left and right panels of fig . [ fig:7 ] , where the corresponding difference is about 5 orders of magnitude . the difference between the two processes is explained by the different @xmath46 dependence of @xmath138 , shown in figs . [ fig:13 ] and [ fig:6 ] , respectively . the average values of @xmath46 for @xmath144 ev and 0.5 kev are @xmath145 and @xmath146 , respectively . therefore , @xmath138 increases with @xmath56 . the sharp increase of the probability with @xmath46 in case of neutrino pair emission is a consequence of the strong energy dependence of the total probability which can be traced back to the four - fermion structure of the @xmath147 matrix element . it seems to be interesting to analyze the asymmetry of the emission of electron and muon plus tau neutrino pairs which may be defined as @xmath150 corresponding predictions for two regions of @xmath3 are shown in fig . [ fig:15 ] for different values of the wave field photon energies . in the region of @xmath151 , the probabilities are calculated as a finite sum of partial harmonics with @xmath152 ( cf . ( [ iii11_a ] ) ) . the prediction for the @xmath153 reaction ( see appendix [ b ] ) is shown by the dot - dashed curve labelled by `` free '' . predictions for @xmath139 and 1.55 ev are virtually identical . for @xmath110 , all predictions are close to the results for the @xmath153 reaction : @xmath154 . all asymmetries are small in absolute value and negative , meaning a slight dominance of the emission of @xmath155 neutrinos compared to electron neutrinos . at large @xmath3 , the probabilities are calculated as an infinite number of partial harmonics ( cf . ( [ w_asympt ] ) ) . for @xmath139 and 1.55 ev the range of variation of @xmath46 is @xmath156 and @xmath157 , respectively . that is , the variable @xmath46 is small and therefore , the asymmetries at these two energies are close to each other and close to the value @xmath158 which results from the asymptotic expression eq . ( [ iii19 ] ) . for @xmath159 kev and @xmath160 the asymmetry is positive and increases with @xmath3 . note that , within the considered range of @xmath3 , eq . ( [ iii19 ] ) does not apply and @xmath161 is smaller than its asymptotic value @xmath162 which would be found from eq . ( [ iii19 ] ) . we predict a distinct dominance of emission of electron neutrino pairs compared to the sum of muon and tau neutrino pairs in the kev - range of the photon energy and at large @xmath3 . the asymmetry increases as @xmath163 in the interval of @xmath164 . finally , we note that our non - perturbative calculation of @xmath161 is strongly different from the prediction of pqed ( cf . dot - dashed curve in the left panel of fig . [ fig:15 ] ) in all considered intervals of @xmath56 and @xmath3 unless @xmath165 . in summary we have considered @xmath166 ( @xmath60 ) emission off an electron in a strong electromagnetic wave field in a wide range of energy of the wave field photon energy @xmath56 and reduced field intensity @xmath3 . similarly to previous work we expressed the emission probability as a sum of partial harmonics , where each harmonic describes the interaction of an electron in - state with @xmath14 field photons coherently . for the first time , we made a summation over harmonics up to a quite large number of harmonics and found that , at large values of field intensity , @xmath123 , which can be achieved in current and future laser facilities , the convergence is rather weak . therefore , we have elaborated a method allowing for a complete summation of all partial harmonics . the method is tested for one - photon emission , i.e. non - linear compton scattering . using this new approach we calculated neutrino pair emission in a region of @xmath56 and @xmath3 which can be reached experimentally in near future . we have shown that , at large @xmath3 , the difference between the finite and complete sums of partial harmonics reaches a few orders of magnitudes . in case of neutrino pair emission we also analyzed the non - trivial asymmetry between the production of electron and @xmath1 neutrino pairs . we found that the asymmetry depends strongly on initial conditions expressed via @xmath56 and @xmath3 . at low @xmath56 , the asymmetry is negative corresponding to the dominance of emission of the @xmath1 neutrinos , while at large @xmath56 the asymmetry changes the sign indicating the dominance of electron neutrino pairs . finally , we note that all calculations ( and conclusions ) have been done for the sake of simplicity for the initial electron energy @xmath120 mev , which corresponds to the energy of the superconducting electron accelerator elbe in fz dresden - rossendorf @xcite . one of the key variable in the considered processes , @xmath167 , where @xmath168 and @xmath169 are the photon and electron four - momenta , respectively , directly depends on @xmath170 . therefore , the values of emission probabilities would also depend on @xmath170 . the corresponding evaluation of this dependence as well as the analysis of other processes will be done in forthcoming papers . a further step towards realistic estimates is related to the inclusion of a temporal shape of the external laser wave field , as considered e.g. in @xcite . the authors appreciate t. e. cowan and d. seipt for fruitful discussions . the methods employed in section ii are guided by the one - photon emission process off a volkov electron , i.e. the non - linear compton effect . to expose the similarities and differences we recall the essential steps and clarify further the notation . let us consider the emission of a photon off an electron moving in a plane electromagnetic wave described in the introduction . this process is described by the @xmath61 matrix element @xcite @xmath171 where @xmath172 is the energy of the emitted photon with four - momentum @xmath11 and @xmath173 is the electron wave function in the initial ( final ) state given by volkov s solution of the dirac equation @xmath174 \,{\exp}\left [ -i\int\limits_0^{k \cdot x } \frac{e(p\cdot a)}{(k \cdot p ) } \,d\phi'\right ] \frac{u_p}{\sqrt{2q_0}}{\rm e}^{-iq \cdot x}~ , \label{ii10}\end{aligned}\ ] ] where @xmath175 denote dirac matrices and @xmath176 is the quasi - momentum @xmath177 of the dressed electron with effective mass @xmath178 note that eq . ( [ ii9 ] ) employs the furry picture : the field @xmath4 is considered as external classical ( background ) field , and the emission of a photon with wave four - momentum @xmath11 and polarization @xmath179 is described in lowest order of perturbation theory , see diagram ( a ) in fig . [ fig:2 ] . the dependence of the potential @xmath180 on @xmath181 in eq . ( [ ii10 ] ) results in the following structure of the @xmath61 matrix element @xmath182 where a fourier decomposition is used : @xmath183 thus , the amplitude is represented as a sum of an infinite number of terms which are referred to as partial harmonics . each harmonic can be attributed to the absorption ( emission ) of @xmath14 photons from ( into ) the external field @xmath180 characterized by the wave four - vector @xmath168 . for the photon emission off an electron the corresponding conservation law reads @xmath184 , cf . ( [ ii13 ] ) . in case of on - shell photons , @xmath185 and @xmath186 hold and therefore @xmath187 . correspondingly , the differential probability is the infinite sum of partial contributions : @xmath188 where the partial harmonics @xmath189 are expressed through bessel functions @xmath190 of the first kind @xcite @xmath191 with @xmath192 as an example , we exhibit in fig . [ fig:3 ] the differential probabilities for photon emission by @xmath103 mev electrons colliding head - on with a @xmath193 m ( @xmath105mev ) laser beam . this process can be studied experimentally , e.g. , at superconducting electron accelerator elbe in conjunction with the 150 tw laser draco in fz dresden - rossendorf @xcite . the solid curves correspond to the partial contributions which come from the coherent absorption of @xmath14 photons with @xmath194 . the prediction of perturbative qed compton scattering , described by the klein - nishina formula ( see appendix [ a.2 ] ) , is shown by the dashed curves . one can see that for small @xmath3 , the results for @xmath112 coincide practically with the result of pqed , cf . also @xcite . however , even in this case the contribution of higher harmonics increases the phase space and modifies the total probability . when @xmath3 increases , the difference between non - perturbative calculations and pqed is rather large , even for @xmath112 . in general , the modification of the kinematical limit follows eq . ( [ ii17 ] ) : @xmath18 increases with @xmath14 and decreases with @xmath3 . in fig . [ fig:4 ] we show the total probability of one - photon emission as a function of @xmath3 for the first 25 harmonics . for convenience , we show reduced probability @xmath113 , where @xmath114 is the density of initial electrons divided by the electron mass in the electron rest frame . on can see that for small @xmath3 , @xmath49 , the total probability is saturated by the contribution of the first two harmonics . however , at large @xmath3 , @xmath195 , the contribution of higher harmonics becomes large and the convergence of the total probability as a function of the number of harmonics @xmath14 is weak . figure [ fig:5 ] illustrates the convergence of the emission probability with increasing @xmath3 for the case of @xmath119 ev and @xmath120 mev , where the number of harmonics increases up to @xmath121 . one can see some saturation as long as @xmath196 . when @xmath3 increases , @xmath197 must further increase , in principle , as @xmath198 @xcite . for the calculation of the total probability for large but finite @xmath3 we use the method of ref . @xcite based on utilizing the properties of the bessel functions @xmath199 at large values of @xmath14 and @xmath200 and replacing the sum of @xmath14 contributions by an integral over @xmath14 . ultimately , this procedure reduces to the following transformations @xmath201 where @xmath202 is the airy function , and @xmath203 the corresponding probability is expressed in the following form : @xmath204\frac{1}{t } \left(y\phi^2(y ) - { \phi'}^2(y)\right)\right]~.\nonumber\\ & & \label{ii18}\end{aligned}\ ] ] contrary to ref . @xcite and related papers we do not put @xmath48 in the above integral , which allows to calculate the probabilities at large but finite values of @xmath3 in a wide range of @xmath46 . in fig . [ fig:6 ] we show @xmath138 as a function of @xmath46 for different values of @xmath3 . the crosses and stars correspond to the asymptotic values of @xmath205 at @xmath135 and @xmath206 , respectively @xcite : @xmath207 in fig . [ fig:7 ] we show the total probability @xmath138 calculated for a wide region of possible experimental conditions ranging from a photon energy of @xmath139 ev up @xmath159 kev as a function of @xmath3 , for @xmath3 varying from 0.1 up to @xmath208 . this interval covers possible experimental conditions illustrated in fig . [ fig:1 ] . for @xmath123 , @xmath138 is evaluated using the asymptotic expression of eq . ( [ ii18 ] ) . for @xmath209 , the probability is calculated as a sum of partial harmonics , in our case up to @xmath121 . at the matching point @xmath210 , the difference between the two expressions is less than 5% . to prove the equivalence of the probabilities defined above to usual partial cross sections we recollect here also the perturbative treatment of the scattering of a photon off an electron ( compton scattering , see middle and right panels in fig . [ fig:2 ] ) . the cross section reads in standard notation @xmath211 where @xmath212 is the square of total energy and @xmath213 denotes the square of the momentum transfer . @xmath214 is the invariant amplitude which consists of direct and crossed terms , schematically depicted in fig . [ fig:2 ] ( b ) and ( c ) , respectively , @xmath215~ , \label{ii2}\end{aligned}\ ] ] where @xmath216 and @xmath217 denote the polarization vectors of incoming ( @xmath175)and outgoing ( @xmath218 ) photons , respectively , @xmath219 and @xmath220 are dirac spinors of incoming and outgoing electrons , normalized as @xmath221 , and @xmath222 is the transition operator @xmath223 in eq . ( [ ii1 ] ) averaging over the initial and summation over the final spin states are provided . the corresponding calculation is described in text books ( for example in @xcite ) . the result is the klein - nishina formula @xmath224 where @xmath225 is the `` classical '' electron radius and @xmath226 for the further analysis it is convenient to use the invariant variables @xmath12 and @xmath227 and to employ the probability @xmath228 of the @xmath229 reaction , instead of the cross section @xmath230 , @xmath231 where @xmath232 is the photon density ( or inverse volume per one photon ) , defined similarly to the dependance of @xmath3 on @xmath19 mentioned in the introduction , @xmath233 inspection of eqs . ( [ ii4 ] ) , ( [ ii7 ] ) and ( [ ii8 ] ) shows that compton scattering is independent of @xmath3 . variation of @xmath3 ( or field intensity @xmath19 ) changes the density of photons @xmath232 , and therefore only determines the overall normalization of @xmath228 . consider first neutrino pair production in the reaction @xmath234 . the differential cross section reads @xmath235 where @xmath236 is the solid angle of outgoing neutrino in the @xmath0 rest frame . the invariant amplitude @xmath237 in lowest order ( tree level ) of the glashow - salam - weinberg model is described by the feynman diagrams depicted in fig . 4 . accordingly , the electron neutrino pairs are produced via exchange of both the charged @xmath76 and neutral @xmath75 vector bosons , while the muon and tau neutrino pairs are produced only through the neutral boson exchange when assuming individual lepton number conservation . in the local limit , where all the momenta involved in the process are much smaller than the masses of the intermediate vector bosons , by making use of a fierz transformation for the charged boson exchange , one can obtain `` universal '' effective interactions depicted in fig . [ fig:8b ] as direct ( a ) and exchange ( b ) terms . this interaction is described by the effective lagrangian in eq . ( [ iii3 ] ) . then , the invariant amplitude is expressed as @xmath238 , \label{iii5}\end{aligned}\ ] ] where @xmath239 is defined by eq . ( [ iii6 ] ) , and the transition operator @xmath240 is defined by eq . ( [ ii3 ] ) with the substitution @xmath241 summation over spins of the neutrinos ( with the assumption @xmath242 ) and integration over the solid angle in eq . ( [ iii1 ] ) leads to @xmath243 = \frac{8}{3}(q_\alpha q_\beta -g_{\alpha\beta } q^2 ) , \label{iii8}\end{aligned}\ ] ] and one can rewrite eq . ( [ iii1 ] ) as @xmath244 . \label{iii82}\end{aligned}\ ] ] a. i. nikishov and v. i. ritus , sov . jetp * 19 * , 529 ( 1964 ) [ zh . eksp . teor . * 46 * , 776 ( 1964 ) ] . a. i. nikishov and v. i. ritus , sov . jetp * 19 * , 1191 ( 1964 ) [ zh . eksp . fiz . * 46 * , 1768 ( 1964 ) ] ; + v. i. ritus , sov . jetp * 29 * , 532 ( 1969 ) [ zh . eksp . fiz . * 56 * , 986 ( 1969 ) ] . n. b. narozhnyi , a. i. nikishov , and v. i. ritus , sov . jetp * 20 * , 622 ( 1965 ) [ zh . eksp . teor . fiz . * 47 * , 930 ( 1964 ) ] . g. a. mourou , t. tajima , and s. v. bulanov , rev . phys . * 78 * , 309 ( 2006 ) ; + y. i. salamin , s. x. hu , k. z. hatsagortsyan , and c. h. keitel , phys . rep . * 427 * , 41 ( 2006 ) ; + m. marklund and p. k. shukla , rev . mod . phys . * 78 * , 309 ( 2006 ) ; + g. gregori _ _ , high energy density phys . * 6 * , 166 ( 2010 ) .	the emission of @xmath0 pairs off electrons in a polarized ultra - intense electromagnetic ( e.g. laser ) wave field is analyzed . we elaborate on the significance of non - linear electrodynamics effects ( i.e. , multi - photon processes ) and the peculiarities of neutrino production . special attention is devoted to the convergence of the reaction probabilities as a function of the number of absorbed photons . expressions for large field intensities are provided . the asymmetry between the probabilities of electron and @xmath1 neutrino production depends on initial conditions such as energy of the wave field photons and the field intensity . these findings differ from the lowest order perturbative calculation of the reaction @xmath2 .
large transverse momentum phenomena in deep inelastic scattering reactions have been extensively studied by the h1 and zeus collaborations at hera . large-@xmath7 particle and jet spectra have been measured and compared to next - to - leading order ( nlo ) qcd calculations . large-@xmath7 photons have also been observed , in an inclusive way @xcite or in correlation with a hadronic jet @xcite . this latter reaction has been the subject of theoretical studies some ten years ago @xcite . the recent data from zeus @xcite lead us to extend these studies and to compare the complete nlo qcd results with the @xmath8-jet cross sections . in principle , prompt photon production in deep - inelastic scattering ( dis ) is a very simple process : it goes via the compton scattering of a virtual photon on a quark : @xmath9 and requires only the knowledge of the distribution function of a quark in the proton . including higher - order ( ho ) corrections considerably complicates the picture and new objects have to be introduced . for example , in the scattering @xmath10 , the @xmath11 pair may be produced quasi - collinearly to the virtual photon , one of the parton in the pair being then scattered at large @xmath12 : this configuration generates the virtual photon structure function ( resolved photon ) at lowest order associated to a large logarithm . it is then necessary to resum such large logarithms and introduce the all order photon structure function . furthermore , in the above process or in @xmath13 , the final photon may be produced collinearly to a final state quark or antiquark ( bremsstrahlung ) leading to a large logarithmic enhancement , thus generating the photon fragmentation fonction . thus one is lead to distinguish four types of processes , all contributing already at leading order ( lo ) : the direct - direct ( d - d ) one where both the virtual photon and the final real photon couple directly to the hard sub - process ; the resolved - direct ( r - d ) where the virtual photon couples to the hard subprocess through its hadronic ( perturbative or non perturbative ) components ; the direct - fragmented ( d - f ) and the resolved - fragmented ( r - f ) ones where the final photon appears as a fragment of a jet unlike in the previous two cases where it is isolated . at hera , all four processes corresponding to four topologies have essentially the same order of magnitude @xcite . however when dealing with isolated photon , the isolation criteria necessary to suppress the background from @xmath14 , considerably reduces the fragmentation components d - f and r - f . the above discussion on the four topologies is valid as long as we can define a virtual photon structure function resumming all the large logarithms @xmath15 @xcite where @xmath7 is a characteristic transverse momentum of the reaction ( for instance that of the observed photon in the proton virtual - photon center - of - mass frame ) and @xmath16 the initial photon virtuality . these terms appear in the calculation of ho corrections to the born direct cross sections . if @xmath7 is not large enough ( @xmath17 ) it is of course not useful to subtract these logarithms from the direct ho corrections in order to resum them in the virtual photon structure function . on the other hand for @xmath18 this approach is useful : indeed in this case the resolved cross sections have the same structure as a hadronic cross section involving two parton distributions convoluted with hard subprocesses . ho corrections are known , they are large and can be easily implemented . the natural frame to observe large-@xmath7 phenomena and to calculate the corresponding nlo cross section in deep - inelastic scattering ( dis ) is the proton virtual - photon center - of - mass system ( hadronic frame or cm@xmath0 ) . the large @xmath19 of the final photon provides the scale which can be compared with the photon virtuality ; a large ratio @xmath20 defines the kinematical range in which the photon structure function formalism is useful . such an approach , but without the introduction of the virtual photon structure function , can be found in @xcite . it contains detailed studies on the jet algorithms and the scale choice dependence of the cross sections . as the kinematical conditions are often specified in the laboratory frame and as a large-@xmath7 in the laboratory does not necessarily implies a large @xmath19 in the cm@xmath0 , a lower limit @xmath21 can also be imposed by the experiments . this condition will preserve the validity of a perturbative calculation and the possibility to define a virtual photon structure function . the production of jets and of forward @xmath22 has been measured with this convention by h1 @xcite and zeus @xcite . on the other hand , several experiments have also used the laboratory frame ( lab frame ) to present their results @xcite without imposing the requirement @xmath23 . as we shall see , the approach involving the definition of the resolved cross section is not always under control , and we have to content ourselves with the calculations of unsubtracted direct contribution . thus we loose the possibility to supplement them with ho corrections . in this paper we consider dis reactions in which an isolated photon and a jet are observed in the final state , extending the approach used in the inclusive case @xcite with only a photon observed in the final state . we discuss both cases , when the transverse momenta are defined in the cm@xmath0 or in the lab frames . this study is triggered by recent detailed zeus results @xcite . unfortunately no @xmath19 constraint has been introduced by this collaboration , thus forbidding , in some kinematical domains , direct comparisons with complete nlo predictions . the comparison with inclusive isolated cross section done in our previous paper was favored by the h1 kinematics @xcite having a large domain where the condition @xmath24 with @xmath25 was verified . the situation is less favorable with the zeus kinematics having a larger range in @xmath16 . the observation of a photon and a jet in the laboratory does not necessarily imply a large invariant mass squared @xmath26 of this system ( when the photon and the jet are almost parallel ) . therefore the addition of a jet is not sufficient to prevent configuration with @xmath27 . this fact leads us to introduce a cut - off @xmath28 in our calculations and to study the stability of our results when @xmath28 goes to zero . if , in some kinematical domains , the results are not sensitive to @xmath28 we will compare them with zeus data . the plan of this paper is the following . the next section is devoted to the calculation of the gamma - jet cross section in the cm@xmath0 . by this we mean that the isolation algorithm for the photon and jet algorithm are defined in this frame . the jet algorithm also contains conditions such as `` jet of highest @xmath7 '' in the cm@xmath0 . as we fix the kinematical variable @xmath29 and @xmath30 in the laboratory we require the condition @xmath31 . we discuss in detail the introduction of the resolved component . the four contributions ( d - d , d - f , r - d and r - f ) are calculated and their respective importance is discussed . then , in section 3 , we discuss calculations performed in the laboratory frame . by this we mean that the isolation algorithm and jet algorithm are now defined in the laboratory where the kinematical boundaries are also fixed . we furthermore study the differences between the standard cone algorithm and the democratic one for the photon isolation and for the jet definition . we have to introduce a cut - off @xmath28 in order to avoid possible instability of the cross section . finally section 4 is devoted to the comparison with the recent zeus data @xcite . in this section , devoted to calculations in the cm@xmath0 , we study the four contributions to the @xmath8-jet cross section discussed in the introduction . such a frame looks like the laboratory frame of a hadronic collider ; the colliding particles have collinear trajectories and the large transverse momentum @xmath19 of the observed final particle fixes the large scale appearing in the distribution functions and in the strong coupling constant . however to be close to the experimental conditions we impose constraints on the observed kinematical variables , transverse momenta and rapidity , in the laboratory frame . the particle momenta are transformed to the cm@xmath0 frame in which we define the isolation algorithm and the jet algorithm . as discussed in the introduction we impose a cut on the photon momentum @xmath31 in order to avoid instabilities of the cross section and to guarantee the validity of the perturbative regime . in this section we use the cone algorithm @xcite for the isolation and the longitudinal @xmath32-algorithm @xcite to define a jet . when two jets are present , we observe the jet of higher @xmath33 . for the photon the radius of the isolation cone is @xmath34 and we require a ratio @xmath35 where @xmath36 is the hadronic energy contained in the isolation cone . for the jet the radius @xmath37 of the @xmath32-algorithm is set equal to @xmath38 . another algorithm , the so - called democratic algorithm will be discussed in section 3 . we use the cteq6 m proton structure @xcite and the virtual photon structure function presented in ref . @xcite and used in refs . @xcite and @xcite . the fragmentation function is that of the bfg collaboration ( set ii ) @xcite . details on the nlo calculation can be found in ref . we work in the @xmath39 scheme for factorization and renormalization with @xmath40 mev and @xmath41 . the factorization and the renormalization scales are taken equal to @xmath42 . the kinematical constraints are defined in the laboratory . we adopt those of ref . the colliding electron and proton have energies @xmath43 gev and @xmath44 gev corresponding to the center - of - mass energy @xmath45 gev . the photon momentum has to lie within the ranges @xmath46 gev and @xmath47 . the jet momentum is required to have 2.5 gev @xmath48 35 gev and @xmath49 . constraints on the final electron are : @xmath50 gev , @xmath51 ( the @xmath52 axis is pointing toward the proton direction ) . finally the photon virtuality is @xmath53 gev@xmath4 . the numerical calculations are carried out using the adaptative monte carlo code basis @xcite . in the d - d contribution the real and virtual photons are directly coupled to the hard subprocess and the born term , @xmath54 , is particularly simple : it consists of the compton scattering diagrams . the cross section has a @xmath55 behavior when the photon is emitted close to the forward direction ( collinear to the initial quark ) and a @xmath56 behavior when it is emitted collinear to the final quark . in the cm@xmath0 the divergent configurations , in @xmath57 or @xmath58 , are forbidden by the requirement @xmath59 . real ho corrections to the born term are given by amplitudes involving an extra gluon coupled to the quark line . when this gluon is collinear to the initial quark the cross section is divergent . this divergence is subtracted and absorbed in the quark distribution . when this gluon is emitted at large @xmath19 , the initial virtual photon can produce a collinear quark - antiquark pair . this is the origin of the virtual photon structure function . after integration on the final quark phase space one gets the expression @xmath60 where the symbol @xmath61 indicates a convolution between the born cross section and the antiquark distribution in the virtual photon - dependent factors in the logarithm argument are not indicated . they are discussed in ref . @xmath62 is the dglap branching function @xmath63 $ ] . in the large-@xmath64 regime , @xmath65 , we can neglect @xmath16 in the born cross section . in order to resum the structure function we subtract @xmath66 from the d - d ho term thus defining the ho@xmath67 corrections . the subtracted term builds , through the inhomogeneous dglap evolution equation , the resummed structure function @xmath68 @xcite , the basis of the resolved contributions . we choose @xmath69 so that the structure function vanishes when @xmath70 , and is proportional to @xmath71 for @xmath65 . some numerical results are displayed in fig . [ fig : ygam - dir - cm ] for the cross section @xmath72 with the cut @xmath73 gev . the kinematics has been term subtracted and the cut @xmath74 is implemented ( see text ) . _ _ ] given at the beginning of section 2 ; @xmath75 is the rapidity in the laboratory . as expected the cross section is largest in the forward direction due to the dominance of the @xmath76 term in the scattering amplitude . we notice that the ho@xmath77 corrections are negative , as already observed in the inclusive case @xcite . on the contrary , the unsubtracted ho corrections are close to zero , the difference ho - ho@xmath77 being the contribution of the lowest order resolved term that will be discussed in the next subsection . in fact , the nlo unsubtracted result can not be distinguished from the d - d lo curve in fig . [ fig : ygam - dir - cm ] . the results in that figure strongly depend on the value of the cut - off @xmath28 . this point is illustrated in table 1 where the total @xmath39 cross section is displayed . we consider two domains in @xmath16 . we see that for low value of @xmath16 the nlo cross section is stable with respect to the @xmath28 value while it is not the case when considering the whole @xmath16 range . in the former case , the lorentz boost between the laboratory and the cm@xmath0 frames , @xmath78 with @xmath79 the usual dis variable ( scaled virtual - photon energy in the laboratory ) , is small and @xmath80 remains relatively large , and close to its @xmath81 value in the laboratory , so that the cross section is insensitive to the cut - off . on the contrary for large values of @xmath16 , @xmath80 may reach small values where perturbation theory is not reliable ( very large ho corrections ) and where the theoretical predictions become sensitive to the cut - off . the errors associated to the nlo cross sections are the values resulting from the monte - carlo integration procedure ( for the born cross sections the errors are at the per - mil level ) . .__variation of the d - d @xmath8-jet cross section in pbarns with @xmath28 in two @xmath16 ranges . the isolation and jet criteria are implemented in the cm@xmath0 . _ _ [ cols="^,^,^,^,^ " , ] however we note that this dependence is weaker when @xmath82 gev@xmath4 . with this kinematical requirement we obtain a more stable cross section . the r - d contribution at nlo ( long - dash - dotted line ) to the cross section @xmath83 is displayed in fig . [ fig : ygam - res - lab ] for @xmath73 gev : it has the same features as in fig . [ fig : ygam - res - cm * ] , namely very slightly decreasing with @xmath84 and one finds large ho corrections throughout the whole range ( about a factor 2 ) . the @xmath75 distribution of the d - f component , with @xmath73 gev , is shown in fig . [ fig : ygam - res - lab ] . it is essentially flat and rather small , the same features as in fig . [ fig : ygam - res - cm * ] . the ho contribution is roughly independent of @xmath75 and doubles the born term . the distribution is relatively stable under the @xmath85 parameter : if taken to zero the d - f component is 13% larger in the backward bin decreasing to no variation in the most forward bin . we have in the r - f ho contribution a piece similar to that in the r - d ho one . a large-@xmath7 parton ( in the lab ) can be emitted collinearly with an initial parton in the resolved virtual photon and then fragments into a photon . again this corresponds to a collinear configuration in the cm@xmath0 frame that we cut by the requirement @xmath86 . therefore this contribution has a strong dependence on this cut - off parameter . however , as seen in fig . [ fig : ygam - res - lab ] it is the smallest of all four components and will give a negligible contribution to the full isolated @xmath8-jet cross section . after this lengthy discussion on each of the components which build the @xmath8-jet cross section we are now ready to compare the nlo calculation to the most recent zeus data . it is worth stressing that only the sum of all four components have a physical meaning and can be compared to data as the relative weight of each component depends on the unphysical scales . it has been stressed above that measuring the transverse components in the laboratory frame rather than the cm@xmath0 frame does not garantee the perturbative stability of the calculation : indeed it appears from the above discussion that , if the direct components are rather stable under the @xmath87 parameter , this is not the case for the resolved ones because the zeus phase space does not garantee that the cm@xmath0 momenta in the hard subprocesses remain large and also does not eliminate all the collinear singularities associated to the photon radiation from a quark at large laboratory transverse momentum . for these reasons we will bracket the theoretical predictions distribution : comparison of the full nlo calculation to zeus data @xcite . the leptonic contribution is as quoted in @xcite . _ _ ] by two cross sections : one obtained with @xmath88 gev , which garantees the perturbative stability but excludes some @xmath8-jet configurations included in the data , the other with @xmath89 gev which accounts for a larger phase space at the cost of letting the theoretical predictions err in the non - perturbative regime . the results of our approach are compared with data in figs . [ fig : ygam - zeus - lab]-[fig : ptjet - zeus - nores - lab ] . in all plots we include the contribution from the leptonic radiation of the photon ( @xmath90 ) based on the djangoh @xcite generator , as estimated in @xcite . we add it to the hadronic results obtained in this work , with a numerical accuracy of better than 2% , to obtain cross sections comparable to the data . the interference contribution between leptonic and hadronic emissions is negligible @xcite . furthermore no hadronization corrections are applied to our estimates . [ fig : ygam - zeus - lab ] we display the photon rapidity spectrum and we note the very good agreement with the data when the cutoff @xmath88 gev is selected . using the lower cut - off uniformely increases the predictions by about 12% . in contrast , the predictions of @xcite , as quoted in ref . @xcite , where the resolved component is treated at lowest order , systematically underestimate the photon rapidity spectrum by about 20% . + distribution : comparison of the full nlo calculation to zeus data @xcite . the leptonic contribution is as quoted in @xcite . _ _ ] distribution : comparison of the partial nlo calculation , with the unresummed resolved component , to zeus data . the data and the leptonic contribution are as quoted in @xcite . _ _ ] the same pattern is found in fig . [ fig : pt - zeus - lab ] displaying the photon transverse momentum spectrum : very good agreement with @xmath88 gev , overestimate with @xmath89 gev . the interesting feature however is that both cut - offs lead to similar results in the high @xmath91 tail where we expect the nlo resummed treatment of the resolved component to be most appropriate ( @xmath92 ) . there , our predictions can be taken as solid predictions from nlo perturbative qcd as all relevant scales are large in the cm@xmath0 frame . comparing to the estimates based on @xcite and quoted in @xcite we are above : in fact they systematically underestimate the data while still being compatible with them at large @xmath91 . + turning to the @xmath16 distribution in fig . [ fig : q2-zeus - lab ] , our results confirm the discussion above : excellent agreement and stability under a change of the cut - off at small @xmath16 , overestimate of the data at the larger values . finally we conclude this phenomenological discussion with the @xmath93 spectrum , in fig . [ fig : ptjet - zeus - lab ] , where we notice a small excess in the predictions for the smallest two bins , @xmath94 gev , with a large instability under change of cut - off in the first bin . in relation with this observation one notes a large higher correction ( up to a factor 2 compared to the born term ) to the resolved component . the origin of this is the fact that the collinear fragments of the virtual photon , which would be rejected by a cut - off in the cm@xmath0 frame can be seen as jets in the laboratory . in contrast , for @xmath95 gev we observe good agreement with data as well as a very good stability under change of the cut - off which indicates that nlo perturbation theory is valid . to further explore the small @xmath93 regime we calculate the cross section without resummation on the resolved component ( see fig . [ fig : ptjet - zeus - nores - lab ] ) : agreement with data is achieved for the first two bins while above 6 gev the cross section is systematically underestimated . a last remark concerning the comparison between theory and experiment at small @xmath12 is of a more practical nature : the theoretical jets constructed at the partonic level are compared with the experimental ones at the hadronic level and hadronization corrections not taken into account at nlo qcd must be important . we have discussed at length the full nlo qcd predictions for @xmath8-jet final states in deep - inelastic scattering at hera and compared with zeus data where transverse momenta are defined in the laboratory frame . the motivation for studying these final states is to avoid the photon - quark final state collinear singularity present in the born term @xmath96 when the photon , produced longitudinally in the cm@xmath0 frame , is detected at large transverse momentum in the laboratory frame due to the transverse boost of the virtual photon . we have seen that the full nlo qcd calculation , taking into account the ho corrections to the resummed resolved component , again hits this collinear singularity problem which occurs in the large @xmath16 , low @xmath91 or @xmath93 regions . so , in deep inelastic scattering , doing large @xmath12 phenomenology with momenta defined in the laboratory is not enough to cure this collinear problem since it re - enters through the back door when considering higher order diagrams as are needed for precision calculations . of course , working with cm@xmath0 coordinates , or with a cut - off in the cm@xmath0 frame , one would not have encountered this difficulty . nevertheless , in some well identified kinematical regions in the laboratory ( low @xmath16 or large @xmath12 ) comparaison between experiment and the full nlo theory is astonishingly good unlike the case when considering partial nlo calculation where the resolved component is treated at lowest order . of course further data involving a cut in the cm@xmath0 frame should permit a sounder comparison between theory and experiment . to be complete a study of the scale dependence of the theoretical estimates should have been done . however in view of the problems discussed above this is not justified at this stage . it is enough to remember that the choice of scales used in this work ( @xmath97 ) gave very good agreement with data for related processes , namely @xmath22 production in dis @xcite and photoproduction ( @xmath98 ) of isolated photons @xcite ( see @xcite for the theoretical discussion ) . in this respect , we are not as pessimistic as the authors of @xcite who observe , based on comparison of older data with previous theoretical calculations , that isolated photon production both in dis and photoproduction is underestimated in the nlo qcd approach . we would like to thank peter bussey for carefully reading the manuscript and for his useful remark on the @xmath99 spectrum . 99 h1 collab . , f. d. aaron et al . , epj * c54 * ( 2008 ) 371 . zeus collab . , s. chekanov et al . , phys b687 * ( 2010 ) 16 . zeus collab . , h. abramowicz et al . , phys * 715 * ( 2012 ) 88 . a. gehrmann - de ridder , g. kramer , h. spiesberger , nucl . * b578 * ( 2000 ) 326 . a. gehrmann - de ridder , t. gehrmann , e. poulsen , epj * c47 * ( 2006 ) 395 . p. aurenche , rahul basu , m. fontannaz , epjc * 71 * ( 2011 ) 1616 . m. fontannaz , epj * c38 * ( 2004 ) 297 . h1 collab . , a. aktas et al . , epj * c36 * ( 2004 ) 441 . h1 collab . aaron et al . , epj * c65 * ( 2010 ) 363 . zeus collab . , s. chekanov et al . , nucl * b786 * ( 2007 ) 152 . s. catani , m. fontannaz , j. ph . guillet , e. pilon , jhep * 1309 * ( 2013 ) 007 . ellis and d.e . soper , phys . rev . * d48 * ( 1993 ) 3160 . j. pumplin , d.r . stump , j. huston , h.l . lai , p.m. nadolsky , w.k . tung , jhep * 0207 * ( 2002 ) 012 . p. aurenche , rahul basu , m. fontannaz , r. m. godbole epj * c42 * ( 2005 ) 43 . l. bourhis , m. fontannaz , j. ph . guillet , epj * c2 * ( 1998 ) 529 . p. aurenche , rahul basu , m. fontannaz , r. m. godbole epj * c34 * ( 2004 ) 277 . p. aurenche , rahul basu , m. fontannaz , epj * c57 * ( 2008 ) 681 . s. kawabata , comp . * 88 * ( 1995 ) 309 . jetphox , z. belghobsi , m. fontannaz , j .- ph . guillet , g. heinrich , e. pilon , m. werlen , phys . rev . * d79 * ( 2009 ) 114024 ; the code is available at http://lapth.cnrs.fr/phox@xmath100family/jetphox.html . p. chiappetta , m. greco , j.ph . guillet , s. rolli , m. werlen , nucl . * b412 * ( 1994 ) 3 . + p. aurenche , m. fontannaz , j.ph . guillet , b.a . kniehl , e. pilon et m. werlen , eur . j. * c13 * ( 2000 ) 347 . a. gehrmann - de ridder and e.w.n . glover , nucl . * b517 * ( 1998 ) 269 . glover , a.g . morgan , z.phys . * c62 * ( 1994 ) 311 . k. charchula , g.a . schuler , h. spiesberger , comp . . comm . * 81 * ( 1994 ) 381 . zeus collaboration , h. abramowicz _ et al . b730 * ( 2014 ) 293 . m. fontannaz , j .- ph . guillet , g. heinrich , epj * c21 * ( 2001 ) 303 ; + m. fontannaz , g. heinrich , epj * c34 * ( 2004 ) 191 . p.r . newman , m. wing , rev . * 86 * ( 2014 ) 1037 .	we present the complete next - to - leading order calculation of isolated prompt photon production in association with a jet in deep - inelastic scattering . the calculation involves , direct , resolved and fragmentation contributions . it is shown that defining the transverse momenta in the proton virtual - photon frame ( cm@xmath0 ) , as usually done , or in the laboratory frame ( lab ) , as done in some experiments , is not equivalent and leads to important differences concerning the perturbative approach . in fact , using the latter frame may preclude , under certain conditions , the calculation of the next - to - leading order correction to the important resolved component . a comparaison with the latest zeus data is performed and good agreement is found in the perturbatively stable regions . + lpt - orsay-14 - 85 + lapth-229/14 + to 1 truecm * photon - jet cross sections in deep - inelastic scattering * 2 truecm * p. aurenche@xmath1 and m. fontannaz@xmath2 * 3 truemm _ @xmath3 lapth , universit de savoie , cnrs + bp 110 , chemin de bellevue , 74941 annecy - le - vieux cedex , france _ 3 truemm _ @xmath4 laboratoire de physique thorique , umr 8627 du cnrs , + universit paris - sud , btiment 210 , 91405 orsay cedex , france _ 2 truecm @xmath5 e - mail : [email protected] @xmath6 e - mail : [email protected]
recently , a large class of manganese oxides ( rmno@xmath7 , @xcite and rmn@xmath0o@xmath1 , @xcite with r = y , tb , dy , etc . ) has been discovered to be multiferroic , with strong magnetoelectric ( me ) coupling . the me coupling leads to various novel physical effects , such as the `` colossal magnetodielectric '' ( cmd ) effects and magneto - spin - flop effects . @xcite for example , in tbmn@xmath0o@xmath1,@xcite the me coupling is so strong , that the electric polarization can be reversed by applying a magnetic field . @xcite the remarkable me effects revealed in these materials have attracted great attention @xcite because of the fascinating physics and their potential applications in novel multifunctional me devices . the crystal structure of tbmn@xmath0o@xmath1 is orthorhombic , with four chemical formula units per primitive cell ( 32 atoms in total ) , containing mn@xmath8o@xmath9 octahedra and mn@xmath10o@xmath1 pyramids , as shown in fig.[fig : structure ] . tbmn@xmath0o@xmath1 shows several magnetic phase transitions accompanied with the appearance of electric polarizations and dielectric anomalies , when cooling down from the room temperature . @xcite starting from an incommensurate antimagnetic ( icm ) ordering at @xmath11 = 43k with a propagation vector @xmath12 ( @xmath130.50 , 0 , 0.30 ) , the structure locks into commensurate antimagnetic ( cm ) state at @xmath14 = 33k with @xmath15=(0.5 , 0 , 0.25 ) , during which spontaneous polarization arises at @xmath16 = 38k . @xcite when the temperature lowers to @xmath17 = 24 k , magnetic order becomes icm again , with a sudden decrease of polarization and a jump of the @xmath15 vector to ( 0.48 , 0 , 0.32 ) . the spontaneous polarization increase again , when continuing to cool to about 10k . @xcite during the magnetic phase transitions , a peak at @xmath16 and a step at @xmath17 of the dielectric constant was observed,@xcite indicating strong me coupling in this compound . it was demonstrated the electric polarization can be reversed by applying magnetic field . @xcite experimental data show that the structure of tbmn@xmath0o@xmath1 has space group _ pbam _ , @xcite which includes spatial inversion ( @xmath18 ) symmetry . it is therefore puzzling that the material can develop spontaneous electric polarizations . it has been suspected @xcite that the actual symmetry group of tbmn@xmath0o@xmath1 is _ _ pb__2@xmath19__m _ _ , allowing polarization along the @xmath3 axis . indeed , there are several experiments supporting this hypnosis . @xcite for example , some raman modes were found to be ir active in tbmn@xmath0o@xmath1 , @xcite and the anomalies of atomic displacement parameters ( adp ) have been observed.@xcite nevertheless , no _ direct _ evidence of the lower symmetry has yet been found . @xcite theoretically , the microscopic origin of the strong me coupling and the electric polarization is still under intensive debates . @xcite the me coupling could originate either from the symmetric superexchange interactions , or from the antisymmetric exchange interactions . @xcite the antisymmetric exchange comes from the spin - orbit coupling , and the noncollinearity of the spin structure is an essential ingredient for this mechanism . @xcite however it was shown in ref . that the largest electric polarization in tbmn@xmath0o@xmath1 is associated with the cm state that is almost collinear . @xcite in our recent work , @xcite we determined the ground - state structure of tbmn@xmath0o@xmath1 using the first - principles methods . the results show that the ground - state structure is indeed of polar space group @xmath2 and the electric polarization equals @xmath4 @xmath5@xmath6 . in the calculations , we use collinear spin approximation and ignore the spin - orbit interaction , suggesting that me coupling in tbmn@xmath0o@xmath1 is due to the symmetric superexchange interactions . the aim of the present work to examine rigorously the ground state structural , electronic and lattice dielectric properties of tbmn@xmath0o@xmath1 using first - principles calculations to provide solid ground for further investigations . the rest of paper is organized as follows . after a brief discussion of the first - principles methods and the approximations used in the calculations in sec . [ sec : methodology ] , we provide a detailed analysis of the ground - state structural and electronic properties in sec . [ sec : structure ] , [ sec : e_structure ] . in sec . [ sec : phonon ] , we calculate all zone center optical phonon frequencies and the oscillator strengths of ir modes . the results are in excellent agreement with the known experimental ir and raman spectra . in sec . [ sec : polarization ] , we calculate electric polarization in tbmn@xmath0o@xmath1 . we then derive an effective hamiltonian to explain the microscopic mechanisms of the ferroelectricity and the giant magnetoelectric coupling . we conclude in sec . [ sec : summary ] . our calculations are based on the standard density - functional ( dft ) theory with spin - polarized generalized gradient approximation ( gga ) . we adopt perdew - burke - ernzerhof functional@xcite implemented in the vienna _ ab initio _ simulations package ( vasp ) . @xcite a plane - wave basis and projector augmented - wave ( paw ) pseudopotentials @xcite are used , with mn 3__p__3__d__4__s _ _ , and tb 5__p__5__d__6__s _ _ electrons treated self - consistently . a 500 ev plane - wave cutoff results in good convergence of the total energies . we relax the structure until the changes of total energy in the self - consistent calculations are less than 10@xmath20 ev , and the remaining forces are less than 1 mev / . experimentally , tbmn@xmath0o@xmath1 is found to be incommensurate anti - ferromagnetic ( afm ) below 24 k , with the propagation vector @xmath21 . to accommodate the magnetic structure , one needs a huge supercell , which is computationally prohibitive . instead , we use a 2@xmath221@xmath221 supercell , equivalent to approximating the propagation vector @xmath23 . the validity of this approximation has been justified in our previous work . @xcite for the supercell we used , a @xmath24 monkhorst - pack k - points mesh converges very well the results . it was demonstrated in ref . that the largest electric polarization is associated with a commensurate magnetic ( cm ) state that is almost collinear . @xcite therefore , in the calculations , we use the collinear spin approximation and ignore the spin - orbit coupling . our results agree very well with the known experiments , indicating that these approximations capture the essential physics in tbmn@xmath0o@xmath1 . o@xmath1 unit cell , showing mn@xmath8o@xmath9 octahedra and mn@xmath10o@xmath1 pyramids . ] .the calculated lattice constants compared with experimental data . @xmath25 and @xmath26 are the structures of spin configurations @xmath27 and @xmath28 in fig . [ fig : spinconfigurations ] , respectively . [ cols="^,^,^,^",options="header " , ] [ tab : diff_js ] in this appendix , we discuss in more details about the magnetic structures of tbmn@xmath0o@xmath1 and the ( super)exchange interactions @xmath29s . @xcite to determine the @xmath29s , we calculate the total energies of eight scs shown in fig . [ fig : spinconfigurations ] , under the fixed _ high _ symmetry structure @xmath30 . the total energies of these scs are listed in table [ tab : diff_js ] , among which sc @xmath28 has the lowest total energy . we then use a heisenberg model with nn interaction , @xmath31 to fit @xmath29s and to check whether the @xmath28 is indeed the the ground state sc . here , @xmath32 is the spin of the @xmath33-th mn ion and @xmath34 denotes the exchange integral between two nn @xmath33-th and @xmath35-th atoms . @xmath36 , @xmath37 and @xmath38 are fitted using four afm scs , and the results are listed in table [ tab : diff_js ] . since @xmath39 and @xmath40 is not relevant here , they are not fitted . we calculate the total energies of the four fm scs from eq.([eq : heisenbergmodel ] ) . these results are in reasonable good agreement with the first - principles calculations , suggesting that the heisenberg model is valid even in the fm phase when applied to tbmn@xmath0o@xmath1 . from table [ tab : diff_js ] , we find all the three exchange integrals are negative , i.e. , of afm type , and @xmath41 . therefore the ground state sc are mainly determined by @xmath37 and @xmath38 . it is easy to see that sc @xmath28 ( and @xmath42 ) indeed has the lowest energy in this model . in both scs @xmath28 and @xmath42 , mn@xmath8 form an afm square lattice in the @xmath43 plane , whereas mn@xmath10 couples to mn@xmath8 either antiferromagnetically via @xmath37 along the @xmath27 axis or with alternating sign via @xmath36 along the @xmath3 axis . the zigzag chains along the @xmath3 axis can be simplified as mn@xmath10-mn@xmath8-mn@xmath10 segments linked via @xmath38 superexchange interactions . inside the segments mn@xmath10 and mn@xmath8 interact through superexchange @xmath36 , whereas the chains couple to each other through @xmath37 . the afm chains along the @xmath27 axis are of two types , labeled as i and ii respectively in fig . [ fig : chains ] . adjacent chains i and ii along the @xmath3 axis couple to each other via @xmath36 , in which half of them are frustrated . the scs @xmath28 and @xmath42 have several energetically degenerate scs as shown in fig . [ fig : chains ] . there are two relative positions between chain i and ii , @xmath44 and @xmath45 . `` @xmath44 '' ( `` @xmath45 '' ) means that the mn@xmath8 ions in chain i have the same spins to the mn@xmath8 ions on their _ left _ ( _ right _ ) side in chain ii in each unit cell . figure [ fig : chains ] shows 4 different combinations of chains i and ii . figure [ fig : chains](a ) and ( b ) are the ground sc @xmath28 and @xmath42 respectively , whereas fig.[fig : chains](c ) and ( d ) are in fact two domain wall structures @xmath46 and @xmath47 . the two domain walls are energetically degenerate to @xmath28 and @xmath42 in the nn heisenberg model , because the @xmath36 exchange interactions between chain i and ii cancel out no matter what their relative positions are . however , after turning on the spin - lattice coupling , the crystal structure @xmath30 will relax to @xmath26 or @xmath48 , and the energy degeneracy between the domain wall structures ( @xmath46 and @xmath47 ) to scs @xmath28 and @xmath42 will be lifted .	the structural , electronic and lattice dielectric properties of multiferroic tbmn@xmath0o@xmath1 are investigated using density functional theory within the generalized gradient approximation ( gga ) . we use collinear spin approximations and ignore the spin - orbit coupling . the calculated structural parameters are in excellent agreement with the experiments . we confirm that the ground state structure of tbmn@xmath0o@xmath1 is of space group @xmath2 , allowing polarizations along the @xmath3-axis . the spontaneous electric polarization is calculated to be @xmath4 @xmath5@xmath6 . the calculated zone - center optical phonons frequencies and the oscillator strengths of ir phonons agree very well with the experimental values . we then derive an effective hamiltonian to explain the magnetically - induced ferroelectricity in this compound . our results strongly suggest that the ferroelectricity in tbmn@xmath0o@xmath1 is driven by the magnetic ordering that breaks the the inversion symmetry , without invoking the spin - orbit coupling .
the complex error function also widely known as the faddeeva function can be defined as @xcite @xmath2 where @xmath3 is the complex argument . it is a solution of the following differential equation @xcite @xmath4 where the initial condition is given by @xmath5 the complex error function is principal in a family of special functions . the main functions from this family are the dawsons integral , the complex probability function , the error function , the fresnel integral and the normal distribution function . the dawsons integral is defined as @xcite @xmath6 it is not difficult to obtain a relation between the complex error function and the dawson integral . in particular , comparing right sides of equations and immediately yields @xmath7 another closely related function is the complex probability function . in order to emphasize the continuity of the complex probability function at @xmath8 , it may be convenient to define it in form of principal value integral @xcite @xmath9 or @xmath10 the complex probability function has no discontinuity at @xmath11 and @xmath12 since according to the principal value we can write @xmath13 where @xmath14 $ ] . there is a direct relationship between the complex error function and the complex probability function . in particular , it can be shown that these functions are actually same on the upper half of the complex plain @xcite @xmath15 \ge 0.\ ] ] separating the real and imaginary parts of the complex probability function leads to @xmath16 and @xmath17 respectively , where the principal value notations emphasize that these functions have no discontinuity at @xmath11 and @xmath12 . in particular , in accordance with equation we can write @xmath18 and @xmath19 as it follows from the identity , for non - negative @xmath20 we have @xmath21 the real part @xmath22 of the complex probability function is commonly known as the voigt function that is widely used in many disciplines of applied mathematics @xcite , physics @xcite and astronomy @xcite . mathematically , the voigt function @xmath22 represents a convolution integral of the gaussian and cauchy distributions @xcite . the voigt function is widely used in spectroscopy as it describes quite accurately the line broadening effects @xcite . although the imaginary part @xmath23 of the complex probability function also finds many practical applications ( see for example @xcite ) , it has no a specific name . therefore , further we will regard this function simply as the @xmath24-function . other associated functions are the error function of complex argument @xcite @xmath25 \quad \leftrightarrow \quad { \rm{erf}}\left ( z \right ) = 1 - { e^ { - { z^2}}}w\left ( { iz } \right),\ ] ] the plasma dispersion function @xcite @xmath26 the fresnel integral @xcite @xmath27/2 \end{aligned}\ ] ] and the normal distribution function @xcite @xmath28 . \end{aligned}\ ] ] due to a remarkable identity of the complex error function @xcite @xmath29 it is sufficient to consider only i and ii quadrants in order to cover the entire complex plane ( see appendix a ) . this property can be explicitly observed by rearranging this identity in form @xmath30 thus , due to this remarkable identity of the complex error function we can always avoid computations involving negative argument @xmath20 by using simply the right side of this equation . therefore , we will always assume further that @xmath31 . since the integral on right side of the equation has no analytical solution , the computation of the complex error function must be performed numerically . there are many approximations of the complex error function have been reported in the scientific literature @xcite . however , approximations based on rational functions may be more advantageous to gain computational efficiency due to no requirement for nested loop involving decelerating exponential or trigonometric functions dependent upon input the parameters @xmath32 and @xmath20 . although the rational approximations of the complex error function are rapid and provide high - accuracy over the most area of the complex plain , all of them have the same drawback ; their accuracy deteriorates with decreasing parameter @xmath20 ( in fact , the computation of the complex error function at small @xmath33 is commonly considered problematic for high - accuracy computation of the voigt / complex error function not only in the rational approximations @xcite ) . in this work we propose a new rational approximation of the dawson integral and show how its algorithmic implementation effectively resolves such a problem in computation of the complex error function at small @xmath33 . in our recent publication @xcite we have shown that a new methodology of the fourier transform results in a rational approximation @xmath34 where the corresponding expansion coefficients are given by @xmath35 @xmath36 @xmath37 and @xmath38 is the step . in algorithmic implementation it is convenient to rewrite the equation above in form @xmath39 using the residue calculus , from approximation we can obtain the rational approximation for the @xmath24-function ( see appendix b ) @xmath40.\ ] ] as @xmath41 $ ] from the identity it follows that @xmath42 = \frac{{\sqrt \pi } } { 2}l\left ( x \right).\ ] ] consequently , according to this identity and equation we can write @xmath43.\ ] ] since the argument @xmath32 is real , it would be reasonable to assume that this equation remains valid for a complex argument @xmath3 as well if the condition @xmath33 is satisfied . therefore , substituting the rational approximation of the dawson integral into identity results in @xmath44 , \qquad { \mathop{\rm im}\nolimits } \left [ z \right ] < < 1 . \end{aligned}\ ] ] the representation of the approximation can be significantly simplified as given by @xmath45 < < 1 , \end{aligned}\ ] ] where the expansion coefficients are @xmath46 @xmath47 and @xmath48 as we can see from equation only the @xmath49-function needs a nested loop in multiple summation . therefore , most time required for computation of the complex error function is taken for determination of the @xmath49-function . however , this approximation is rapid due to its simple rational function representation . although the first term @xmath50 in approximation is an exponential function dependent upon the input parameter @xmath51 , it does not decelerate the computation since this term is not nested and calculated only once . consequently , the approximation is almost as fast as the rational approximation . the relatively large area at the origin of complex plane is the most difficult for computation with high - accuracy . according to karbach _ @xcite the newest version of the _ roofit _ package , written in c / c++ code , utilizes the equation ( 14 ) from the work @xcite in order to cover accurately this area , shaped as a square with side lengths equal to @xmath52 . in the algorithm we have developed , instead of the square we apply a circle with radius @xmath53 centered at the origin . this circle separates the complex plane into the inner and outer parts . only three approximations can be applied to cover the entire complex plain with high - accuracy . the outer part of the circle is an external domain while the inner part of the circle is an internal domain consisting of the primary and secondary subdomains . these domains are schematically shown in fig . 1 . external domain is determined by boundary @xmath54 and represents the outer area of the circle as shown in fig . in order to cover this domain we apply the truncated laplace continued fraction @xcite @xmath55 that provides a rapid computation with high - accuracy . it should be noted , however , that the accuracy of the laplace continued fraction deteriorates as @xmath56 decreases . the inner part of the circle bounded by @xmath57 is divided into two subdomains . most area inside the circle is occupied by primary subdomain bounded by @xmath58 . for this domain we apply the rational approximation that approaches the limit of double precision when we take @xmath59 , @xmath60 and @xmath61 ( see our recent publication @xcite for description in determination of the parameter @xmath38 ) . the secondary subdomain within the circle is bounded by narrow band @xmath1 ( see fig . 1 along @xmath32-axis ) . the rational approximation sustains high - accuracy within all domain @xmath62 required for applications using the hitran molecular spectroscopic database @xcite . however , at @xmath63 its accuracy deteriorates by roughly one order of the magnitude as @xmath20 decreases by factor of @xmath64 . the proposed approximation perfectly covers the range @xmath33 . in particular , at @xmath59 , @xmath60 and @xmath61 this approximation also approaches the limit of double precision as the parameter @xmath20 tends to zero . in order to quantify the accuracy of the algorithm we can use the relative errors defined for the real and imaginary parts as follows @xmath65 - { \mathop{\rm re}\nolimits } \left [ { w\left ( { x , y } \right ) } \right]}}{{{\mathop{\rm re}\nolimits } \left [ { { w_{ref}}\left ( { x , y } \right ) } \right ] } } } \right\|\ ] ] and @xmath66 - { \mathop{\rm im}\nolimits } \left [ { w\left ( { x , y } \right ) } \right]}}{{{\mathop{\rm im}\nolimits } \left [ { { w_{ref}}\left ( { x , y } \right ) } \right ] } } } \right\|,\ ] ] where @xmath67 is the reference . the highly accurate reference values can be generated by using , for example , algorithm 680 @xcite , algorithm 916 @xcite or a recent c++ code in _ package from the cern s library @xcite . consider figs . 2 and 3 illustrating logarithms of relative errors of the considered algorithm for the real and imaginary parts over the domain @xmath68 , respectively . the rough surface of these plots indicates that the computation reaches the limit of double precision @xmath69 . in particular , while @xmath70 the accuracy can exceed @xmath71 . however , there is a narrow domain along the along @xmath32 axis in fig . 2 ( red color area ) where the accuracy deteriorates by about an order of the magnitude . apart from this , we can also see in fig . 3 a sharp peak located at the origin where accuracy is @xmath0 . figures 4 and 5 depict these areas of the plots magnified for the real and imaginary parts , respectively . it should be noted , however , that the worst accuracy @xmath0 is relatively close to the limitation at double precision computation . furthermore , since the corresponding areas are negligibly small as compared to the entire inner circle area , where the approximations and are applied , their contribution is ignorable and , therefore , does not affect the average accuracy @xmath72 . thus , the computational test reveals that the obtained accuracy at double precision computation for the complex error function is absolutely consistent with _ cernlib _ , _ libcerf _ and _ roofit _ packages ( see the work @xcite for specific details regarding accuracy of these packages ) . a matlab code for computation of the complex error function is presented in appendix c. a rational approximation of the dawson integral is derived and implemented for rapid and highly accurate computation of the complex error function . the computational test we performed shows the accuracy exceeding @xmath0 in the domain of practical importance @xmath1 . a matlab code for computation of the complex error function covering the entire complex plane is presented . this work is supported by national research council canada , thoth technology inc . and york university . it is not difficult to show that the complex error function can be expressed alternatively as ( see equation ( 3 ) in @xcite and @xcite , see also appendix a in @xcite for derivation ) @xmath73 consequently , from this equation we have @xmath74dt } \\ & = \frac{1}{{\sqrt \pi } } \int\limits _ { - \infty } ^\infty { { e^ { - { t^2}/4}}\left [ { { e^{i\left ( { x + iy } \right)t } } + { e^ { - i\left ( { x + iy } \right)t } } } \right]dt}. \end{aligned}\ ] ] using the eulers identity @xmath75 where @xmath76 , from the equation it follows that @xmath77 t } \right)dt } = 2{e^ { - { { \left ( { x + iy } \right)}^2}}}\ ] ] or @xmath78 as we have shown recently in our publication @xcite , the real part of the complex error function can be found as @xmath79 = \frac{{w\left ( { x , y } \right ) + w\left ( { - x , y } \right)}}{2}\ ] ] and since @xcite @xmath80 = k\left ( { x , y = 0 } \right ) = { e^ { - { x^2}}},\ ] ] the substitution of the approximation at @xmath11 into equation leads to @xmath81 according to the definition of the @xmath24-function , at @xmath11 we can write @xmath82 consequently , substituting the approximation for the exponential function into equation yields @xmath83dt . \end{aligned}\ ] ] the integrand of the integral is analytic everywhere except @xmath84 isolated points on the upper half plane @xmath85 therefore , using the residue theorems formula @xmath86 } } , \ ] ] where @xmath87 denotes a contour in counterclockwise direction enclosing the upper half plain ( for example as a semicircle with infinite radius ) and @xmath88 integrand of the integral , we obtain the rational approximation of the @xmath24-function . .... function ff = fadfunc(z ) % this program file computes the complex error function , also known as the % faddeeva function . it provides high - accuracy and covers the entire % complex plane . the inner part of the circle \|x + iy\| < = 8 is covered by % the equations ( 10 ) and ( 14 ) . derivation of the rational approximation % ( 10 ) and its detailed description can be found in the paper [ 1 ] . the new % approximation ( 14 ) shown in this paper computes the complex error % function at small im[z ] < < 1 . the outer part of the circle \|x + iy\| > 8 % is covered by the laplace continued fraction [ 2 ] . the accuracy of this % function file can be verified by using c / c++ code provided in work [ 3 ] . % % references % [ 1 ] s. m. abrarov and b. m. quine , a new application methodology of the % fourier transform for rational approximation of the complex error % function , arxiv:1511.00774 . % http://arxiv.org/abs/1511.00774 % % [ 2 ] w. gautschi , efficient computation of the complex error function , % siam j. numer . , 7 ( 1970 ) 187 - 198 . % http://www.jstor.org/stable/2949591 % % [ 3 ] t. m. karbach , g. raven and m. schiller , decay time integrals in % neutral meson mixing and their efficient evaluation , % arxiv:1407.0748v1 ( 2014 ) . % http://arxiv.org/abs/1407.0748 % % the code is written by sanjar m. abrarov and brendan m. quine , york % university , canada , december 2015 . % * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * % all parameters in this table are global and can be used anywhere inside % the program body . n = 1:23 ; % define a row vector sigma = 1.5 ; % the shift constant h = 6/(2pi)/23 ; % this is the step % ------------------------------------------------------------------------- % expansion coefficients for eq . ( 10 ) % ------------------------------------------------------------------------- an = 8pih^2n.exp(sigma^2 - ( 2pihn).^2).sin(4pihnsigma ) ; bn = 4hexp(sigma^2 - ( 2pihn).^2).cos(4pihnsigma ) ; cn = 2pihn ; % ------------------------------------------------------------------------- % expansion coefficients for eq . ( 14 ) % ------------------------------------------------------------------------- alpha = 8pihnsigma.exp(-(2pihn).^2).sin(4pihnsigma ) ; beta = 2exp(-(2pihn).^2).cos(4pihnsigma ) ; gamma = ( 2pihn).^2 ; % end of the table % * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ext_d = coeff(end)./z ; % start computing using the last coeff for m = 1:length(coeff ) - 1 ext_d = coeff(end - m)./(z - ext_d ) ; end ext_d = 1i / sqrt(pi)./(z - ext_d ) ; end int_d(secd ) = exp(-z(secd).^2 ) + 2ihexp(sigma^2)z(secd ) . ... theta(z(secd).^2 + sigma^2 ) ; % compute using eq . ( 14 ) int_d(~secd ) = rappr(z(~secd)+1isigma ) ; % compute using eq . ( 10 ) thf = 1./z ; for k = 1:23 thf = thf + ( alpha(k ) + beta(k)(z - gamma(k)))./ ... ( 4sigma^2gamma(k ) + ( gamma(k ) - z).^2 ) ; end end end j.p . boyd , evaluating of dawsons integral by solving its differential equation using orthogonal rational chebyshev functions , appl . comput . , 204 ( 2 ) ( 2008 ) 914 - 919 . g. pagnini and f. mainardi , evolution equations for the probabilistic generalization of the voigt profile function , j. comput . , 233 ( 2010 ) 1590 - 1595 . http://dx.doi.org/10.1016/j.cam.2008.04.040 b.m . quine and j.r . drummond , genspect : a line - by - line code with selectable interpolation error tolerance j. quant . transfer 74 ( 2002 ) 147 - 165 . http://dx.doi.org/10.1016/s0022-4073(01)00193-5 l.e . christensen , g.d . spiers , r.t . menzies and j.c jacob , tunable laser spectroscopy of @xmath91 near @xmath92 : atmospheric retrieval biases due to neglecting line - mixing , j. quant . transfer , 113 ( 2012 ) 739 - 748 . http://dx.doi.org/10.1016/j.jqsrt.2012.02.031 a. berk , voigt equivalent widths and spectral - bin single - line transmittances : exact expansions and the modtran5 implementation , j. quant . transfer , 118 ( 2013 ) 102 - 120 . http://dx.doi.org/10.1016/j.jqsrt.2012.11.026 b.m . quine and s.m . abrarov , application of the spectrally integrated voigt function to line - by - line radiative transfer modelling . transfer , 127 ( 2013 ) 37 - 48 . http://dx.doi.org/10.1016/j.jqsrt.2013.04.020 s.j . mckenna , a method of computing the complex probability function and other related functions over the whole complex plane . space sci . , 107 ( 1 ) ( 1984 ) 71 - 83 . http://dx.doi.org/10.1007/bf00649615 s.m . abrarov and b.m . quine , master - slave algorithm for highly accurate and rapid computation of the voigt / complex error function , j. math . research , 6 ( 2 ) ( 2014 ) 104 - 119 . http://dx.doi.org/10.5539/jmr.v6n2p104 s.m . abrarov and b.m . quine , sampling by incomplete cosine expansion of the sinc function : application to the voigt / complex error function , appl . comput . , 258 ( 2015 ) 425 - 435 . 10.1016/j.amc.2015.01.072[http://dx.doi.org/ 10.1016/j.amc.2015.01.072 ] s.m . abrarov and b.m . quine , efficient algorithmic implementation of the voigt / complex error function based on exponential series approximation , appl . comput . , 218 ( 5 ) ( 2011 ) 1894 - 1902 . h. amamou , b. ferhat and a. bois , calculation of the voigt function in the region of very small values of the parameter @xmath95 where the calculation is notoriously difficult , amer . chem . , 4 ( 2013 ) 725 - 731 . http://dx.doi.org/10.4236/ajac.2013.412087 l.s . rothman , i.e. gordon , y. babikov , a. barbe , d.c . benner , p.f . bernath , m. birk , l. bizzocchi , v. boudon , l.r . brown , a. campargue , k. chance , e.a . cohen , l.h . coudert , v.m . devi , b.j . drouin , a. fayt , j .- flaud , r.r . gamache , j.j . harrison , j .- hartmann , c. hill , j.t . hodges , d. jacquemart , a. jolly , j. lamouroux , r.j . le roy , g. li , d.a . long , o.m . lyulin , c.j . mackie , s.t . massie , s. mikhailenko , h.s.p . mler , o.v . naumenko , a.v . nikitin , j. orphal , v. perevalov , a. perrin , e.r . polovtseva and c. richard , the hitran2012 molecular spectroscopic database , j. quant . transfer , 130 ( 2013 ) 4 - 50 . http://dx.doi.org/10.1016/j.jqsrt.2013.07.002 s.m . abrarov and b.m . quine , a rational approximation for efficient computation of the voigt function in quantitative spectroscopy , j. math . research , 7 ( 2 ) ( 2015 ) , 163 - 174 .	in this work we show a rational approximation of the dawson integral that can be implemented for high - accuracy computation of the complex error function in a rapid algorithm . specifically , this approach provides accuracy exceeding @xmath0 in the domain of practical importance @xmath1 . a matlab code for computation of the complex error function with entire coverage of the complex plane is presented . + * keywords : * complex error function , faddeeva function , dawsons integral , rational approximation
the recently discovered ring - like structure ( the monoceros / canis major ring ) in the anti - center direction of the galaxy ( e.g. newberg et al . 2002 , ibata et al . 2003 ) could be the consequence of accretion of dwarf galaxies by the milky way ( cf . martin et al . 2004 , bellazzini et al . 2006 and references therein ) . helmi et al . ( 2003 ) suggested that the ring may be a tidal arc produced by stars stripped away from the parent satellite galaxy during a recent interaction . this would produce an asymmetric structure above and below the disc , limited in galactic longitude and with a significant velocity gradient in the galactic longitude direction . if the ring is symmetric without velocity gradient , it could be produced by ancient minor mergers . from the analysis of the asymmetries in the population of galactic m - giant stars , martin et al . ( 2004 ) have reported detection of the canis major dwarf galaxy ( cma ) located at 7 - 8 kpc from the sun and centered at @xmath3 and @xmath4 . cma was identified as strong elliptical - shaped over - density of m - giants by the comparison of star counts in northern and southern galactic hemispheres . the @xmath5 colour magnitude diagram ( cmd ) of a field located at @xmath6 from the center of the cma by bellazzini et al . ( 2004 ) revealed a population of an intermediate / old ( @xmath7 gyr ) and moderately metal deficient stellar system located a distance of @xmath8 kpc . these results have also been supported by the cmd presented by martinez - delgado et al . a blue plume ( bp ) of possibly young stars or blue stragglers has also been detected in the cmd . martin et al . ( 2005 ) further reported existence of cma on the basis of distance - radial velocity gradient among cma stars that can be explained as the effect of on - going tidal disruption of the stellar system . momany et al . ( 2004 , 2006 ) suggested that the cma over - density is an effect of the galactic warp . the southern stellar maximum warp occurring near @xmath3 ( for @xmath9 kpc ) brings down the milky way mid - plane by @xmath10 in this direction . the regularity and consistency of the stellar , gaseous and dust warp argues strongly against a recent merger scenario for cma . they present evidence to conclude that all observed parameters of cma are consistent with it being a normal milky way outer - disk population . carraro et al . ( 2005 ) and moitinho et al . ( 2006 ) have also detected bp population ( age @xmath11 myr ) in the background of several open clusters in the third galactic quadrant . they concluded that bp population is associated to norma - cygnus spiral arm and the over - density towards cma is simply a projection effect of looking along the nearby local arm . the detection of the cma over - density by martin et al . ( 2004 ) has attracted the attention towards the third galactic quadrant of the milky way . young open clusters are one of the best tools to study the spiral arms and structure of the milky way ( becker & fenkart 1970 ; pandey et al . 1988 , 1990 ) . relevant parameters like their distance , age , reddening towards the cluster can be measured more accurately than for single stars . we have embarked on a wide field ccd photometric survey around open clusters using the kiso schmidt telescope ( japan ) , since extensive studies of the coronal regions of open clusters have not been carried out so far mainly because of non - availability of photometry in a large field around open star clusters . we have undertaken a @xmath12 photometric study of a large area around open clusters using the @xmath13 @xmath12 mounted on the 105-cm schmidt telescope of the kiso observatory which gives @xmath14 field . results on few open clusters have been reported by pandey et al . ( 2001 , 2005 , 2006 ) and sharma et al . we also noticed undesired background contamination in the cmds of open clusters located in the second and third galactic quadrant , therefore in the light of above discussions it is considered worthwhile to study the background population towards the target clusters . we have carried out ccd photometry of several clusters , in the galactic longitude range @xmath15 , during 1999 to 2001 using the 105 cm schmidt telescope of the kiso observatory . the ccd camera used a site 2048@xmath162048 pixel@xmath17 tk2048e chip having a pixel size 24 @xmath18 m . at the schmidt focus ( f/3.1 ) each pixel of ccd corresponds to @xmath19 and the entire chip covers a field of @xmath14 on sky . the read out noise and gain of the ccd are 23.2 @xmath20 and 3.4 @xmath20/adu respectively . the analysis presented by pandey et al . ( 2001 , 2005 , 2006 ) and sharma et al . ( 2006 ) indicates accuracy and consistency of the data with the previously published work . the data is homogeneous and errors are @xmath21 mag at @xmath22 mag . here we used the data set where @xmath23 band observations are available because @xmath23 band allow an accurate estimation of reddening towards the cluster region . details of the clusters are given in table 1 . we have supplemented present data with the data available in the literature . carraro et al . ( 2005 ) from their ccd data of eleven open clusters , in the longitude range @xmath24 , have reported a detection of a young stellar population ( @xmath25 myr ) located at a distance of @xmath26 kpc . they also estimated reddening towards the bp direction by using @xmath27 two colour diagrams . the result for eleven clusters namely ngc 2302 , ngc 2362 , ngc 2367 , ngc 2383 , ngc 2384 , ngc 2432 , ngc 2439 , ngc 2533 , be 33 , ru 55 and to 1 have been taken directly from carraro et al . @xmath28 ccd photometric data for 5 clusters have been taken from phelps & janes ( 1994 ) . the data for another two clusters ( ngc 2099 and ngc 2168 ) is taken from kalirai et al . ( 2001 , 2003 ) and webda . all the relevant parameters of clusters taken from the literature are given in table 2 . thus in the present paper we studied bp towards 26 clusters distributed in the galactic longitude range @xmath29 . [ cols=">,>,>,^,^,^,^,^,>,^,^ " , ] [ multiblock footnote omitted]1 phelps & janes ( 1994 ) ; 2 bragaglia et al . ( 2006 ) ; 3 kalirai et al . ( 2001 , 2003 ) ; 4 carraro et al . ( 2005 ) ; 5 webda first of all we determined the reddening towards the direction of cluster region using the @xmath27 two - colour diagram ( tcd ) . the reddening is obtained by using the slide - fit method and assuming a normal reddening law i.e. , @xmath30 = 0.72 . details of the procedure are given in earlier works ( pandey et al . 2001 , 2005 ) . the reddening determination using the @xmath27 tcds is independent of distance of the objects . the colour - magnitude diagrams ( cmds ) are used to derive the distance to the clusters . following the traditional method , the fitting was done visually . the estimated random errors in determination of distance modulus of star clusters varies from @xmath31 0.10 mag to @xmath31 0.30 mag ( see e.g. kalirai et al . 2001 , 2003 ; pandey et al . 2001 , 2005 ; phelps & janes 1994 ; sagar et al . 2001 ; sanner et al . 2001 ) the tcds for six clusters and five clusters obtained from the kiso schmidt observations and from data given by phelps & janes ( 1994 ) are shown in figs . 1 and 2 respectively . a careful inspection of tcds indicates presence of different populations as reported by carraro et al . the reddening for the cluster population is obtained by shifting the unreddened ms along a normal reddening vector @xmath32 to match with the observed clusters ms . these fits are shown as thin line in figs 1 and 2 . the unreddened ms is further shifted to match the more reddened observed sequence which is shown by continuous thick line in figs 1 and 2 . this population is referred as blue plume ( bp ) by carraro et al . then we selected probable bp stars lying along the continuous thick line and the same were plotted as open circles in the tcds and ( @xmath33 ) cmds of the cluster region , also shown in lower panels of figs 1 and 2 . it can be seen on the tcds that the spectral type of bp stars in all the cmds varies from early b to late a and f as already indicated by carraro et al . however , the tcds are entangled for clusters members and bp stars of late b- and early a - type stars . after estimating the reddening , the cmds are used to estimate the distance to the cluster and bp by visual fitting the theoretical zams to the observations . the cmds for 11 clusters along with visually fitted zams are shown in lower panels of figs 1 and 2 . in the case of ngc 129 the background populations could not be disentangled using the tcd and cmd , however the tcd and cmd both clearly show the presence of background population . here we would also like to point out that the distance estimation of background population towards the clusters tr 1 and be 7 is based on few points as observations are available for smaller region ( @xmath34 arcmin square ) . however a well defined sequence for the background population can be seen . upper right panel of fig . 2 shows cmds for the clusters ngc 1912 and ngc 2099 where ccd observations in @xmath35 band are not available , hence @xmath36 is obtained using the photoelectric and/or photographic data . various parameters obtained from the cmds are given in tables 1 and 2 . columns 5,6,7 of the tables give the reddening , distance and age of the clusters respectively . whereas the results for bp populations are given in columns 8 and 9 of tables 1 and 2 . column 11 gives the source of the data . estimated random errors in determination of distance modulus of bp population is @xmath31 0.3 mag . the distance of the most of the bps obtained in the present work ( columns 9 , tables 1 and 2 ) in the longitude range @xmath37 is @xmath38 kpc . column 10 gives the reddening estimates towards the cluster direction from the far - infrared background ( firb ) maps of schlegel et al . ( 1998 ) , which are in general higher than the photometric estimates . there are several studies which have compared interstellar extinction in different region of the galaxy with firb reddening map and concluded that the firb map overestimates the extinction ( dutra , santiago & bica 2002 ; dutra et al . 2003 a , b ; joshi 2005 ) . the observations taken through kiso schmidt telescope cover an area of about @xmath39 on the sky , which gives an opportunity to study the field star population located out side the cluster region . in fig . 3 we have plotted the tcds and cmds for the field region located near the four clusters . as expected the sequence for cluster stars in the tcds and cmds , shown in fig . 3 , is diluted but the prominent sequence for the bps can be easily noticed just like in figs 1 and 2 . the reddened zams for the same @xmath36 and @xmath40 as used in fig . 1 is also shown in fig . fig 3 further supports the bps in the galactic latitude range @xmath41 . here it is worthwhile to mention our preliminary results regarding the cluster do 25 ( @xmath42 . 4 shows tcd and @xmath43 cmd for the cluster region . the tcd indicates a variable reddening , varying from @xmath44 to @xmath45 mag , in the cluster region . the cluster region is strongly affected by the foreground population . the distance and age to the cluster do 25 are found to be 5.8 @xmath46 0.5 kpc and 6 myr . the distance and age of the the cluster are comparable to those of bps towards the direction of the cluster do 25 . an inspection of the cmd of three open clusters namely ngc 2302 , ngc 2362 and ngc 2439 by carraro et al . ( 2005 , their figure 1 ) indicates that there seems to exist a younger population between the cluster and the bp population highlighted by the shaded region in fig . we shifted the zams ( assuming the same @xmath36 ) fitted by carraro et al . ( 2005 ) upwards by 1.25 mag , 1.75 mag and 1.0 mag respectively in the case of clusters ngc 2302 , ngc 2362 and ngc 2439 , which indicates a distance @xmath47 , 4.8 and 6.9 kpc for the intermediate background population toward the direction of clusters ngc 2302 , ngc 2362 and ngc 2439 respectively . the new fitting in the case of ngc 2302 and ngc 2362 is shown in fig . 5 by thick lines . these populations may be the part of extension of perseus arm . the bp is noticed in the cmds of all the clusters listed in tables 1 and 2 . in the case of ngc 2302 , ngc 2362 , ngc 2439 we noticed some intermediate background population lying in between the cluster and bps . the distance from the sun to the bps is in the range @xmath48 kpc and the bps seem to have a younger population of spectral type @xmath49 and later . the distribution of bps in the longitude range @xmath50 , listed in tables 1 and 2 are plotted in fig . 6 in x - y , x - z and @xmath51-z plane of the galaxy . the two curves shown in the upper panel of the figure are location of the outer ( norma - cygnus ) arm and perseus arm taken from russeil ( 2003 ) . momany et al . ( 2006 ) have estimated a fwhm @xmath311.7 kpc for the norma - cygnus arm . a strip of about 1.7 kpc wide to represent the outer arm is also shown . the bps towards the clusters ngc 1528 , ngc 2301 , ngc 2323 , ngc 2432 , ngc 2533 and ru 55 indicate that these may be part of the perseus arm , whereas bps towards the clusters be 7 , be 17 , be 33 , ngc 129 ngc 1907 , ngc 1912 , ngc 1960 , ngc 2099 , ngc 2168 , ngc 2367 , ngc 2383 , ngc 2384 , ngc 7654 , to 1 and cluster do 25 in the longitude range @xmath50 seem to follow the norma - cygnus ( outer ) arm . although in our sample we do not have objects in the outer arm in the galactic longitude range @xmath52 , however negueruela and marco ( 2003 ) have reported that beyond @xmath53 there is number of open clusters which clearly delineates the outer arm . they conclude that the existence of structure at d=5 - 6 kpc over the @xmath54 range seems rather secure . on the x - z and @xmath51-z projected diagrams , the bps in the range @xmath55 remain close to the formal galactic plane and the outer arm seems to descend for @xmath56 indicating a signature of galactic warp as discussed by carraro et al . ( 2005 ) and moitinho et al . the bp populations towards three clusters ngc 2432 , ngc 2533 and ru 55 that seem to deviate and lie above the galactic plane in @xmath51 - z diagram are in fact closer to the sun and follow the expected extension of the perseus arm ( carraro et al . 2005 ) . the galactic warp was revealed early from hi data ( kerr 1957 ) , which was further supported from ( 1 ) the @xmath57 distribution of various spiral arm tracer populations , e.g. open star clusters etc , as a function of longitude ( e.g. pandey & mahra 1987 , pandey et al . 1988 , 1990 ; djorgovski & sosin ( 1989 ) and ( 2 ) the ratio of star counts in northern and southern hemisphere as a function of longitude ( e.g. lopez - corredoira et al . the maximum upward and downward bending occurs at @xmath58 and @xmath3 respectively . to study the distribution of reddening material towards the bps we plotted interstellar absorption , @xmath59 mag / kpc for the galactic longitude range @xmath60 and @xmath61 as a function of @xmath62 distance in fig . . a gaussian fit to the distribution yields a peak of @xmath63 at @xmath64 pc and @xmath65 pc for the range @xmath60 and @xmath66 respectively . it will be worthwhile to compare the distribution of reddening material towards the bps with the distribution of reddening material near the sun . for this purpose we followed the approach described by pandey & mahra ( 1987 ) and obtained distribution of @xmath63 as a function of @xmath62 in 11 zones of galactic longitude within 3 kpc from the sun using the catalogue of open clusters compiled by dias et al . we obtained the value of @xmath62 for which the absorption is at maximum ( @xmath67 ) using the gaussian fit . in fig . 8 we plot the value of @xmath67 as a function of longitude . a least - square solution for sinusoidal function manifests that the plane defined by reddening material shows a maximum upward tilt towards @xmath68 or maximum downward tilt towards @xmath69 . this result is in agreement with our previous findings ( pandey & mahra , 1987 ) . 8 indicates that in the longitude range @xmath70 ( mean @xmath51 = @xmath71 ) , @xmath67 has a mean value of about -25 pc and at @xmath72 , the @xmath67 is found to be @xmath73 pc . the obtained values of @xmath67 towards above mentioned longitudes are in good agreement with the results obtained for the bps in fig . 7 . here we have to keep in mind that the downward maximum tilt towards @xmath3 found in fig . 8 should be considered as a local downward maximum for the population located within few kpc of the sun . therefore the bps towards @xmath3 having @xmath74 may be a part of warped galactic spiral arms . recently momany et al . ( 2006 ) using 2mass clump and red giant stars at mean heliocentric distances of @xmath75 = 2.8 , 7.3 and 16.6 kpc have confirmed southern hemisphere warp maximum around @xmath3 . therefore the bps around @xmath3 and having @xmath76 kpc may also be a part of warped outer galactic arm . within 3 kpc of the sun as a function of galactic longitude . , width=316,height=192 ] the over - density of stars in the direction of cma has been explained as a nearby galaxy is being cannibalized by the milky way ( martin et al . 2004 , bellazzini et al . 2004 , 2006 ) a wide sequence behind the ms of cluster ngc 2477 ( @xmath77 ) , which is near to the center of cma ( @xmath3 , @xmath78 ) , is attributed to the cma system ( bellazzini et al . . however carraro et al . ( 2005 ) pointed out that the same type of populations can also be seen in the cmd of ngc 2168 ( @xmath79 , @xmath80 ) which is far away from the cma . here it is further interesting to mention that cmd of the cluster ngc 7654 ( @xmath81 , @xmath82 ) , given by pandey et al . ( 2001 ) , also show a similar type of sequence behind the cluster ms which is found to be situated in the outer arm at a distance of @xmath83 kpc . a comparison of the cmds of above mentioned three clusters is shown in fig . 9 , which clearly indicates a similarity between the cmds of the clusters . here it is interesting to point out that the bp towards two clusters at @xmath84 , namely ngc 581 and tr 1 are found to be located between @xmath85 to 9.5 kpc . the work by russeil ( 2003 ) also indicates some population behind the outer arm at @xmath86 . to conclude , the cmds of the open clusters and nearby field regions used in the present analysis show a significant stellar population in background of the clusters , which may be part of the outer spiral arm as well as of perseus arm . the structure of norma - cygnus ( outer ) arm seems to be continued from @xmath1 to @xmath2 . here it is worthwhile to mention that the detection of bp population over a large range of longitudes is not against the dwarf galaxy hypothesis . all the existing models of the disruption of cma predicts that the remnant and its tidal stream should embrace the whole galactic disc ( see e.g. , martin et al . 2005 ; pennarrubia et al . 2005 ) . we acknowledge the support given by dst ( india ) and jsps ( japan ) to carry out the wide field ccd photometry around open clusters . we are also thankful to the kiso observatory for allotting time at the kiso schmidt telescope . authors are thankful to dr . m. bellazzini for useful comments that improved the scientific contents of the paper . becker , w. , & fenkart , r. , 1970 , in iau symp . 38 , the spiral structure of the galaxy , ed . w. becker & g. contopoulos ( dordrecht : reidel ) , 205 bertelli g. , bressan a , chiosi c. , fagotto f. , & nasi e. , 1994 , a&as 106 , 275 bellazzini m. , ibata r. , monaco l. , martin n. , irwin m. j. , & lewis g. f. , 2004 , mnras , 354 , 1263 bellazzini m. , ibata r. , martin n. , lewis g. f. , conn b. , & irwin m. j. , 2006 , mnras , 366 , 865 bragaglia a. , tosi m. , andreuzzi g. , & marconi g. , 2006 , mnras , 368 , 1971 carraro g. , vzquez r. a. , moitinho a. , & baume g. , 2005 , apj , 630l , 153 djorgovski s. , & sosin c. , 1989 , apj , 341 , 13 dias w. s. , alessi b. s. , moithinho a. , lepine j. r. d. , 2002 , a&a , 389 , 871 dutra c. m. , santiago b. x. , bica e. , 2002 , a&a , 381 , 219 dutra c. m. , ahumada a. v. , claria j. j. , bica e. , barbuy b. , 2003a , a&a , 408 , 287 dutra c. m. , santiago b. x. , bica e. , barbuy b. , 2003b , mnras , 338 , 253 helmi a. , navarro j. f. , meza a. , steinmetz m. , & eke v. r. , 2003 , apj , 592l , 25 ibata r. a. , irwin m. j. , lewis g. f. , ferguson a. m. n. , & tanvir n. , 2003 , mnras , 340 , 21 joshi , y. c. , 2005 , mnras , 362 , 1259 kalirai j.s . , ventura p. , richer h.b . , fahlman g.g . , durrell p.r . , dantona f. , & marconi g. , 2001 , aj , 122 , 3239 kalirai j.s . , fahlman g.g . , richer h.b . , & ventura p. , 2003 , aj , 126 , 1402 kerr f. j. , 1957 , aj , 62 , 93 lopez - corredoira m. , cabrera - lavers a. , garzn f. , hammersley p. l. , 2002 , a&a , 394 , 883 martin n. f. , ibata r. a. , bellazzini m. , irwin m. j. , lewis g. f. , & dehnen w. , 2004 , mnras , 348 , 12 martin n. f. , ibata r. a. , conn b. c. , lewis g. f. , bellazzini m. , & irwin m. j. , 2005 , mnras , 362 , 906 martinez - delgado d , butler d. j. , rix h. , franco y. i. , penarrubia j. , alfaro e. j. , & dinescu d. i. , 2005 , apj , 633 , 205 moitinho a. , vzquez r. a. , carraro g. , baume g. , giorgi e. e. , & lyra w. , 2006 , mnras , 368 , 77 momany y. , zaggia s. r. , bonifacio p. , piotto g. , de angeli f. , bedin l. r. , & carraro g. , 2004 , a&a , 421 , 29 momany y. , zaggia s. , gilmore g. , piotto g. , carraro g. , bedin l. r. , & de angeli f. , 2006 , a&a , 451 , 515 negueruela i. , & marco a. , 2003 , a&a , 406 , 119 newberg h. j. et al . , 2002 , apj , 569 , 245 pandey a. k. , & mahra h. s. , 1987 , mnras , 226 , 635 pandey a. k. , bhatt b. c. , & mahra h. s. , 1988 , a&a , 189 , 66 pandey a. k. , bhatt , b. c. , & mahra , h. s. , 1990 , a&a , 234 , 128 pandey a. k. , nilakshi , ogura k. , sagar r. , & tarusawa k. , 2001 , a&a , 374 , 504 pandey a. k. , upadhyay k. , ogura k. , sagar r. , mohan v. , mito h. , bhatt h.c . , & bhatt b.c . , 2005 , mnras , 358 , 1290 pandey a. k. , upadhyay k. , sharma s. , ogura , k. , sandhu t. s. , & mito h. , 2006 , submitted pennarrubia , j. et al . , 2005 , apj , 626 , 128 phelps r. l. , & janes , k. a. 1994 , apjs , 90 , 31 russeil d. , 2003 , a&a , 397 , 133 sagar , r. , munari , u. , & de boer , k. s. , 2001 , mnras , 327 , 23 sanner , j. , brunzendorf , j. , will , j. -m . , & geffert , m. , 2001 , a&a , 369 , 511 schlegel , d. j. , finkbeiner , d.p . , & davis , m. , 1998 , apj , 500 , 525 sharma s. , pandey a. k. , ogura k. , mito h. , tarusawa k. , & sagar r. , 2006 , aj , in press	we present colour - magnitude diagrams of open clusters , located in the range @xmath0 , manifesting stellar populations in the background of clusters . some of the populations are found to be located beyond the perseus arm and may be the part of norma - cygnus ( outer ) arm . the outer arm seems to be continued from @xmath1 to @xmath2 . the background populations follow the downward warp of the galactic plane around @xmath3 . [ firstpage ] open clusters and association : general , distances - galaxy : structure
it is postulated in quantum physics that quantities we observe are eigenvalues of operators representing the dynamics of the quantities . therefore , the energy spectra , i.e. the eigenvalues , are required to be real and bounded from below so that the system has a stable lowest - energy state . to satisfy such requirements , it was conjectured that the operators must be hermitian . non - hermitian hamiltonians have been commonly associated with complex eigenvalues and therefore decay of the quantities . however , it turned out that hermiticity is not necessarily required by a hamiltonian system to satisfy the postulate @xcite . of particular examples have been systems exhibiting the so - called parity - time ( @xmath0 ) symmetry , suggested by bender and co - workers @xcite . a necessary condition for a hamiltonian to be @xmath0symmetric is that its potential @xmath1 should satisfy the condition @xmath2 . optical analogues of such systems were proposed in @xcite using two coupled waveguides with gain and loss . note that such couplers were already studied in @xcite . the following successful experiments @xcite have stimulated extensive studies on @xmath0symmetric dimers , which are a finite - dimensional reduction of schrdinger equations modelling , e.g. , bose - einstein condensates with @xmath0symmetric double - well potentials @xcite . nontrivial characteristics of the systems allow them to be exploited , e.g. , for all - optical switching in the nonlinear regime , lowering the switching power and attaining sharper switching transition @xcite as well as a unidirectional optical valve @xcite . @xmath0symmetric analogues in coupled oscillators have also been proposed theoretically and experimentally recently @xcite . note that coupled oscillators with gain and loss have already been considered in @xcite . in this paper , we consider the following equations of motion @xcite @xmath3 where the dot represents differentiation with respect to the evolution variable , which is the propagation direction @xmath4 for nonlinear optics or the physical time @xmath5 for bose - einstein condensates , @xmath6 is the nonlinearity coefficient and @xmath7 is the gain - loss parameter . here , we consider two cases , i.e. when @xmath8 and @xmath9 . for the former case , one can scale the coefficient such that @xmath10 . it was shown numerically in @xcite that the nonlinearity suppresses periodic time reversals of optical power exchanges between the sites , leading to the symmetry breaking and a sharp beam switching to the waveguide with gain . when @xmath11 , eq . has two conserved quantities @xmath12 which are commonly referred to as the power and the hamiltonian / energy , respectively . using the liouville - arnold theorem ( or liouville - mineur - arnold theorem ) @xcite , is integrable since the degree of freedom is equal to the number of conserved quantities . by defining the site - occupation probability difference @xmath13 kenkre and campbell @xcite showed that @xmath14 satisfies a @xmath15-equation , which explains the presence of josephson tunneling and self - trapped states , with the latter corresponding to @xmath14 being sign - definite , as well as the transition between them . it was later shown @xcite that @xmath14 also satisfies the pendulum equation . when @xmath16 , eq . is still integrable @xcite . system is actually a special case of a notably integrable dimer derived in ( see also a brief review of integrable oligomers in @xcite ) . the conserved quantities in that case are @xmath17 it was reported that the general system could be reduced to a first - order differential equation with polynomial nonlinearity and it possesses blow - up solutions that was observed numerically . in this paper , we consider and show that it can be reduced to a pendulum equation with a linear potential and a constant drive . the same equation has been obtained recently , parallel to and independently from this work , by kevrekidis , pelinovsky and tyugin @xcite and by barashenkov , jackson and flach @xcite through a different formalism . the linear potential and constant drive explain the presence of unbounded solutions . we exploit the strong relation between the problem and the geometry of circles . we also discuss the qualitative pictures of all solutions of the governing equations . in section [ sec2 ] , we rewrite the governing equations in terms of power , population imbalance , and phase difference between the wavefields in the channels . in the section , we also derive a constant of motion . in section [ sec3 ] , we analyse the characteristics of the fixed points , which are the time - independent solutions of the system . in section [ sec4 ] , we reduce our system in section [ sec2 ] further into one equation . here , we show that the problem is described by a pendulum equation with a linear potential and a constant drive . in the section , we analyse the pendulum equation qualitatively through its phase - portrait . in section [ sec5 ] , we discuss the phase - portrait of the system that is composed of trajectories with the same value of a constant of motion . the constant corresponds to power , that is a conserved quantity when @xmath11 . the case of linear systems is discussed in section [ sec6 ] . finally , we conclude our work in section [ conc ] . writing @xmath18 and @xmath19 in polar form @xmath20 and defining the variable of phase difference between @xmath18 and @xmath19 @xmath21 the equations of motion can then be expressed in terms of , and as ( see the appendix ) @xmath22 we can limit the phase difference to be in the interval @xmath23 . note that the argument angle @xmath24 ( and hence @xmath25 ) is undefined when @xmath26 ( or @xmath27 ) . taking @xmath28 , the conditions for equilibrium points are @xmath29 this shows that no equilibrium points can exist with @xmath30 , therefore demonstrating the non - existence of the self trapped state that is observed for the case when @xmath11 . when @xmath31 it follows that no equilibrium points exist . this is a threshold of total @xmath0-symmetry breaking where no periodic solutions can exist . however , unbounded trajectories will be shown to always exist for any @xmath7 . after some manipulations ( see the appendix for the details ) , it is possible to find a constant of motion of , , and that is given by @xmath32 where @xmath33 is a constant . we can impose the constant to be non - negative with no loss of generality . it is important to note that one can draw similarities of to the cosine rule for triangles with three edge length : @xmath34 , @xmath33 and @xmath35 ( see fig [ sk2 ] ) . \(a ) ( b ) ( c ) by varying @xmath25 , trajectories of form circles with radius @xmath33 centred on a point a unit distance from the origin . three different sketches corresponding to different values of @xmath33 are given in fig . [ sk2 ] . equation ( [ eq : c ] ) can also be rearranged to give @xmath36 from which a few things can be said about the constant @xmath33 and how this affects what values @xmath25 can take . firstly @xmath33 must be real to give a real solution . if @xmath37 then @xmath38 . if @xmath39 then @xmath25 is bounded such that @xmath40 . the @xmath41 solution in and can exist only for @xmath33 within this range . only when @xmath42 , @xmath43 can equal @xmath44 , in which case @xmath45 . if @xmath46 then @xmath25 is unbounded it is always natural to first analyse the behaviour of the time - independent solutions . using the conditions for the equilibrium points eq . ( [ eq : eqpt ] ) alongside ( [ conmo ] ) , we can find the value of the power , @xmath47 , at equilibrium points for different values of @xmath33 , i.e.@xmath48 applying the inequality @xmath49 to eq . , eq . is also the minimum power that a ( generally time - dependent ) solution can attain . the value @xmath37 is particular as explicit solutions can be found . in this case @xmath50 , @xmath51 and @xmath52 , where @xmath53 is a constant determined by the initial conditions . this gives one example of where @xmath0-symmetry is broken for any non - zero @xmath54 . if @xmath55 , no equilibrium points can exist as the power is complex - valued . hence , all solutions are unbounded . when @xmath56 , an equilibrium point lies on the boundary of @xmath25 with @xmath57 . for trajectories with @xmath58 , there are two equilibrium points at @xmath57 . this is shown in fig . [ sk2 ] as the intersections of the circles and the dashed lines given by eq . ( [ eq : eqpt ] ) . the first of these is a centre and the second is a saddle node with the corresponding power @xmath59 and @xmath60 , respectively . exploiting the geometry of the triangles and circles in fig . [ sk2 ] , one can obtain that @xmath61 for the case of @xmath42 , the trivial equilibrium point at @xmath62 is found to represent a centre for these trajectories . there is also an equilibrium point located at @xmath63 and @xmath64 , that can be straightforwardly shown to be a saddle node . solutions with @xmath46 also have two equilibrium points located at @xmath63 and @xmath65 . the former represents a saddle node with @xmath66 and the latter is a centre with @xmath67 . from the results above , we obtain that generally speaking the larger @xmath33 is , the larger the power of stable solutions can be . in order to analyse the general solutions of the governing equations , it is useful to define a new variable @xmath68 , i.e. the arc angle , that parametrises the circles as indicated in fig . the variable can be expressed by @xmath69 where @xmath70 and @xmath71 . these equations can be rearranged to yield @xmath72 the relationships are plotted in fig . [ sk4 ] . using the remark following , note that is a lower bound to the power of solutions . after carrying out some rearranging and differentiation ( see the appendix ) , we find that @xmath73 where @xmath74 is , technically , a constant of motion . we use the word `` technically '' because care must be taken in how @xmath68 is defined . because we are choosing to define @xmath75 we must remember that as a trajectory crosses this boundary , the value of @xmath74 changes by an amount @xmath76 to ensure that the power , @xmath47 , remains continuous . if @xmath68 should be defined unbounded then @xmath74 would indeed remain constant for all time . it so happens that for @xmath28 , @xmath74 has similarities to @xmath47 with @xmath11 , where it is well known that the power remains constant for all trajectories . \(a ) ( b ) the true value of defining the variable @xmath68 and the constant @xmath33 is made apparent when we substitute @xmath77 with equations ( [ eq : cpsi1 ] ) and ( [ eq : k ] ) into equation ( [ eq : deltadot ] ) . the equations of motion then reduce to a second order differential equation @xmath78 this is a main result of this paper . when @xmath79 , we indeed obtain the equation for a pendulum similarly to @xcite that was derived through a different approach . the presence of @xmath16 introduces a linear potential and a constant drive into the pendulum equation . it may be possible to solve eq . ( [ eq : oscillator1 ] ) in terms of jacobi elliptic functions . asymptotic solutions of the equation were derived in @xcite . however , instead we will study the qualitative behaviours of the solutions in the ( @xmath68,@xmath80)-phase plane . the first integral of eq . ( [ eq : oscillator1 ] ) can be obtained by using eq . ( [ eq : cpsi3 ] ) , i.e.@xmath81 the phase - portrait can then be plotted rather easily and is demonstrated in fig . [ sk5 ] for @xmath82 . the phase plane can have two topologically different structures , one with and one without a stable region . [ padd1 ] shows how this phase plane , combined with fig . [ sk4]a ( cf . ) can be plotted in a three - dimensional graph . this then gives an idea of how solutions appear in the @xmath83 phase plane . when @xmath84 there are boundaries on @xmath25 which correspond to @xmath85 . these are represented by vertical dashed lines in fig . trajectories that cross these boundaries will behave differently , in the @xmath86 plane , from those trajectories which do not . this results in a more diverse set of solutions when @xmath84 than for otherwise . one can note that when @xmath42 these vertical dashed lines merge at @xmath87 to correspond to the point where @xmath25 is in fact undefined . it is useful to know the value of the constant @xmath74 for the trajectories that touch the boundary at @xmath88 . such values are found by using eq . ( [ eq : k ] ) and eq . ( [ eq : cpsi3 ] ) to be @xmath89 how a trajectory behaves depends on how its @xmath74-value compares to the four main values of @xmath74 , namely @xmath90 , @xmath91 , @xmath92 and @xmath93 , where @xmath94 with @xmath95 and @xmath96 being the values of @xmath68 at the equilibrium points , i.e.@xmath97 which can be obtained using - ( or similarly exploiting the geometric pictures in fig . [ sk2 ] ) . by plotting out these values of @xmath74 against @xmath33 for a fixed @xmath54 one sees that there is a point where @xmath90 and @xmath93 cross . three types of solution exist for all @xmath84 . the first of these is a stable trajectory that never touches a boundary for @xmath25 . this corresponds to @xmath98 . second are trajectories with @xmath99 . these cross one boundary of @xmath25 . the final type is an unstable trajectory with @xmath100 . in addition to the trajectories above , there are two types of solutions that can not coexist for the same @xmath33-value . this is because one solution exists when @xmath101 and the other when @xmath102 , which is obvious that both conditions can not be met at the same time . when @xmath101 , there is additionally another type of stable trajectories that crosses both boundaries of @xmath25 this is demonstrated in fig . however , if @xmath102 then instead there is another type of unstable trajectory , as shown in in fig . all of the different types of trajectory are sketched in fig . [ sk3 ] . \(a ) ( b ) \(a ) ( b ) \(a ) ( b ) + ( c ) ( d ) in the conservative case @xmath11 , the power @xmath47 is independent of time , see . in that case , plots in figs . [ padd1 ] and [ sk3 ] correspond to varying power , i.e. each trajectory has different value of @xmath47 , and a fixed energy , i.e. the same value of first integral . however , it is not the common practice as usually one plots trajectories corresponding to the same value of power . in the general case when @xmath16 , @xmath47 is no longer constant . it is therefore not possible to compose a similar phase portrait consisting of trajectories with constant @xmath47 . however , we find that one could instead plot phase - portraits with the same value of @xmath74 , which interestingly runs parallel with the different solutions observed at the same values of @xmath47 for @xmath11 , see . when comparing periodic trajectories that all have the same value of power @xmath74 , it is useful to study the ( @xmath68,@xmath47)-phase plane , which qualitatively can be obtained by plotting and analysing ( see fig . [ sk4]b ) and . by doing this , we find that different values of @xmath74 can give different results in the ( @xmath25,@xmath14)-phase plane . there are three important @xmath74 values to consider . the first is the minimum value @xmath74 can take . the second is the point where the equilibrium point at @xmath63 bifurcates from a centre to a saddle node . this is found from substituting @xmath56 into eq . ( [ eq : ksaddle ] ) . the third important value of @xmath74 is when the separatrix passing through the saddle point has a @xmath33-value of one , found from substituting @xmath42 into eq . ( [ eq : ksaddle ] ) . these values of @xmath74 are given respectively by @xmath103 substituting @xmath11 into the above expressions gives the important values of @xmath74 as 0 , 1 and 2 . realising that here @xmath104 makes clear that these are generalisations of the critical values found for @xmath47 when @xmath11 , where @xmath105 corresponds respectively to the presence of only trivial solution , the emergence of a pair of fixed points @xmath106 , and the emergence of rotational ( or running ) states @xcite . when @xmath107 all three of these values are equal , which is the threshold of total @xmath0-symmetry breaking , i.e. there are no periodic , bounded solutions for @xmath31 . note that for any @xmath108 , there are two equilibrium points . the @xmath33-values of the equilibria give the upper and lower bounds the periodic paths can have for a certain @xmath74 . \(a ) ( b ) ( c ) in fig . [ sk6 ] , we sketch the possible phase - portraits of the system for @xmath16 within three different intervals determined by the critical values of @xmath74 above . note that the figures are actually composed of trajectories taken from the phase portraits with different values of @xmath33 sketched in fig . they can also be seen as the dynamical regimes for arbitrary initial conditions , showing the regions of periodic and unbounded solutions , that we refer to here as stability and instability regions . similar figures obtained using comprehensive numerical computations were presented in @xcite . in panel ( a ) , all the solutions are periodic . as @xmath74 increases , there will be a critical value above which the only equilibrium point becomes unstable . when @xmath11 , this is where a pair of fixed points with nonzero @xmath14 emerges . for @xmath16 , instead we have unbounded solutions . the instability region is nevertheless contained within the region of stable , periodic solutions . when @xmath74 is increased further passing , the stability region that contains the instability region vanishes completely . sukhorukov , xu , and kivshar @xcite also presented the dependence of the minimal input intensity on the gain / loss coefficient and the corresponding phase difference required for nonlinear switching , i.e. the minimum intensity above which the solutions would be unbounded . according to the analytical results presented herein , those would correspond to the separatrices of ( see fig . [ sk5 ] ) . analytical expressions should be derivable , that we leave for the interested reader . after analysing the nonlinear dimer with @xmath8 , finally we discuss the linear equations described by with @xmath9 . this is particularly interesting because the system has been realised experimentally in @xcite . transforming the equations into the polar forms by similarly defining @xmath47 , @xmath14 and @xmath25 , we obtain the same equations except ( [ eq : thetadot ] ) , which is now given by @xmath109 following the similar reduction , we obtain that instead of the constant of motion is given by @xmath110 this , instead of circles , is an equation for a straight line passing a distance @xmath33 from the origin . by parametrising the line with the new variable @xmath68 defined by @xmath111 we find that ( cf . and ) @xmath112 where @xmath74 is also a constant of integration . using the above equations combined with the identity @xmath113 ( obtained from and ) in equation ( [ eq : deltadot ] ) gives @xmath114 this is a forced , simple harmonic equation and therefore can be solved explicitly to give @xmath68 in terms of @xmath5 . due to the second equation in , we therefore can conclude that the oscillating power reported in , e.g. , @xcite has internal frequency @xmath115 . we have studied analytically linear and nonlinear @xmath0-symmetric dimers , where we described the whole dynamics of the system . the effect of nonlinearity that induces @xmath0-symmetry breaking for gain / loss parameters below that of the linear system , which in a previous work was referred to as nonlinear suppression of time - reversals @xcite , has been analysed as well . our analytical study may offer a new insight into the global dynamics of directional waveguide couplers with balanced gain and loss or bose - einstein condensates in a double - well potential with a balanced sink and source of atoms . in additions to our qualitative analysis , one could extend the study here to the analytical expression of the solutions of the nonlinear @xmath0-symmetric dimer , that may be expressed in terms of jacobi elliptic functions . in that case , the oscillation frequency of the power in the nonlinear system would be obtained . here , we only derive the frequency of the linear system . differentiating the expression gives @xmath116 which by substituting into the governing equations yields @xmath117 from the real parts and @xmath118 using the product rule for differentiation we can also obtain the following four equations @xmath119 using the above equations and noting that @xmath120 eqs . , , and can be immediately obtained . next , one can compare and to obtain that @xmath121 therefore , we can define a constant @xmath33 such that @xmath122 which is nothing else but eq . . after parametrisation of the constant of motion by @xmath68 , i.e. eqs . - , one can differentiate to obtain that @xmath123 now combining it with yields @xmath124 using , we see that @xmath125 which can be integrated to give . a. ruschhaupt , f. delgado , and j. g. muga , j. phys . a 38 , l171 ( 2005 ) . r. el - ganainy , k. g. makris , d. n. christodoulides , and z. h. musslimani , opt . 32 , 2632 ( 2007 ) . s. klaiman , u. guenther , and n. moiseyev , phys . lett . 101 , 080402 ( 2008 ) . y. j. chen , a. w. snyder , and d. n. payne , ieee j. quantum electron . 28 , 239 ( 1992 ) . a. guo , g. j. salamo , d. duchesne , r. morandotti , m. volatier - ravat , v. aimez , g. a. siviloglou , and d. n. christodoulides , phys . 103 , 093902 ( 2009 ) . c. e. ruter , k. g. makris , r. el - ganainy , d. n. christodoulides , m. segev , and d. kip , nature physics 6 , 192 ( 2010 ) . miroshnichenko , b.a . malomed , and yu.s . kivshar , phys . a 84 , 012123 ( 2011 ) . h. cartarius and g. wunner , phys . a 86 , 013612 ( 2012 ) . graefe , j. phys . 45 , 444015 ( 2012 ) . j. schindler , a. li , m.c . zheng , f. m. ellis , and t. kottos , phys . a 84 , 040101(r ) ( 2011 ) . h. ramezani , j. schindler , f. m. ellis , u. gnther , and t. kottos , phys . a 85 , 062122 ( 2012 ) . z. lin , j. schindler , f. m. ellis , and t. kottos , phys . a 85 , 050101(r ) ( 2012 ) . j. schindler , z. lin , j.m . lee , h. ramezani , f.m . ellis and t. kottos , j. phys . a : math . theor . 45 , 444029 ( 2012 ) .	the coupled discrete linear and kerr nonlinear schrdinger equations with gain and loss describing transport on dimers with parity - time ( @xmath0 ) symmetric potentials are considered . the model is relevant among others to experiments in optical couplers and proposals on bose - einstein condensates in @xmath0symmetric double - well potentials . it is known that the models are integrable . here , the integrability is exploited further to construct the phase - portraits of the system . a pendulum equation with a linear potential and a constant force for the phase - difference between the fields is obtained , which explains the presence of unbounded solutions above a critical threshold parameter . the behaviour of all solutions of the system , including changes in the topological structure of the phase - plane , is then discussed .
the dynamics of any stellar system may be characterized by the following three dynamical time scales : ( i ) the crossing time @xmath0 , which is the time needed by a star to move across the system ; ( ii ) the two - body relaxation time @xmath1 , which is the time needed by the stellar encounters to redistribute energies , setting up a near - maxwellian velocity distribution ; ( iii ) the evolution time @xmath2 , which is the time during which energy - changing mechanisms operate , stars escape , while the size and profile of the system change . several ( different and precise ) definitions exist for the relaxation time . the most commonly used is the half - mass relaxation time @xmath3 of spitzer ( 1987 , eq . 2 - 62 ) , where the values for the mass - weighted mean square velocity of the stars and the mass density are those evaluated at the half - mass radius of the system ( see meylan & heggie 1997 for a review ) . in the case of globular clusters , @xmath0 @xmath4 , @xmath1 @xmath4 100 , and @xmath2 @xmath4 10 . table 1 displays , for open clusters , globular clusters , and galaxies , some interesting relations between the above three time scales . for open clusters , crossing time @xmath0 and relaxation time @xmath1 are more or less equivalent , both being significantly smaller than the evolution time @xmath2 . this means that most open clusters dissolve within a few gigayears . for galaxies , the relaxation time @xmath1 and the evolution time @xmath2 are more or less equivalent , both being significantly larger than the crossing time @xmath0 . this means that galaxies are not relaxed , i.e. , not dynamically evolved . it is only for globular clusters that all three time scales are significantly different , implying plenty of time for a clear dynamical evolution in these stellar systems , although avoiding quick evaporation altering open clusters . consequently , star clusters open and globular represent interesting classes of dynamical stellar systems in which some dynamical processes take place on time scales shorter than their age , i.e. , shorter than the hubble time , providing us with unique dynamical laboratories for learning about two - body relaxation , mass segregation from equipartition of energy , stellar collisions , stellar mergers , and core collapse . all these dynamical phenomena are related to the internal dynamical evolution only , and would also happen in isolated globular clusters . the external dynamical disturbances tidal stripping by the galactic gravitational field influence equally strongly the dynamical evolution of globular clusters . . dynamical time scales for open clusters , globular clusters and galaxies [ cols="<,^,^,^ , < " , ] mass segregation was one of the early important results to emerge from computer @xmath5-body simulations of star clusters . see , e.g. , von hoerner ( 1960 ) who made the first @xmath5-body calculations with @xmath5 = 4 , 8 , 12 , and 16 bodies . the heavier stars would gradually settle towards the center , increasing their negative binding energy , while the lighter stars would preferentially populate the cluster halo , with reduced binding energy . later , direct integrations using many hundreds of stars showed the same tendency . soon it was also realized that computation of individual stellar motions could be replaced by statistical methods . the same mass segregation was observed in models which integrated the fokker - planck equation for many thousands of stars ( e.g. , spitzer & shull 1975 ) . mass segregation is expected from the basic properties of two - body relaxation . the time scale for dynamical friction to significantly decrease the energy of a massive star of mass @xmath6 is less than the relaxation time scale for lighter stars of mass @xmath7 by a factor @xmath8 ( see , e.g. , eq . 14.65 in saslaw 1985 ) . as massive stars in the outer regions of a cluster lose energy to the lighter ones , they fall toward the center and increase their velocity . the massive stars continue to lose the kinetic energy they gain by falling and continue to fall . the lighter stars , on the other hand , increase their average total energy and move into the halo . as light stars rise through the system , their velocity decreases , altering the local relaxation time for remaining massive stars . will this mass segregation process have an end , i.e. will the system reach an equilibrium ? two conditions would have to be satisfied : mechanical equilibrium determined by the scalar virial theorem : @xmath9 and thermal equilibrium determined by equipartition of energy among components of different mass @xmath10 : @xmath11 all species must have the same temperature , so there is no energy exchange among the different species . from a pure observational point of view , mass segregation has now been observed clearly in quite a few open and globular clusters . these observational constraints are essentially photometric : different families of stars , located in different areas of the color - magnitude diagram ( cmd ) , exhibit different radial cumulative distribution functions . such an effect , when observed between binaries and main sequence stars or between blue stragglers and turn - off stars , is generally interpreted as an indication of mass segregation between subpopulations of stars with different individual masses . we present hereafter examples of observations of mass segregation in three different kinds of star clusters : ( i ) in the very young star cluster r136 , ( ii ) in a few open clusters , and ( iii ) in a few globular clusters . the large magellanic cloud star cluster ngc 2070 is embedded in the 30 doradus nebula , the largest hii region in the local group ( see meylan 1993 for a review ) . the physical size of ngc 2070 , with a diameter @xmath4 40 pc , is typical of old galactic and magellanic globular clusters and is also comparable to the size of its nearest neighbor , the young globular cluster ngc 2100 . with an age of @xmath4 4 @xmath12 10@xmath13 yr ( meylan 1993 , brandl 1996 ) , ngc 2070 appears slightly younger than ngc 2100 which has an age of @xmath4 12 - 16 @xmath12 10@xmath13 yr ( sagar & richtler 1991 ) . brandl ( 1996 ) obtained for r136 , the core of ngc 2070 , near - ir imaging in @xmath14 bands with the eso adaptive optics system adonis at the eso 3.6-m telescope . they go down to @xmath15 = 20 mag with 0.15 resolution over a 12.8 @xmath12 12.8field containing r136 off center . they present photometric data for about 1000 individual stars of o , b , wr spectral types . there are no red giants or supergiants in their field . brandl ( 1996 ) estimate from their total @xmath15 magnitude that the total stellar mass within 20 is equal to 3 @xmath12 10@xmath16 , with an upper limit on this value equal to 1.5 @xmath12 10@xmath17 . a star cluster with a mass of this range and a typical velocity dispersion of @xmath4 5 would be gravitationally bound , a conclusion not immediately applicable to ngc 2070 because of the important mass loss due to stellar evolution experienced by a large number of its stars ( see kennicutt & chu 1988 ) . mass segregation may have been observed in r136 , the core of ngc 2070 . from their luminosity function , brandl ( 1996 ) determine , for stars more massive than 12 , a mean mass - function slope @xmath18 = 1.6 [ @xmath18(salpeter ) = 1.35 ] , but this value increases from @xmath18 = 1.3 in the inner 0.4 pc to @xmath18 = 1.6 for 0.4 pc @xmath19 @xmath20 @xmath19 0.8 pc , and to @xmath18 = 2.2 outside 0.8 pc . the fraction of massive stars is higher in the center of r136 . brandl ( 1996 ) attribute these variations to the presence of mass segregation . given the very young age of this system , which may still be experiencing from violent relaxation , the cause of this mass segregation is not immediately clear . it may be due to a spatially variable initial mass function , a delayed star formation in the core , or the result of dynamical processes that segregated an initially uniform stellar mass distribution . obviously , the older the cluster , the clearer the mass segregation effect . one of the first such clear cases was observed by mathieu & latham ( 1986 ) in m67 which , with an age of about 5 gyr , is one of the oldest galactic open clusters . they studied the radial cumulative distribution functions of the following three families of stars : single stars , binaries , and blue stragglers , the latter being possibly the results of stellar mergers . the radial cumulative distribution functions of binaries and blue stragglers are similar and significantly more concentrated than the distribution function of the single stars . in such a dynamically relaxed stellar system , this result may be explained only by mass segregation between stars of different individual masses . in one of the most recent such studies , raboud & mermilliod ( 1998 ) have observed some clear presence of mass segregation ( see fig . 1 ) in three open clusters ngc 6231 , the pleiades , and praesepe which , with ages equal to 4 , 100 , 800 myr , respectively , are significantly younger than m67 . the presence of mass segregation in the pleiades and praesepe open clusters is expected given the fact that their relaxation times are shorter than their ages . this is not the case for ngc 6231 , where the presence of mass segregation may be as problematic as it is in the case of r136 . because of their very high stellar densities , globular clusters have been hiding for decades any clear observational evidence of mass segregation , expected to be present essentially in their crowded central regions . differences in the radial distributions of stars of different luminosities / masses have finally been definitely observed with hst , providing conclusive observational evidence of mass segregation in the central parts of globular clusters . one of the most serious and detailed such studies is the one by anderson ( 1997 ) who has used hst / foc and hst / wfpc2 data to demonstrate the presence of mass segregation in the cores of three galactic globular clusters : m92 , , and . anderson has first determined the luminosity function of each cluster at two different locations in the core . then he has compared these luminosity functions with those from king - michie multi - mass models , in the cases with and without mass segregation between the different stellar species . 3 displays the comparison between the observed luminosity function ( dots ) and the model predictions with ( continuous lines ) and without ( dashed lines ) mass segregation , at the center of ( left panel ) and at one core radius from the center ( right panel ) . the two different models differ strongly over a large range in magnitude ( 18 - 26 mag ) . the observed luminosity function shows a very clear agreement with the model containing mass segregation , and rules out completely any model without mass segregation ( anderson 1997 ) . the globular cluster m92 displays results very similar to those obtained for . this is not surprising given the fact that both clusters have rather similar structural parameters and concentrations , providing similar central relaxation times of the order of 100 myr . this is not the case for , which is the most massive galactic globular cluster and has a central relaxation time of about 6 gyr . as expected , the two model luminosity functions ( with and without mass segregation ) computed at the center of differ only slightly , and the observed luminosity function is right between the curves of the two models . the two model luminosity functions computed at 16 from the center ( at about 5 core radii ) do not differ significantly and consequently agree similarly with the observations . as expected , , which has had hardly any time to become dynamically relaxed , even in its center , exhibits a very small amount of mass segregation ( anderson 1997 ) . as seen above , the various stellar species of a star cluster must have the same temperature in order to have equipartition of energy . spitzer ( 1969 ) derived a criterion for equipartition between stars of two different masses @xmath21 and @xmath22 . let us consider the analytically tractable case where the total mass of the heavy stars , @xmath23 , is much smaller than the core mass of the system of the lighter stars , @xmath24 , and the individual heavy stars are more massive than the light stars , @xmath22 @xmath25 @xmath21 . in such a case , equipartition will cause the heavy stars ( e.g. , binaries and/or neutron stars ) to form a small subsystem in the center of the core of the system formed by the light ( e.g. , main sequence ) stars . in equipartition , @xmath26 , where @xmath27 represents the central one - dimensional dispersion of the light stars . it can be easily seen ( e.g. , binney & tremaine 1987 ) that equipartition can not be satisfied unless the following inequality holds : @xmath28 where @xmath29 and @xmath30 are dimensionless constants . when @xmath23 become too large , the inequality is violated , there is the `` equipartition instability '' ( spitzer 1969 ) , which has a simple physical explanation : when the mass in heavy stars is too large , these stars form an independent high - temperature self - gravitating system at the center of the core of light stars . in a realistic system with a distribution of stellar masses , the chief effect of the equipartition instability is to produce a dense central core of heavy stars , which contracts independently from the rest of the core . however , as this core becomes denser and denser , the gravothermal instability dominates over the equipartition instability ( antonov 1962 , lynden - bell & wood 1968 ) and the cluster experiences core collapse ( makino 1996 ) . from an internal point of view , the dynamical evolution of star clusters is driven by two - body relaxation , mass segregation , equipartition instability , and core collapse . from an external point of view , the dynamical evolution of star clusters is driven by the dynamical disturbances due to the crossing of the galactic plane , which create tidal tails . in whatever location , these stellar systems are dynamically never at rest . the globular cluster m15 has long been considered as a prototype of the collapsed - core star clusters . high - resolution imaging of the center of m15 has resolved the luminosity cusp into essentially three bright stars . post - refurbishment hubble space telescope star - count data confirm that the 2.2 core radius observed by lauer ( 1991 ) and questioned by yanny ( 1994 ) , is observed neither by guhathakurta ( 1996 ) with hst / wfpc2 data nor by sosin & king ( 1996 , 1997 ) with hst / foc data . this surface - density profile clearly continues to climb steadily within 2 . it is not possible to distinguish at present between a pure power - law profile and a very small core ( sosin & king 1996 , 1997 ) . consequently , among the galactic globular clusters , m15 displays one of the best cases of clusters caught in a state of deep core collapse . sosin & king ( 1997 ) have estimated the amount of mass segregation in the core of m15 from their hst / foc data : the mass functions at 20 and 5 from the center clearly show substantial mass segregation for all stars with masses between 0.55 and 0.80 . -0.5truecm @xmath31 the mf at @xmath20 = 20 is best fit by a power - law with slope @xmath18 = @xmath32 0.26 , -0.5truecm @xmath31 the mf at @xmath20 = 5 is best fit by a power - law with slope @xmath18 = @xmath33 0.25 . these two slopes differ at the 5-@xmath27 level . once compared with models , the amount of mass segregation is somewhat less than predicted by a king - michie model , and somewhat greater than predicted by a fokker - planck model . see also king ( 1998 ) in the case ngc 6397 . mass segregation is also present in kinematical data , i.e. , in the radial velocities and proper motions of individual stars . so far , radial velocities have been obtained essentially only for the brightest stars , giants and subgiants , which have very similar masses . it is only recently that internal proper motions of individual stars have been obtained in globular clusters . the following team ( pi . g. meylan , with cois . d. minniti , c. pryor , e.s . phinney , b. sams , c.g . tinney , joined later by j. anderson , i.r . king , and w. van altena ) have acquired hst / wfpc2 images in the core of in three different epochs ( oct . 1995 - nov . 1997 - oct . 1999 ) defining a total time baseline of 4 years . the choice of the @xmath34 = fw300 filter prevents saturation for the brightest stars and allows simultaneous measurement of proper motions for the brightest stars as well as for stars more than two magnitudes bellow the turn - off ( meylan 1996 ) . for each epoch we have 15 images with careful dithering . each measurement of the position of a star has a different bias since in each pointing the star is measured at a different pixel phase . we use an iterative process on positions and local psf determinations . we achieve a position accuracy of 0.020 pixel for a single image , amounting to 0.006 pixel for the mean of 15 images . this corresponds to 0.3 mas in the pc frame and 0.6 mas in the wf2 , wf3 , and wf4 frames , for about 14,000 stars in the core of ( anderson & king in preparation ) . preliminary results show a clear difference between the proper motions of blue stragglers and stars of similar magnitudes : the former are significantly slower than the latter . since blue stragglers are either binaries or mergers , with masses higher than the turn - off mass , the above difference unveils the first kinematical observation of mass segregation in a globular cluster ( meylan in preparation ) . anderson j. , 1997 , phd thesis , university of california , berkeley antonov v.a . , 1962 , vest . 7 , 135 ; english translation : antonov , v.a . , 1985 , in dynamics of star clusters , iau symp . 113 , eds . goodman j. & hut p. , ( dordrecht : reidel ) , p. 525 binney j. , tremaine s. , 1987 , galactic dynamics , ( princeton : princeton university press ) brandl b. , sams b.j . , bertoldi f. , , 1996 , apj , 466 , 254 guhathakurta p. , yanny b. , schneider d.p . , bahcall j.n . , 1996 , aj , 111 , 267 kennicutt r.c . , chu y .- h . , 1988 , aj , 95 , 720 king i.r . , anderson j. , cool a.m. , piotto g. , 1998 , apj , 492 , l37 lauer t.r . , holtzman j.a . , faber s.m . , , 1991 , apj , 369 , l45 lynden - bell d. , wood r. , 1968 , mnras , 138 , 495 makino j. , 1996 , apj , 471 , 796 meylan g. , 1993 , in the globular cluster - galaxy connection , asp conference series vol . smith g.h . & brodie j.p . , ( san francisco : asp ) , p. 588 meylan g. , heggie d.c . , 1997 , a&ar , 8 , 1 - 143 meylan g. , minniti d. , pryor c. , tinney c. , phinney e.s . , sams b. , 1996 , in the eso / stsci workshop on _ science with the hubble space telescope - ii _ , p. benvenuti , f.d . macchetto , e.j . schreier , ( baltimore : stsci ) , p. 316 raboud d. , mermilliod j .- c . , 1998 , a&a , 333 , 897 sagar r. , richtler t. , 1991 , a&a , 250 , 324 saslaw w.c . , 1985 , gravitational physics of stellar and galactic systems , ( cambridge : cambridge university press ) sosin c. , king i.r . , 1996 , in dynamical evolution of star clusters : confrontation of theory and observations , iau symp . 174 , eds . hut p. & makino j. ( dordrecht : kluwer ) , p. 343 sosin c. , king i.r . , 1997 , aj , 113 , 1328 spitzer l. , 1969 , apj , 158 , l139 spitzer l. , 1987 , dynamical evolution of globular clusters , ( princeton : princeton university press ) spitzer l. jr . , shull j.m . , 1975 , apj , 201 , 773 von hoerner s. , 1960 , z. f. a. , 50 , 184 yanny b. , guhathakurta p. , bahcall j.n . , schneider d.p . , 1994 , aj , 107 , 1745 * comment by s. portegies zwart * from a theoretical point of view , it is not always clear what observers considered as the center of a star cluster and what theorists should use as the center . one can use , for example , the geometric center , the area with the highest luminosity density , number density , mass density . * comment by g.m . * from an observational point of view , the determination of the center of a globular cluster is also difficult and uncertain . ideally , the algorithm used should determine the barycenter of the stars , not of the light . in the case of a collapsed globular cluster like m15 , which has a very small unresolved core , the task is difficult because of the very small number of stars detectable in such a small area . the uncertainty in the position of the center is of the order of the core radius value , i.e. , about 0.2 . in the case of a globular cluster like , various methods give positions differing by 2 - 3 . such a large uncertainty is nevertheless acceptable , given the core radius value of about 25 . * question by h. zinnecker * is the mass segregation observed in the 30 doradus cluster due to dynamical evolution or due to preferential birth of the more massive stars near the cluster center ? can we distinguish between these two possibilities ? * answer by g.m . * it is not known if the mass segregation is the consequence of dynamical evolution or of a flatter imf in the center . i fail to see any reliable way to distinguish between these two possibilities . * question by p. kroupa * the globular clusters and do not appear to have a pronounced binary sequence in color - magnitude diagrams , whereas other globulars have pronounced binary sequences . does this imply different dynamical histories ? * answer by g.m . * it would be interesting to compare the locations where these various color - magnitude diagrams have been obtained . in the case of , the excellent photometry we obtained is for stars right in the center , where encounters and collisions operate and probably decrease the fraction of binaries . it would be interesting to make a precise comparative study , for a few globular clusters , for which we would have data from the same instrument and reduced with the same software , in fields at the same relative distance from the center . * question by d. calzetti * about studies which find flatter imfs in the centers of clusters : do you think that these studies may suffer from effects of crowding towards the cluster center , and therefore , find a flatter imf because of this ? * answer by g.m . * yes , definitely ! crowding is always present in photometry of globular clusters , especially in observations from the ground . nevertheless , i think that some careful photometric studies using hst data have provided reliable results in relation to imf slope ( see , e.g. , king , 1998 , apj , 492 , l37 ) . it would be an interesting study , which , as far as i know , has not been done yet . it is partly due to the difficulties in determining precisely the center of the core and to the difficulties in determining precisely outer isophotes which suffer from very low star counts and are strongly polluted by foreground stars and background galaxies .	star clusters open and globulars experience dynamical evolution on time scales shorter than their age . consequently , open and globular clusters provide us with unique dynamical laboratories for learning about two - body relaxation , mass segregation from equipartition of energy , and core collapse . we review briefly , in the framework of star clusters , some elements related to the theoretical expectation of mass segregation , the results from n - body and other computer simulations , as well as the now substantial clear observational evidence . # 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
let us consider fundamental hyperbolas along @xmath13 and @xmath14 in minkowski s space - time ( at one space dimension with the axis , @xmath15 @xmath16 of system k with light velocity @xmath17 . the two hyperbolas determining the units of measure ( @xmath18 are called hyperbolas of scale or calibration ( @xcite ) . we focus the attention on the along @xmath14 hyperbola defined with invariance of _ space interval _ by a _ passive _ lorentz transformation ( @xmath19 ) ( 5 ) @xmath20 @xmath21 ( * fig1 * , light asymptotes and standard representation of primed axis @xmath22 @xmath23 in scissors or _ hyperbolic rotation _ of system k ) @xmath24 given that @xmath25 ( the hyperbolic radius of curvature ) can be as large as we wish , the calibration hyperbolas disappear at the infinity and we have only one invariant @xmath26 in standard configuration . let us however note that , only the _ finite _ interval involves , according to minkowski , that _ space by itself and time by itself are doomed to fade away into mere shadows , and only a kind of union of the two will preserve an independent reality_. indeed an infinite interval ( @xmath27 @xmath28 @xmath29 , see 21 ) should mean that the independent space is given for any @xmath16 and therefore the return of the shadow ( absolute space @xmath14 , @xmath30 ) . in order to stay in relativistic configuration suppose now that we have a very small but non nul constant @xmath31 in such a way that @xmath32 such a reformed minkowski s metric no longer is flat because we have an hyperbolic global curvature @xmath33 in this case we have a constant radius @xmath34 of the universe and therefore a static universe incompatible with the observed universe in dilation . however , by the same argument with the other hyperbola ( @xmath13 ) , we could claim that we should have an observed dilated time ( in k , * fig1 * ) of universe ( see 3.4 ) . let us now underline that the right branch @xmath35 of @xmath14 hyperbola @xmath36 can also represent a _ worldline _ of an uniformly ( @xmath37 accelerated particle @xmath38 according to rindler ( * fig2 * , @xcite ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ consider a rod of arbitrary length resting along the x axis of minkowski space . a time @xmath39 we wish to give one point of the rod a certain constant proper acceleration and we want the rod as a whole to move rigidly , i.e. in such a way that the proper length of each its infinitesimal elements is preserved . it turns out that each point of the rod must then move with a different though also constant proper acceleration , the necessary acceleration increasing in the negative direction and become infinite at a well - defined point of the rod ; the rod can evidently not to be extended beyond or even quite up to that point , since an infinite proper acceleration corresponds to motion at the speed of light . if we arrange things so that this cutoff point lies originally at the origin the equation of motion of the point originally at @xmath40 @xmath41is ( 2 ) we take x as a convenient spatial coordinate on the rod . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ @xmath42 ( end of quotation ) . born s rigid motion suppose a set of hyperbolic worldlines of points belonging to a rigid rod @xmath43 this basic motion with constant interval , is induced by successive @xmath44 ( active lorentz transformation , 2.2 ) which boosts the particle fixed in @xmath45 ( the rigid rod ) to a larger velocity . given that in standard configuration the proper acceleration @xmath46 can be as small as we wish @xmath47 and so the hyperbola disappears at the infinity : it only remains the light cone that defines the standard minkowski s metric with only one invariant with @xmath48 ( 14)@xmath49 the geodesics are _ straight lines _ in _ flat standard _ minkowski s metric however in special relativity ( 3 - 2 ) the acceleration is a _ spacelike _ four - vector whose norm must necessarily be larger than zero @xmath50 . suppose now the existence of a minimal acceleration @xmath8 ( 3 ) which corresponds to the relativistic inexistence of an infinite interval ( 2 ) . @xmath51 where @xmath52 is the observable radius of hubble and @xmath8 is the observable acceleration of milgrom ( @xcite ) . we rediscover the empirical standard values of constants @xmath53.@xmath54 , @xmath55 @xmath56 @xcite)@xmath57 the cosmological constant ( @xmath58 , standard value defined by static 1917 einstein s model ) is a second invariant in our new sr which connects two previously independent empirical quantities @xmath59 and @xmath8 . this new completed sr that we suggest to call _ cosmological ( special ) relativity _ ( * cr * ) introduces in physics a _ scalar _ field of acceleration _ _ _ _ @xmath8 that is non - gravitionnal because the fundamental acceleration depends on the distance and not on the square of the distance . the constant @xmath2 can be see either from the left side or the right side of equations ( 4 ) . in the first case ( left side ) we have a geometrical interpretation : the geodesic is no longer a straight space - time world line but an hyperbolic spacetime worldline that can be connected with hyperbolic trajectories of pioneer sounds ( @xcite ) . exactly as in euclidean geometry @xmath60 represents the global constant curvature of circle , @xmath61 @xmath62 represents in minkowskian geometry the _ global _ constant curvature of the hyperbola . but the difference is that the radius @xmath63 of the circle can be as great as we wish @xmath64 ( globally the curvature is zero ) the maximal interval - distance is @xmath52 in cr ( globally the curvature is non - zero ) the basic heuristic in cr is based on the relativistic transformation of circle into hyperbola ( that is well known , * fig1 * ) and therefore also on the relativistic transformation of non - relativistic circular uniformly accelerated _ motion _ into hyperbolic rectilinear uniformly accelerated motion ( that is not very well known , * fig3 * ) . this principle of correspondence circle - hyperbola * ( pcch ) * plays a basic role in our analysis ( 5 figures ) . in the second case ( right side ) we have not only a cosmological interpretation of milgrom s acceleration @xmath65 ( @xcite ) but also , very curiously , hubble physical constant @xmath66 @xmath67 . the presence of this constant of expansion in ( 4 ) and therefore in ( 3 & 2 ) is very strange because minkowski s space - time is renowned incompatible with expanding universe . the fact that the hubble radius is a constant ( @xmath68 is not in contradiction with the observed expansion of the universe because we will prove that this radius is an * hyperbolic horizon * which involves the law of hubble ( 37 ) . however there is a serious problem in @xmath69 because our theory is a purely logical building based on the identification ( fig1 & fig2 ) between minkowski s hyperbola ( @xmath19 without explicit acceleration ) and born s hyperbola ( @xmath44 , with an acceleration ) . this identification first involves the coincidence with @xmath70 and @xmath71 , center of rotation in both cases in cr ( 3 ) does not correspond to the most sophisticated model of rigid motion ( 2 , according to rindler the end ( @xmath71 ) of rod is a photon ) . rindler s representation of einstein s contraction involves that @xmath71 moves on the asymptote ( its own infinite hyperbola ) with respect to @xmath70 . this is the reason why the two hyperbolas are generally not identified . our identification involves a physical synthesis , based on * hyperbolic rotation motion * , between @xmath19 and @xmath44 and therefore the determination of a * fundamental acceleration in lorentz boost * ( 4 ) . einstein introduced his theory of contraction of length in 1905 ( second paragraph ) by using rigid rods ( @xcite ) . born defined in 1909 the _ rigid motion _ of each points of einstein s rigid rod , by using accelerated hyperbolic motion of a particle ( @xcite ) . rindler found in 1966 an original metric in cosmology on the basis of einstein - born s rigid motion in special relativity ( @xcite ) . einstein considers two systems ( rigid rods ) , @xmath72 and @xmath45 , in standard configuration . lorentz transformation ( lt ) , at one space dimension is : @xmath73 with instantaneous coincidence of @xmath74 at @xmath75 and space - time @xmath26 units @xmath76 , @xmath77 . in einstein s paragraph 2 ( @xcite ) , the system @xmath78 is in uniform translation with respect to the system @xmath72 for _ any _ @xmath16 ( @xmath79 and @xmath80 einstein uses _ a passive lt _ in order to define the length in k of a moving rigid rod @xmath81 of @xmath45 with the simultaneity of two events @xmath82 in k ( the difference of space coordinate of two ends @xmath71 and @xmath83 of the rigid rod at the same time in @xmath72 ) . by passive using of first @xmath84 ( 5a ) , we find the coordinates in @xmath72 of @xmath71 and @xmath83 , respectively for finite ( 6a ) and differential length ( 6b ) : @xmath85 we note that einstein s contraction suppose @xmath39 and not @xmath86 . let us summarize briefly born s procedure . in sr , the acceleration is not an invariant ( @xmath87 and @xmath88 and only a _ proper _ constant acceleration in k @xmath89 @xmath87 can be defined at the condition to do @xmath90 in the components of acceleration four - vector @xmath91 @xmath92 @xmath93 @xmath94 @xmath95 ( at one space dimension ) : @xmath96 the comoving system with the particle @xmath83 is thus at each instant a new system , @xmath45 , defined by successive @xmath44 with @xmath97 but @xmath98 ( @xmath99 ) and with @xmath100 but @xmath101 . in each k system , the equation of the @xmath102 axis is @xmath86 . the origin @xmath103 of successive @xmath45 is generally located in @xmath83(born ) ; we will adopt the standard configuration.(rindler ) , by locating the particle @xmath83 at a distance @xmath40 ( born - rindler s notations,@xcite ) from @xmath71 in k at the coordinate @xmath104 in @xmath72 ( @xmath105 . the differential equation @xmath106 is @xmath107 @xmath108 . a first integration is carried out with @xmath109 at @xmath110 : @xmath111 or : @xmath112 a second integration yields @xmath113 with @xmath114 @xmath115.and therefore the right branch of hyperbola : @xmath116 the accelerated hyperbolic motion is resolved by successive @xmath44 into an infinity of instantaneous inertial motions * ( fig2). this is born s definition of _ rigid motion _ of the rod : at each instant @xmath117 , the entire rod ( and the infinitesimal rigid rod @xmath118 ) is at rest in its proper system k(in born - rindler s notations @xmath119 . it is then impossible to separate the motion of the particle @xmath120 from the motion of the system @xmath45(the rigid rod @xmath81 ) that drags it along . in other words , it is impossible to separate successive @xmath44 ( succession of events ) from successive boosts of systems @xmath45 . ebr s rigid motion is a relativistic interpretation of _ rectilinear non - relativistic accelerated _ motion which involves that the rod @xmath121 is in motion with respect to the origin @xmath70 . so in rindler s diagram @xmath122 the end of the rod @xmath71 for @xmath123 is in motion on its own infinite hyperbola ( fig2 * ) , i.e. the asymptote.we underline that ebr s rigid motion is not correlated ( pcch , 1 ) with a rotation because in any rotation ( including hyperbolic rotation ) we have a center or a fixed point ( @xmath124 . this is the reason why rindler adopts a basic equation which result from differentiating for constant @xmath125 @xmath126 introducing einstein s contraction in rigid motion . @xmath127 k. the problem in einstein - born s theory is that the entire rod @xmath121 is defined at constant @xmath86 ( simultaneity ) and einstein s contraction supposes also at constant @xmath128 ( also simultaneity , rindler s representation with @xmath129 * fig2 * ) . the quadratic form ( 6a ) is not the quadratic form ( 9a ) : there is an asymmetry ( * fig1 * ) . we underline that in born s theory the infinitesimal proper length @xmath118 ( @xmath130 ) appears as an invariant exactly as minkowski s proper time @xmath131 ( @xmath132 ) appears as an invariant ( see note 3 ) . but in standard sr this analogy must be limited because if the space interval defines a distance exactly as the time interval defines a duration , we obtain a dilation ( 9b ) in both cases . rindler gives in 1966 a cosmological interpretation of born s theory . he focuses at first the attention on the fact that there is in each point of the rod a a different though also constant proper acceleration but also on a singularity because the rigid rod ( the proper length ) ends in a photon ( infinite acceleration @xmath133 ) . in order to develop analogy with schwarschild metric , rindler then introduces a parameter @xmath134 for observers with clocks on the rod ( given that @xmath117 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it can be seen that the proper acceleration of the point @xmath135 of the rode is @xmath136 . hence an observer at @xmath135 feels a constant gravitational field of intensity @xmath137 the observers on the rod can so synchronize their clocks that each sees all the other clocks neither gain nor lose relative to his own ; each observer must simply speed - up his proper clock by a factor equal to the reciprocal of his coordinate x. let @xmath134 denote this new time . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ with rindler s time parameter @xmath138 the hyperbolic velocity and not the proper time @xmath139 ) , one obtains a non - minkowskian metric @xmath140 ellis and williams underline that rindler s metric ( @xmath141 @xmath142 is based on the boost - invariance of the proper length of the rod or boost - invariance of flat space - time(@xcite)@xmath57 the fact that the boost - invariance involves a non - minkowskian space - time is not consistent with the fact that the boosts in question are ... lorentz boosts . we need therefore a complete theory of @xmath44 or lorentz boosts ( _ hyperbolic rotations _ ) . rindler finally does not use the parameter @xmath143 because the proper time @xmath144 @xmath145 , in a state of acceleration of @xmath146 systems , is a different one in each point of the rod in prerelativistic kinematics at rest in the system k. if we push in @xmath86 the rod with a finite constant acceleration in o , instantaneously p is in moving with the same finite constant acceleration . the two ends @xmath71 and @xmath83 of the rods are instantaneously boosted at the velocity @xmath147 and therefore at the same rindler s time @xmath148 this instantaneity at distance however can not be true in a completely relativistic kinematics.@xmath149@xmath150 ebr refers to einstein s contraction ( @xmath19 , his paragraph 2 ) but not directly to einstein s boost ( his paragraph 3 ) . in other words ebr s demi - hyperbola * ( fig2 ) * is not concentrated on the transition from @xmath151-velocity to @xmath147-velocity ( @xmath152 * * ( fig3)*@xmath57 this is a natural interpretation of born s integral hyperbola which is deduced from a differential equation ( 8a ) @xmath107 @xmath153 ( unlike minkowski s hyperbola ) . @xmath41we now focus the attention on the element of _ arc of hyperbola _ @xmath154 ( fig3 ) * , i.e. on einstein s boost . einstein considers two systems k and k , at rest relatively with each other . he places a rigid rod in k , of length @xmath155 , and an identical rigid rod in k , of identical length @xmath156 when both systems are at rest . einstein then boosts the k system at @xmath75 , to bring it to a cruise speed of @xmath147 ; this calls for an acceleration . einstein explicitly states ( in 1907 ) that the eventual consequences of this acceleration @xmath157 on the @xmath129 rod disappear in k as soon as the velocity becomes uniform ( a statement we do not question ) . einstein however does not say anything about this acceleration and how to compute this acceleration @xmath46 . let us now develop into _ three successive phases _ ( * fig3 ) * a theory of acceleration in einstein s boost with @xmath44 @xmath158 . phase 1 the rod first is at rest @xmath159 for @xmath79 phase 2 then in @xmath75 the rod is submitted in each point to a constant acceleration @xmath46 ( _ arc of hyperbolic rotation _ @xmath154 ) . the element of angle that subtends the element of hyperbolic arc is hyperbolic velocity @xmath160 phase 3 finally at @xmath161 the rod reaches its cruise speed of @xmath147 ( _ tangent _ in @xmath83 , @xmath162 is the intersection with asymptote)@xmath57 the end of the rod , the particle @xmath83 ( the end of the rod ) is transported by an active lt from the point @xmath163 @xmath164 @xmath165 to the point @xmath166 @xmath167 in k i.e. @xmath168 given that , in hyperbolic motion the particle @xmath83 and its proper comoving system @xmath169 are completely inseparable , @xmath170is therefore the acceleration in any einstein s boost @xmath171 @xmath147 . our new theory of einstein s boost of the rod @xmath81 is exactly born s hyperbolic motion of particle @xmath120 if and only if we admit that @xmath71 _ _ coincide with _ _ @xmath70 _ _ _ _ at @xmath172 _ _ for any velocity _ _ : @xmath173 * * ( * * this is not true with rindler s representation of contraction , see the contrast * * fig1&2)*@xmath57 our theory of einstein s boost suppose that @xmath44 must be consistent with @xmath174 . hyperbolic rotation with @xmath86 is:@xmath175 @xmath176 in order to obtain this fundamental coherence ( fig1 * & * fig3 * ) , we have to admit that @xmath71(@xmath177 ( with @xmath133 ) reaches instantaneously the @xmath178or @xmath179)@xmath180velocity ( the center of hyperbolic rotation is the same ) : @xmath181 so the acceleration * * * * is clearly translated in hyperbolic motion by an * elongation * ( 13a ) or extension of the space itself from the fixed point @xmath71(@xmath177 into @xmath182 in k : everything happens as if there was a fundamental * elastic * elongation @xmath183 of distance @xmath135 . we suggest to call this motion _ expanding motion_. given that , in non - relativistic rectilinear accelerated motion , the velocity is intrinsically non - constant , we admit a fundamental contrast ( at one space dimension ! ) between a relativistic _ hyperbolic elastic _ motion ( @xmath184 and a non - relativistic _ circular _ motion ( @xmath185 because we have in both cases ( pcch ) a constant velocity ( respectively @xmath186 and c ) . in both cases ( * fig3 * ) we have also an acceleration in the direction of the center @xmath187 : the _ center _ of the circle ( 13b ) with constant distance ( euclidean orthogonality acceleration - velocity , 3 - 2 ) or the _ center _ of hyperbola ( 13b ) with constant interval ( minkowskian orthogonality with velocity ) . in both cases o does not move ( in phase 2 ) with respect to o ( unlike ebr , * fig2 * ) . but if the acceleration in euclidean rotation motion and hyperbolic rotation motion have the same direction , they have not the same sense : the former is centripetal and the latter is * centrifugal * ( * fig3 * ) . so the basic expansion is inscribed in lt . indeed the principle of correspondence circle - hyperbola ( pcch ) involves that the physical hyperbolic motion , induced by @xmath44 , must be * hyperbolic rotation * ( lt ) , in geometrical meaning . elastic motion is not compatible with einstein s contraction in rigid motion and this is the reason why we have to give up einstein s definition of distance . in order to do that we have to define a passive use of lt that leads to a dilation of distance ( 13a ) . we will prove that the best way to precise born s boost - invariance of the distance is to bring invariance of _ proper distance _ into line with invariance of _ proper time _ ( new symmetry in sr , 3 - 4 ) . the essential physical data of the material point in pre - relativistic kinematics are given by the 3-vectors position @xmath188 , velocity @xmath189 and acceleration * @xmath190. in relativistic kinematics , these data are replaced by the space - temporal 4-vectors @xmath191 , @xmath192 and @xmath193 . the importance of the transition from 3-vector @xmath190 * ( pre - relativistic , a galilean invariant ) to that of a 4-vector @xmath193 is completely underestimated . the essential characteristic of @xmath193 is that it is a spacelike 4-vector , orthogonal , in the minkowskian meaning , to the velocity 4-vector @xmath194 , which is time like : @xmath195 we thus have the norms @xmath196 and @xmath197 @xmath28 @xmath198 , that must be _ different from 0 ( 14 ) , _ the zero norm being characteristic of a light - type 4-vector . in the general case of three space dimensions , the components of @xmath199 , @xmath200 @xmath93 @xmath201 @xmath202 describe the motion of a particle@xmath203 in k , which is also a motion in k. let us restrict the 3-space dimensions problem to that of a single space dimension , in which case the space - time 4-vectors become space - time 2-vectors @xmath204 , @xmath205 , @xmath206 . if the particle @xmath83 is now a rest ( fixed ) in @xmath146 ( @xmath207 , @xmath86 ) spacelike 4-vector becomes 2-vector acceleration @xmath208 @xmath209 and invariant norm of acceleration becomes proper acceleration : @xmath210 @xmath211 in * fig3 * the direction of acceleration along @xmath102 must be permanently orthogonal ( in minkowski s meaning ) to the direction of velocity . the norm @xmath212 of _ spacelike _ @xmath213 the proper acceleration in any lorentz boost , can not be equaled to zero ( definition of lightlike 4-vector ) . we now examine the phase of acceleration in order to deduce firstly in k the duration of acceleration @xmath214 and , from the elongated distance , the distance of acceleration @xmath215 @xmath216 with kinematics form @xmath217 ( taylor ) which corresponds to _ relativistic kinetic energy _ @xmath218 of the particle @xmath83 . we obtain a different duration of acceleration for any point @xmath135 of the rod ( @xmath219 @xmath41we now calculate , by a simple integration , this duration _ in proper time _ of the lorentz boost because we have @xmath86 but @xmath220 with @xmath221 and with ( 8) ( in born - rindler s notation we have in our theory for @xmath14 hyperbola : @xmath222 , note 3 ) @xmath223 the transition between @xmath151 and @xmath147 velocity will be shorter @xmath224 ( @xmath225 in proper time in @xmath45 than in time of @xmath72 : @xmath226 by using _ hyperbolic velocity _ @xmath227 ( the well known parameter of rapidity ) we rediscover the hyperbolic coordinates of @xmath14 hyperbola @xmath21 @xmath228 the proper time @xmath229 in order to reach the cruise velocity is a different one in each point of the rod in the phase of acceleration ( phase 2 ) . we have therefore the joint between the second phase and the third phase , the second limit of integration ( because @xmath86 but @xmath230 ) : @xmath231 we rediscover the instantaneity @xmath232 for one end @xmath233 of the rod and the maximal proper time @xmath229 for the other end of the rod . we have the following coupling for any hyperbolic rotation ( hr ) or any lt @xmath234 given that @xmath235 . unlike rindler ( 11 , note 2 ) we admit that the proper time , defined @xmath143 by ( 19 ) is the fundamental parameter on the rod . let us remark that in proper time basic equation is of first order whose solution is an exponential ( 33 ) at the condition that @xmath236 so with our parameter @xmath143 we obtain naturally a minkowskian metric @xmath237 @xmath238 in phase2 . @xmath239 however we know that there is an hidden boost @xmath46 behind the arc of hyperbola in minkowski s metric with @xmath44 ( * fig3 * ) . this unexpected result is geometrically obvious because if you take a point at rest and you apply a series of @xmath240 between two limits of integration ( basic equation 19 ) the successive points will be placed not a straight line but ... on an arc of hyperbola . let us note that in ssr the infinitesimal interval is undeterminate if @xmath241 ( minkowski s standard metric is a non - calibrated metric and need a course of treatment ( 34 ) of renormalization ( note 1 ) ! we come back now to minkowski s hyperbolas ( beginning of this paper ) . we focus the attention on the demi - branch ( * fig1 * ) of hyperbola @xmath242 @xmath80 the distance @xmath243 is defined by the interval between two simultaneous events in @xmath244 @xmath245 obviously @xmath243 in k is no longer a distance , in the usual meaning , but an interval . however , if we consider that a difference of space coordinates @xmath246 in k defines a distance , then such a distance @xmath247 @xmath248 is defined between two _ non - simultaneous events _ in k ( gap of simultaneity : @xmath249 . this is very interesting for astrophysical and cosmological distances 3 - 5 ) . it is clearly a non - einsteinian definition of ( the dilation of ) distance . however this new definition is induced from a perfect _ symmetry in _ @xmath19 _ _ or _ _ in minkowski s hyperbolas when we replace events at the same place by events at the same time in k and duration by distance in @xmath250 . the duration @xmath251 between two events not at the same place or the distance @xmath243 between two events not at the same time is dilated in k ( @xcite ) . both @xmath252 and @xmath253 induce a dilation of distance . minkowski s and born s arc of hyperbola ( _ if _ @xmath254 _ _ ) _ _ , _ _ are completely identical and so the couple ( _ _ @xmath255 , @xmath256 _ _ _ _ @xmath257 has to be a fundamental relativistic couple . _ _ _ _ unlike prerelativistic kinematics , in which there is no fundamental relationship between @xmath258 and @xmath188 @xmath49 there is a relativistic four - vectorial relationship ( ) @xmath259 we verify with @xmath260 and @xmath261 , @xmath262 , @xmath263 the factor onehalf @xmath264 ( @xmath265 @xmath266 disappears in a relativistic theory ( @xmath267 . according to rindler the rod ends in a photon . but in born - rindler s theory a photon dispatched to chase the particle at @xmath39 from @xmath70 never catch up with it because the acceleration remains constant . in our new theory of einstein s boost , the acceleration stops in the third phase and there is an intersection @xmath162 between the tangent and the wordline of the photon . in cr , einstein s boost is structurally an emitter of radiation and this emission is not independent of the definition of distance . in the @xmath45inertial system , let us first consider einstein s rigid rod @xmath268 at rest in k(of length l ) with a source at @xmath233 and a mirror at @xmath162 . a light signal is emitted from @xmath233 at @xmath86 , it is reflected in @xmath269 at @xmath162 and returns to @xmath71 at @xmath270 , the time out@xmath271 being equal to the @xmath272 . einstein s clock synchronization uses three successive physical events,@xmath273 : @xmath274 @xmath275 , @xmath276 . he has synchronized the two clocks at the ends of the rod and defined the simultaneity of two events at a distance . these two simultaneous events @xmath164 @xmath277 and @xmath278 @xmath279 , respectively at @xmath280 and @xmath281 are however not explicitly given in einstein s paper ( @xmath143 represent here minkowski s standard proper time ) . the same system is used , but without einstein s rigid rod ( given a priori ) and with a single clock in @xmath71 . we will suppose that the mirror @xmath282 is at rest in @xmath45 ( a distant reflecting object ) at an unknown distance . a light signal is emitted at @xmath71 at @xmath117 , reflected in @xmath162 in @xmath251 ( duration ) and returns to @xmath283 at @xmath284 . the proper , distance @xmath243 is defined by two simultaneous events in k : @xmath285 the synchronous distance @xmath243 may be measured by a single clock in @xmath233 ( @xmath286 ) with a round trip signal . such a distance , measured by the light time of travel , is not a new technique . we can replace einstein s rigid rod by one half light travel time ( _ two ways or round trip _ @xmath287 ) distance with _ bondi s radar method _ ( @xcite ) . we add a new element : the converse ( poincar ) interpretation of einstein s synchronization involves also that the so formed distance @xmath243 is transformed by lt @xmath248 like the duration @xmath288 . in other words , @xmath243 is a distance that is defined as the proper duration , and is thus the * shortest * in k. as a last analysis and from a historical viewpoint , einstein s work is based on the direct theorem ( the @xmath289 axis ) , while poincar s opened the way to the reciprocal ( the @xmath290 axis or @xmath86 ) . _ this lorentz hypothesis is the immediate translation of michelson s experiment , if the lengths are defined by the time that light takes to travel through them _ ( @xcite , @xcite & @xcite ) . in the same way that simultaneity at a distance can not be defined independently from the velocity of light , the distance itself may not be defined independently from this velocity . this _ proper _ distance @xmath243 ( events at the same time @xmath86 in k ) then becomes an _ invariant _ of lt ( @xmath291 in the same sense as the proper duration @xmath251 ( events at the same place @xmath292 in k ) is an invariant of lt ( @xmath293 . _ in summary poincar s proper length ( hyperbola along _ @xmath294 _ _ ) is the exact symmetric of minkowski s proper time ( hyperbola along ( ot ) ; except the one - half _ _ @xmath264 _ _ factor ( two - ways ) in ( 25 ) . _ _ we remark now that the proportionality duration - distance , according to poincar s basic elongated ellipse@xcite , is also valid for one way trip ( without factor @xmath295 . we have in k because when the phase of acceleration stops ( phase 3 ) , o is in motion with respect to o : @xmath296 so the factor defining the distance by the travel time back is no longer @xmath297 @xmath298 @xmath299 but famous bondi s factor @xmath9 @xmath300 @xmath301 both factors are lorentz factor @xmath302 ( @xmath303 . however in bondi s theory , this factor is introduced in the transformation of the lengthwave and not the transformation of length itself : it is a doppler factor . given that , any one - way length is determined by the one - way light travel time , the transformation of _ length _ and _ wavelength _ are the same . in order to define the distance we must have a source in p. suppose a monochromatic source of lengthwave @xmath304 in @xmath305 . observed at o in @xmath72 we have the lengthwave @xmath306 @xmath307 we rediscover with poincar s dilation of distance the same doppler formula as einstein s one ( longitudinally , the direct calculus with lt is very easy , @xcite ) . but with the theory of lorentz boost we have structurally a redshift ( @xmath308 and not @xmath309 . we suppose here that the acceleration stops ( phase 3 ) and so we have ( the derivation with respect of the proper time @xmath143 is primed ) @xmath310 so in the framework of standard sr bondi s factor is not an expansion factor ( a scale factor ) because it does not depend of proper time ( see cr ) . finally , we remark also that the one - way outgoing signal is still not used until now . now we are able to transform sr into cr ( 1 ) . until now we have a fundamental relationship between acceleration in lorentz boost ( @xmath240 ) and distance ( @xmath253 ) defined by interval in minkowski s space - time ( 23 ) . the cause of this acceleration is not inscribed in the structure of space - time and when the cause ( the force ) no longer acts ( m ) , the acceleration stops @xmath311 and we rediscover the standard sr with uniform cruise velocity or the straight line geodesic . in vacuum flat space - time we can not place on the same plane @xmath312 @xmath313 the @xmath44 and the @xmath19 . without a fundamental acceleration @xmath314 there is no lorentz boost , no active lt , no hyperbola , and finally no complete sr . without hyperbola , it remains the light cone , with only one invariant in standard sr . if we wish take into account the hyperbola , we have to consider the existence of a minimal ( @xcite ) acceleration in the vacuum . @xmath315 the existence of @xmath316 ( 24 ) involves the existence of a fundamental lorentz boost structurally connected with an emission of radiation . given that distance @xmath243 is measured in lightyear , if we send a light signal towards a galaxy we can not wait for the return of the light . our new definition of _ proper distance _ @xmath243 supposes the existence of a _ one way ( without one half ) _ @xmath317 light travel time distance , i.e. a _ non - infinite including unit _ in hyperbola ( 2 ) . the radius of hubble @xmath52 is precisely defined by the one - way travel lifetime @xmath318 of the universe or time of hubble @xmath319 ( emission @xmath320 cosmological background radiation , or photons structurally boosted fifteen billions years ago(_fundamental acceleration involves fundamental emission of cbr _ ) . with poincar s distance ( 25 ) we have : ( @xmath39 , emission ) @xmath321 bondi s mirror @xmath162 becomes penzias - wilson antenna , * fig5 * ) . . so with the basic including hyperbolic unit we have now @xmath322 we will show that @xmath323 on the other side , the introduction of a minimal acceleration @xmath316 involves the deletion of the third phase of cruise velocity : any body in the vacuum undergoes a basic @xmath324 . the sound pioneer @xmath83 follows in curved space - time a * geodesic * which is a generalization of inertial principle what is the relationship between milgrom acceleration @xmath325 @xmath106 and the parameter of acceleration @xmath326 in cosmology ? in other words what is the relationship with our constant hubble radius @xmath52 and the ad hoc scale factor in friedman s metric ( robertson - walker s metric ) , generally noted @xmath4 ( with derivatives with respect to time , @xmath327 , @xmath328 and with @xmath329 ? in cr this factor is naturally equaled to @xmath330 because we have @xmath331 so we have then an accelerated expansion in sr . for the remote galaxies @xmath332 we obtain the same result with the law of expansion in @xmath333 ( factor of bondi ) ( @xmath334 ) @xmath335 @xmath336 so we have : @xmath337 unlike in robertson - walker s metric , we does not need an ad hoc factor in cr because hubble constant @xmath66 is now explicitly connected with the logarithmic derivation of expanding bondi factor @xmath9 in proper time ( primed ) . indeed fundamental equation is in cr is @xmath338 and so the contrast @xmath339 because @xmath340 in the boost . and so the non nul hubble constant involves an exponential expansion compatible with hyperbolic constant horizon . in summary we have switched to the second member @xmath341 friedman s scale factor ( @xmath342 ) by transforming it into constant . then we showed that the scale factor of expansion is bondi s factor @xmath9 and we have therefore to discover a relationship between @xmath52 and @xmath9 ( 36 ) . the lorentz boost describes the continuous transformation from any point of the elastic rod from the velocity @xmath151 to @xmath343 _ _ _ _ the element of proper time @xmath344 depends in cr on hubble constant : with @xmath345 we have a renormalized minkowski s metric ( 21 ) by introducing a fundamental unit @xmath52 @xmath346 locally for _ _ _ _ @xmath347 _ _ ( _ _ @xmath348 @xmath349standard flat metric of ssr is still valid . indeed locally the hyperbola becomes a parabola @xmath350 ) and parabolic einstein s principle of equivalence on the basis of gr is still valid but _ globally _ we have the curvature of fundamental milgrom s ( demi - branch of ) hyperbola @xmath351 ( * fig5 * ) . @xmath352 let us finally show that the geometry is hyperbolic in the meaning of cayley and klein .. with cayley - klein s hyperbolic distance @xmath353 induced from the _ cross - ratio _ formula is given by @xmath354where @xmath63 , is a cartesian distance smaller than @xmath355 the horizon . we have from the global form @xmath356 and therefore @xmath357 with @xmath358 and @xmath59 @xmath359 ( @xmath360 is the proper distance is connected by the other hyperbola ot ( fig1 ) . we have a perfect symmetry between proper time @xmath361 and proper length @xmath362 with the other ( @xmath363 hyperbola @xmath364 and @xmath365 if @xmath366 with element of proper distance @xmath367 ) , we have a basic relationship between @xmath9 and @xmath52 @xmath368 where @xmath9 is the redshift doppler factor of bondi @xcite . this is the inverse form of ( 33 ) . and so we deduced , with @xmath332 the law of hubble ( @xmath151 @xmath369 @xmath370 from integral form , we can therefore deduce globally the law of hubble . in cr , like in esher s drawing , there is a constant hyperbolic horizon and an expanding universe with potentially infinite hyperbolic distances , @xmath371 or @xmath372 ) in time @xmath16 and proper time @xmath143 as well . the universe in cr ( 33 - 36 ) is not static but in steady state@xcite . nothing is changed for a light point because we have obviously @xmath373 is the light sensitive to the curvature in cr or to cosmological constant @xmath2 ( @xcite ) given that light follows in vacuum a straight line and that straight line is in cr hyperbolic straight line ( with curvature ) , the answer is therefore positive . our main heuristic principle is the principle of correspondence non - relativistic and relativistic kinematics ( pcch ) . we have the theater of circular motion and the theater of hyperbolic rotation motion . the roles but not the actors are the same . let us give an example . who plays the role of angular velocity @xmath374 ? in uniform circular motion , unlike constant angular velocity @xmath375 ( in standard units @xmath376 ) , linear velocity @xmath186 depends on radius . in hyperbolic motion linear velocity depend on the radius @xmath377 unlike hubble constant ( in standard units @xmath376 ) which does not depend of distance ( 37 ) @xmath378 last but not least : who plays the role of the transcendent @xmath379 inhyperbolic cr ? the fundamental differential equation in cr , based on @xmath324 , is a first order equation ( 33 - 36 ) . thanks to @xmath31 the solution is not trivial and cr is exactly the completely hyperbolic theory of hyperbolic rotation , i.e. for lorentz boost with milgrom s acceleration . the circular motion governed by @xmath379 is replaced in cr by an expanding hyperbolic motion governing by @xmath380 . this complete substitution characterizes the aesthetic superiority of hyperbolic geometry on the other non - euclidean geometries @xcite . einstein had introduced 1n 1917 the cosmological constant @xmath2 in the first member of einstein s gr equation ; if it is switched to the second member , as is now the general practice , it becomes a term independent from matter in the usual sense of the word . but what is the meaning of a constant , unrelated with matter , in a theory ( gr ) which by essence gives a structure based on matter to the space - time ? we suggest , as a consequence , not to shift it from one member to the other , but to change its relativity in order to reinsert it into its natural environment , in other words , in the vacuum space - time of minkowski . nothing is changed with gr except that @xmath2 must be introduced in einstein s equation . the cosmological constant , @xmath2 , when it is written on the right side , introduces the constant global curvature of the universe ( @xcite ) , as distinct from all the other terms of einstein s gr equation , which correspond to the local curvature ( defined by gravitation in each point of the space - time ) . the fact that the acceleration may not be null is not in contradiction with the local riemann flatness of the pseudo - euclidean space , as derived from the affine connection ( the christoffel ) , under the condition that @xmath2 , defined by cr , be introduced into einstein s gr equation . our use of inverted commas in the title of our paper means that the classical vacuum is not really empty because there is a fundamental field that can be called poincar s relativistic aether or hidden relativistic fluid or in dark energy ( it s the fashion ) . given that nothing is changed between the relation with gr except that the cosmological constant is imposed by cr , we can determine thanks to einstein s gr , and more precisely from einstein s static model of universe , the very weak density of poincar s fluid with poincar s _ non - electromagnetic _ negative pressure ( @xcite ) if einstein is the author of the very well known @xmath382 , poincar is the author of the very unknown @xmath383 in the framework of lt . einstein considered at the end of his life that the cosmological constant @xmath0 in 1917 was his greatest mistake . maybe poincar thought also , at the end of his life , that his greatest mistake was keeping aether in his relativistic dynamics of 1906 ( a relativistic aether but an aether ) ? according to our last formula , maybe both , the old poincar and the old einstein , were wrong . if we put einstein s cosmological constant in his own relativistic kinematics without aether , we rediscover poincar s relativistic aether ( la structure fine de la relativit restreinte , @xcite ) . this is the irony of the history .	* version 1 * * ( results ) * we show that cosmological constant @xmath0 is not optional in gr ( _ general _ relativity ) because it is required by sr ( special relativity ) . this completely unexpected result is obtained by introducing a minimal acceleration @xmath1 ( milgrom ) in einstein boost with lorentz transformation ( lt ) . we prove that hyperbolic rotation ( lt ) is an hyperbolic _ motion _ with a centrifugal acceleration . in cosmological sr with @xmath2 ( csr or cr ) , the universe is not only in expansion ( with the law of hubble ) but even in accelerated expansion ( cosmological parameter @xmath3 . in cr the constant @xmath2 is naturally associated to the bending of light ( ishak and rindler ) . given that the structure of space - time in einstein s gr is determined by the _ presence _ of matter and @xmath2 is associated to the _ absence _ of matter , we associate @xmath2 not to quantum vacuum but to classical vacuum of minkowski s space - time . finally we show that 1917 einstein s cosmological constant @xmath2 corresponds to 1906 poincar s _ non - electromagnetic _ negative pressure and we deduce density of poincar s relativistic fluid . * version 2 ( methodology ) * in * 1 * we switch the scale radius factor @xmath4 of friedman - lematre in the second member of lorentz invariant @xmath5 @xmath6 ( @xmath7 is hubble constant and @xmath8 is milgrom s acceleration ) . in * 2 * we sum up einstein - born - rindler ( ebr ) s model of acceleration in sr . ebr s rigid motion is induced from non - relativistic _ rectilinear _ uniformly accelerated motion . in * 3 * we propose another relativistic theory of acceleration induced from non - relativistic uniformly _ circular _ accelerated motion given that euclidean rotation ( centripetal acceleration ) must be replaced by hyperbolic rotation _ motion _ ( _ centrifugal _ acceleration ) . this new theory of acceleration in lorentz boost involves not a rigid motion but an elastic expanding motion . we introduce a _ new symmetry _ in minkowski s basic diagram of ( scale ) hyperbolas by adopting a new definition of invariant proper distance in such a way that the lted proper distance be a dilated distance ( symmetrically with invariant proper time ) . by giving up einstein s definition of length contraction , minkowski s calibration hyperbola ( passive lt ) and born s acceleration hyperbola ( active lt ) become identical . in * 4 * we introduce a _ basic minimal acceleration _ ( milgrom ) in lorentz boost with fundamental emission of radiation ( cbr ) . by renormalization of minkowski s metric , we show that classical vacuum is perfectly compatible with an expanding universe ( bondi s factor @xmath9 ) where the hubble radius @xmath10 is an hyperbolic horizon . cr is distinct from the doubly special relativity ( dsr ) approach , in that it does not look for a new invariance group adapted to einstein s definition of distance , but for a new definition of distance , adapted to the lt . * figures * we suggest a geometrical version of this paper with * 5 figures. the principle of correspondence circle - hyperbola ( pcch ) is deeply inscribed in hyperbolic rotation ( lt ) . this principle is valid for physical uniformaly accelerated _ circular _ motion @xmath11 and unfformly accelerated _ hyperbolic _ motion @xmath12 . both motion have a center but if the first one is centripetal , the second one is centrifugal*. = 1
the importance of the stem industry to the development of our nation can not be understated . as the world becomes more technology - oriented , there is a necessity for a continued increase in the stem workforce . however , the u.s . has been experiencing the opposite . in the united states , 200,000 engineering positions go unfilled every year , largely due to the fact that only about 60,000 students are graduating with stem degrees in the united states annually [ 17 ] . another obvious indication is the relatively fast growth in wages in most stem - oriented occupations : for computer workers alone , there are around 40,000 computer science bachelor s degree earners each year , but roughly 4 million job vacancies [ 29 ] . therefore , our motivation is to solve this problem of stem workforce shortage by promoting stem education and careers to college students so as so to increase the number of people who are interested in pursuing stem majors in college or stem careers after graduation . in this paper , we present an innovative approach to promote stem education and careers using social media in the form of introducing stem role models to college students . we chose college students as our target population since they are at a life stage where role models are important and may influence their career decision - making [ 15 ] . social media is useful for our study in the following two ways : 1 ) the massive amount of personal data on social media enables us to predict users real life identities and interests so we can identify college students and role models from mainstream social networking websites such as the microblogging website twitter and professional networking website linkedin ; 2 ) social media itself also can serve as a _ natural and effective platform _ by which we can connect students with people already in stem industries . -0.1 in our approach is effective in the following three ways . first , increasing stem presence will inspire students to develop interests in stem fields [ 18 ] . second , the exposure of career stem role models that students can identify with will have positive influence on students , as strongly supported by previous studies [ 12 ] . finally , as a form of altruism , accomplished people are likely to help young people [ 11,6 ] and people who resemble them when they were young [ 21 ] . more importantly , social learning theory [ 1,2 ] , psychological studies , and empirical research have suggested that students prefer to have role models whose race and gender are the same as their own [ 12,30,15 ] as well as who share similar demographics [ 7 ] and interests [ 16 ] . motivated and supported by the findings of these related studies , we select _ gender , race , geographic location , and interests _ as the four attributes that we will use for matching the students with stem role models . in addition , similar interests and close location will further facilitate the potential _ personal connection _ between the students and role models . in particular , we first use social media as a tool to identify college students and stem role models using the data mined from twitter and linkedin . as a popular online network , on the average , twitter has over 350,000 tweets sent per minute [ 27 ] . moreover , in 2014 , 37% social media users within the age range of 18 - 29 use twitter [ 5 ] . this suggests a large population of college users on twitter . in contrast , as world s largest professional network , linkedin only has roughly 10% college users out of more than 400 million members [ 25 ] , but has a rich population of professional users . part of its mission is to connect the world s professionals and provide a platform to get access to people and insights that help its users [ 14 ] . our goal , to connect college students with role models , is _ organically consistent with linkedin s mission and business model_. specifically , we train a reliable classifier to identify college student users on twitter , and we build a program that finds stem role models on linkedin . we employ various methods to extract gender , race , geographic location and interests from college students and stem role models based on their respective social media public profiles and feeds . we then develop a ranking algorithm that ranks the top-5 stem role models for each college student based on the similarities of their attributes . we evaluated our ranking algorithm on 2,000 college students from the 297 most populated cities in the united states , and our results have shown that around half of the students are correctly matched with at least one stem role model from the same city . if we expand our geographic location standard to the state - level , this percentage increases by 13% ; if we look at the college students who are from the top 10 cities that our stem role models come from separately , this percentage increases by 33% . our objective is to do social good , and we expect to promote stem education and careers to real and diverse student population . in order to make a real life impact on the college students after we obtain the matches from the ranking algorithm , we design an implementation to help establish connections between the students and stem role models using social media as the platform . for each student , we generate a personalized webpage with his top-5 ranked stem role models linkedin public profile links as well as a feedback survey , and recommend the webpage to the student via twitter . ultimately , it is entirely up to the student and the role models if they would like to get connected via linkedin or other ways , and we believe these connections are beneficial for increasing interest in stem fields . it is noteworthy that _ linkedin has already implemented a suite of mechanisms to make connection recommendations _ , even though none of which is intended to promote stem career specifically . fig.2 illustrates how our implementation naturally fits into the work flow and business model of linkedin . -0.2 in -0.1 in our study has many advantages . leveraging existing social media ensures that we are able to retrieve a large scale of sampling users and thus our implementation is able to influence a large scale of students . also , due to available apis and existing social media infrastructures , our data collection and our implementation are low cost or virtually free . more importantly , unlike some traditional intervention methods , we recommend stem role models to college students in a _ non - intrusive _ way . we tweet at a student with the link of his personalized webpage , and it depends on himself if he wants to take actions afterwards . finally , our approach is failure - safe in delivery . if there are some twitter users that are classified incorrectly as college students , it has no harmful impact on them even if we promote stem education to them . the major contributions made in this study are fourfold . first , we take advantage of social media to do social good in solving a problem of paramount national interest . second , we take advantage of human psychology , motivation , and altruism . that people are more likely to be inspired by models who are like them , and people who are accomplished are likely to help young people who share similarities with them . third , we have developed a simple yet effective ranking algorithm to achieve our goal and verified its effectiveness using real students . lastly , we design an implementation that seamlessly mashes up with the natural work flow and business model of linkedin to establish the connections between students and role models . stem workforce is significant to our nation , and the shortage in such fields makes promoting stem education and careers indispensable . we review the existing methods of promoting stem education and build on previous research in both computer science and human psychology . previous effort has been made to promote stem education . most existing intervention methods focus on promoting through school educators [ 19 ] , external stem workshops [ 26 ] , and public events such as conferences [ 22 ] . however , very little evidence has shown that these strategies were effective . on the other hand , while none of the methods has utilized the rich database and powerful networking ability of social media , social media - driven approaches have succeeded in many applications , such as health promotion and behavior change [ 33 ] . the abundance of social media data has attracted researchers from various fields . we benefit the most from studies that related to age prediction and user interest discovery . nguyan et al . [ 20 ] studied various features for age prediction from tweets , and guided our feature selection for identifying college students . michelson and macskassy [ 31 ] proposed a concept - based user interest discovery approach by leveraging wikipedia as a knowledge base while xu , lu , and yang [ 32 ] , ramage , dumais , and liebling [ 23 ] both discovered user interest using methods that built on lda ( latent dirichlet allocation)[3 ] or tf - idf [ 24 ] . our study also adopts knowledge from psychological studies that demonstrate the importance of having a role model with similar demographics and interests . karunanayake [ 12 ] discussed the positive effect of having role models with the same race , and it holds across different races ; weber and lockwood [ 30 ] discovered that female students are more likely to be inspired by female role models ; ensher and murphy [ 7 ] indicated that liking , satisfaction , and contact with role models are higher when students perceive themselves to be more similar to them in demographics ; and lydon et al . [ 16 ] suggested that people are attracted to people who share similar interests . these studies help determine the attributes that we selected to match the students with role models . we used the rest api to retrieve twitter data . instead of directly searching for college twitter users among all the general users , we focused on the followers of 112 u.s . college twitter accounts since there is higher percentage of college students among these users . in total , we successfully retrieved more than 90,000 followers . for each user , we extracted the entities of his most recent 200 tweets ( if a user has fewer than 200 tweets , all his tweets were extracted ) and his user profile information , which includes geographic location , profile photo url , and bio . after we filtered out api failures , duplicates , and users with zero tweet or empty profile , we are left with 8,688,638 tweets from 62,445 distinct users . due to the limited information that linkedin api allows us to retrieve , we employed web crawling techniques to obtain the desired information directly from the webpage . we built a program that does automated linkedin public people search and used it to search users based on the most common 1,000 surnames for asians , blacks , and hispanic , and more than 5,000 common american given names . despite some overlapping surnames , the large number of names we searched is still able to ensure the diversity of the potential role models , and our results confirmed that . for each search , the maximum number of users returned is 25 , and we collected the public profile urls of all the returned users . after we deleted the duplicates , we retained 182,016 distinct linkedin users . we employed machine learning techniques to identify twitter college student users ( i.e. from incoming freshmen to seniors ) . first , we labeled our training set . we used regular expression techniques to label college student users and non - college student users . specifically , we studied patterns in users tweets and bio , and constructed 45 different regular expressions for string matching . for example , expressions such as `` i m going to college '' , `` # finalsweek '' , or `` university 19 '' are used to label college students ; and expressions such as `` professor of '' , `` manager of '' , or `` father '' are used to label non - college students . if a user s tweets and bio do not contain any of the 45 expressions , the user is unlabeled . we then manually checked and only counted the correctly labeled users . in the end , we are left with 2,413 labeled users , where 1,103 are college students and 1,310 are non - college students , as well as 60,032 unlabeled users . second , we trained our labeled data set to develop a reliable classifier using the libsvm library [ 4 ] in weka [ 8 ] . we chose svm for our binary classification because it is efficient for the size of our data set . we learned from nguyan et al.s study of language and age in tweets [ 20 ] that the usage of emoji , hashtag , and capitalized expressions such as `` haha '' and `` lol '' are good age indicators . we built on their study and took a step further to use these three features for differentiating college students ( i.e. specific age group ) from general users . we were also curious about whether re - tweet would be another good age indicator , so we also extracted this feature . for each user , each feature is represented by its _ relative frequency _ among the user s tweets : @xmath0 since relative frequencies are continuous , we discretized them into 10 bins with an equal width of 0.1 and assigned them with ordinal integer values for classification . -0.2 in -0.1 in fig.3 demonstrates our analysis of the four features . on average , college student users use emojis and haha / lol more frequently while non - college student users use hashtags more frequently . we note that these results are consistent with the conclusions of a previous study [ 20 ] . however , there is not much difference in re - tweet between these two groups . we experimented training the classifier with and without re - tweet , our 10-folds cross - validation results showed that including re - tweet actually slightly lowers the accuracy of the classifier . thus , we confirmed that re - tweet is a noise and does not help us to differentiate college student users . our final classifier trained from the other three features achieves a high accuracy of 84% . we then used this trained classifier to infer college student users among the unlabeled users . we further labeled 18,351 users as college students , and with our manually labeled college student users , together we have labeled 19,454 college student users in total . our goal is to find diverse stem role models from linkedin in terms of geographic locations and industries . while the definition of a role model is subjective to an individual student , we take an objective view and consider people who have received stem education and work in stem - related industries or have a career in stem industries as role models . we first filtered out users who are outside of the united states and then built a _ role model identification _ program to find stem role models . the program takes in a user s profile url , crawls the contents in `` industry '' and `` education '' fields on the user s profile and only outputs the url if the user is a stem role model . specifically , we divided all 147 linkedin industries into three groups , `` non - stem '' , `` stem '' , and `` stem - related '' . for example , `` biotechnology '' and `` computer software '' are `` stem '' , `` music '' and `` restaurants '' are `` non - stem '' , and `` financial services '' and `` management consulting '' are `` stem - related '' . we only consider those users who are under `` stem '' or under `` stem - related '' with a degree in stem majors as role models . we used the 38 stem majors offered at our university as our standard . after we obtained the profile urls of stem role models , we crawled their entire profiles using the urls . we successfully found 25,637 stem role models from 2,022 distinct locations in the united states , including some places in hawaii and alaska . fig.4 shows a rough visualization of the diverse geographic locations the stem role models come from . the top-10 cities that role models come from are , not surprisingly , san francisco , new york city , atlanta , los angeles , dallas , chicago , washington d.c . , boston , seattle and houston . this section presents the methods we employed to extract the gender , race , geographic location and interests from college students and stem role models as well as our ranking algorithm that matches them based on the similarities of these attributes . we reiterate that our selection of attributes are supported by a variety of previous related studies . these factors can make the most influential pairing because they ensure that a student gets a mentor with a similar background for affinity . moreover , close geographic location and similar interests are valuable for potential real life interaction between the students and role models . we extracted race and gender from both textual and visual features , namely the users names and profile photos . we recognize that there are people who identify themselves with genders other than male and female ; we also recognize that there are a variety of ways for categorizing races . to build a prototype system , we will use male , female for gender categorization , and use white , black , asian , asian pacific islander ( i.e. api ) and hispanic for race categorization . -0.1 in in particular , we used genderize.io , face++ , and demographics to extract these two attributes . genderize.io and demographics predict gender or both gender and race based on the user s given name or full name while face++ predicts both using the user s profile photo . in total , we obtained three gender predictions and two race predictions for each user . each prediction is returned with an accuracy , and in the case the tool fails to predict , the prediction will be null . we picked the gender and race predictions with the highest accuracy . as a result , we extracted the gender of 80% college students and 97% role models , and the race of 46% college students and 92% role models . almost all role models have both attributes since we used their linkedin profiles , where the profile photos are usually high quality and the names are usually real . in contrast , twitter profiles sometimes can contain profile photos with random objects and invented names . fig.5 shows the make - up of those college students and stem role models whose gender and race were successfully extracted . we directly extracted geographic locations from the `` location '' field in twitter and linkedin profiles . the interests extraction is less straightforward and we used other features as proxies for this attribute . we were able to extract the locations of all stem role models since linkedin requires users to have a valid geographic location on their profiles . these locations usually contain the city and the state that role models work in . however , twitter does not have this requirement , and we noticed that not every college student has filled the location field on his profile and some of the filled locations are not valid . in fact , 34% twitter users either did not fill the `` location '' field or provided fake geographic locations ; among those valid locations , roughly 65% are at city - level [ 11 ] . in addition , we observed that many students use the name of their educational institutions as locations , and some locations are not correctly spelled or formatted . for example , a student s location is `` mcallentx '' , which refers to the city mcallen in texas , but not a place called `` mcallentx '' . due to the difference in the nature of linkedin and twitter , we selected different proxies as interests for role models and college students . for role models , we directly extracted the contents in `` interests '' and `` skills '' fields as their interests because skills such as `` web development '' can also be an interest , and people usually are good at things that they are interested in . for college students , we extracted hashtags ( excluding prefix `` # '' ) as interests . a hashtag is a user - defined , specially designated word in a tweet , prefixed with a `` # '' [ 31 ] . originally , we experimented lda topic modeling to discover topics of interests from all college students tweets and intended to use these to define each student s interests . however , due to the noise and non - interest related terms in tweets ( excluding stop words and non - english words ) , most of the terms generated are too generic to be defined as topics of interests . therefore , we extracted one s unique hashtags as proxy for interests . hashtags have been used in characterizing topics in tweets [ 23 ] and have shown to be interest - related to a decent extent [ 32 ] . although high - frequency hashtags are intuitively better representations of one s interests , including all unique hashtags allows us to extract a wilder range of interests . after we extracted interests from both students and role models , we stored everyone s interests as a set which we call _ interest set_. the size of the set varies from user to user depending on the number of interests of that user . we rank all stem role models for each student based on the similarities of their attributes . specifically , for each comparison of a student and a role model , we calculate the similarity of each attribute , and rank the role model based on the arithmetic average across similarities of all attributes . we will now explain our methods used for each comparison . for gender and race , we simply compared if the two people have the same string for gender or race . in our case , there are two strings for gender , `` female '' and `` male '' , and five strings for race , `` white '' , `` black '' , `` asian '' , `` api '' , and `` hispanic '' . therefore , the _ gender similarity _ is either 1 or 0 because two people either have the same gender or not , and the same went for _ race similarity_. for geographic locations , we used string comparison method to measure the similarity of two locations . originally , we experimented two ways to calculate it : the actual distance between two locations based on their latitudes and longitudes , and the levenshtein distance between the two strings that represent the two locations . due the variety of possible expressions of the same location , traditional tool such as geocoder can only correctly convert well - formatted locations that do not contain non - letter characters . for example , a real college student has location `` buffalo state 18 psych majorr '' and it can not be successfully converted into coordinates using geocoder , but clearly that the student studies in buffalo . since our objective is to be able to compare as many locations as possible , we decided to use string comparison , which allows the flexibility of using various location representations for the same place . specifically , we employed levenshtein distance [ 13 ] to calculate the distance between two strings , and the _ levenshtein - based similarity _ ( a ratio between 0 and 1 ) is defined as : @xmath1 where in the case of location , @xmath2 is the string of the student s ( role model s ) location , and we then have our _ location similarity_. a minor problem of this similarity measure is that two geographically different locations might contain similar words and have a high similarity , such as `` washington d.c . '' and `` washington state '' . but this happens relatively rare only if there are enough people from one of the location or both . we used jaccard coefficient [ 10 ] combined with _ levenshtein - based similarity _ to compute the similarity of two _ interest sets_. hashtags are often not real words but a combination of words without spaces . while a real student s hashtag `` computersciencelife '' and a real role model s interest `` computer science '' clearly refer to the same interest in the field of computer science , the two strings are different and have a _ levenshtein - based similarity _ of 0.86 . therefore , in order to capture the overlapping interests between two _ interest sets _ , we need a threshold for _ levenshtein - based similarity _ that decides whether two strings refer to the same interest . after extensive experimenting with real data , we chose our threshold to be 0.8 . our _ interest similarity _ is then defined as : @xmath3 where @xmath4 is the student s ( role model s ) _ interest set_. a potential problem is that since our measurement is string - based but not concept - based , it might not capture the synonymous of interests as overlapping interests . after we calculated the similarities of all four attributes , we combined them by taking the arithmetic average and used that to rank the role models . in the cases of missing values , any unlabeled attributes is not taken into account . for instance , if a student does not have gender information , the arithmetic average will entirely depend on the similarities of his other three attributes . in this section , we verified our ranking algorithm on 2,000 college students from the 297 most populated cities in the united states [ 28 ] . all these students are randomly selected from our database . we manually evaluated their top-5 ranked role models , and we also recommended these role models to them via twitter . although it is desirable to evaluate the ultimate impact of our study , we recognize that this would require tracking the subjects of the study over their career of substantial length ( e.g. , over 10 years ) . therefore , it is beyond the scope of this study , and we decided to use matching accuracy as the performance measure , which is defined as : @xmath5 where n is the second metric , the specific number of role models out of the top-5 that are correctly matched with the student . it represents the granularity level of matching . we consider a student is _ correctly _ matched with a role model if the linkedin user is indeed a stem role model and has the same gender , race and geographic location as the student . we did not evaluate interests since they are often not explicitly stated in social media and it would be too difficult to discover every student s real interests by reading his tweets . we took a careful effort to manually evaluate the matching results of these 2,000 college students by checking their twitter profile pages and the linkedin profile pages of their top-5 ranked stem role models . we utilized all the information on their respective social media profiles to determine their gender , race , and geographic location . in order to determine if someone is indeed a stem role model , we make our best judgment , as a career counselor would , based on the entire linkedin profile , which usually includes demographic background , personal summary , industry , education , working experience and skills . if we failed to determine any of the three attributes of a student , we will have to consider that he is not correctly matched with any role model because we are unable to conduct the evaluation . consequently , for twitter public accounts and students with unlabeled gender , race , or invalid location , they all receive zero correctly matched role models . location should not have been a problem since we selected these students by their locations , but we found that a handful of students have removed or changed their locations after we collected the data . -0.15 in -0.1 in table 1 shows two randomly selected representative examples of the matching results for two students . we consider that the student in the top table was correctly matched with all five role models and the student in the bottom table was only correctly matched with # 3 and # 4 role models at state - level because # 1 role model is not in stem - related occupation and # 2 and # 5 are not asian . none of the role models was correctly matched at city - level . taking into consideration that our limited database of stem role models may have an impact on the performance of the ranking algorithm , we conducted evaluation in four levels : city - level for 297 cities , state - level for top 297 cities , city - level for top-10 cities , and state - level for top-10 cities . among the 2,000 selected students , about a quarter of the selected students are from the top 10 cities . intuitively , we expect more students to be correctly matched with role models at the state - level than city - level . also , we expect students from the top-10 cities to be correctly matched with more stem role models because there should be more diverse role models in these cities . -0.1 in -0.1 in -0.1 in in fig.6 we show the overall matching accuracy in the four levels . we first look at our baseline , the city - level for 297 cities . 42% of the college students were correctly matched with at least one role model . we noticed that around half of them was not matched with any role model and this is partly due to those college students with unlabeled gender and race information . we then noticed that our ranking algorithm performs better in the 10 cities than in the 297 cities for both city and state levels . numerically , the difference increases as the the minimum number of correctly matched role models decreases . if we look at students who were correctly matched with at least one role model , for both city and state levels , the top-10 cities outperforms the 297 cities by 33% and 21% , respectively ; the ranking algorithm achieves a decent accuracy of 57% in both city and state levels for the 10 cities . also , our ranking algorithm performs better in the state - level than in city - level for the 297 cities . with students who were at least correctly matched with one role model , the difference is 13% , which is smaller but still very significant . however , there is almost no difference in state and city levels for the top-10 cities . a possible explanation is that because there are more stem role models of various types in the top-10 cities , the student can usually get matched with stem role models who are from the exact same city . during our evaluation , we are encouraged to see that there is a good variety of stem role models in different industries even for students with the same demographic background . we think this is a positive indicator that the attribute , interests , in fact contributes to our ranking algorithm . -0.1 in -0.1 in in order to make a real - life impact , for each student , we generated a personalized webpage and delivered the link of the webpage via tweeting at him from the official twitter account of our study . fig.7 shows an example of such webpage . it contains the linkedin public profile links of his top-5 role models and a survey regarding the accuracy of our recommendations . we only received a very small number of responses and conducted preliminary analysis . all responses indicated that they are indeed currently college students , a third agree that the recommendations are good and a third indicated that they would be more interested in stem majors / careers if they had role models in stem fields . we would need more responses to validate our implementation , and a potential way to do so is to cooperate with our university , apply the ranking algorithms on students who are twitter users and ask for responses . in this paper , we present an innovative social media - based approach to promote stem education by matching college students on twitter with stem role models from linkedin . our ranking algorithm achieves a decent accuracy of 57% in the city - level for the top-10 cities that the stem role models come from . we also design a novel implementation that recommends the matched role models to the students . to achieve this , we identified college students from twitter and stem role models from linkedin , extracted race , gender , geographic location and interests from their social media profiles , and developed a ranking algorithm to rank the top-5 ranked stem role models for each student . we then created a personalized webpage with the student role models and recommended the webpage to the student via twitter . our recommendation is not imposed on either side . it is the students choice if they want to initiate the connection with the role models via linkedin or other methods ; and it is for the role models to decide if they want to accept their linkedin invitations or other forms of communication . in the case of linkedin , note that if a student decides to approaches a potential role model , he can express why he would like to get connected ( e.g. , interest in stem fields ) , and the role model can make his own judgment . one may worry that our implementation of recommendations may be considered a form of spamming on students , however , our intention is clearly to help their careers , and _ not to profit _ from them . there are several possible extensions of our study in the future . our approach might have a reduced effect for college seniors since it is more difficult for them to switch majors . however , it is not uncommon that students change their career paths after graduation , and in the future we could recommend role models with similar experiences specifically to seniors . we could also expand our target population to high school students or focus on promoting stem education specifically to minority college students . in addition , we could classify stem role models into specific groups such as current stem major college students and experienced stem role models since students might feel more comfortable reaching out to their peers . finally , we could design an application based on our implementation to achieve real - time matching , where a college student could log into our application using their twitter account , and we could collect their data , extract their attributes , and give them stem role model recommendations in real - time . this application could also be generalized to other social media since many methods we used are compatible with other platforms . we hope this study can serve as a starting point to make use of the rich data and powerful networking ability of social media `` by the people '' in order to promote stem education and build positive influence `` for the people '' . this work was supported in part by xerox foundation , and new york state through the goergen institute for data science at the university of rochester . we thank all anonymous subjects for contributing to the evaluation of our system . chang , c .- c . , lin , c .- j . : libsvm : a library for support vector machines , 2001 . duggan , m.,ellison , n. b. , lampe , c . , lenhart , a. , and madden , m. : demographics of key social networking platforms ( january 2015 ) . emmerik , h.v.,baugh , s.g . , euwema , m. c. : who wants to be a mentor ? an examination of attitudinal , instrumental , and social motivational components.career development international 10 , 4 , 310 - 340 ( 2005 ) hecht , b. ; hong , l. ; suh , b. ; chi , e. h. : tweets from justin bieber s heart : the dynamics of the `` location '' field in user profiles . proceedings of acm chi conference on human factors in computing systems ( 2011 ) machi , e. : improving u.s . competitiveness with k-12 stem education and training ( sr 57 ) . a report on the stem education and national security conference october 21 - 23 . ( 2008 ) merrill , c. , custers , r.l . , daugherty , j . , westrick , m. , and zeng , y. : delivering core engineering concepts to secondary level students . journal of technology education 20 , 1 , 48-64 ( 2008 ) nguyen , d . , gravel , , trieschnigg , d. , and meder , t. : `` how old do you think i am ? '' : a study of language and age in twitter . proceedings of international conference on weblogs and social media , pp . 439448 ( 2013 ) xu , z.h . , lu , r.,xiang , l.,yang , q.:discovering user interest on twitter with a modified author - topic model . proceedings of the 2011 ieee / wic / acm international conferences on web intelligence and intelligent agent technology ( 2011 ) heldman , a. b. , schindelar , j. , weaver , j. b. : social media engagement and public health communication : implications for public health organizations being truly `` social '' . public health reviews , 35 , 1 ( 2013 ) hadgu , a.t . ; jaschke , r. : identifying and analyzing researchers on twitter . proceedings of acm conference on web science , pp . 23 - 32 ( 2014 ) mislove , a. ; lehmann . s. ; ahn , y. ; onnela , j. ; rosenquist , j. n. : understanding the demographics of twitter users , proceedings of fifth international aaai conference on weblogs and social media . aaai press , pp.554 - 557 ( 2012 )	stem ( science , technology , engineering , and mathematics ) fields have become increasingly central to u.s . economic competitiveness and growth . the shortage in the stem workforce has brought promoting stem education upfront . the rapid growth of social media usage provides a unique opportunity to predict users real - life identities and interests from online texts and photos . in this paper , we propose an innovative approach by leveraging social media to promote stem education : matching twitter college student users with diverse linkedin stem professionals using a ranking algorithm based on the similarities of their demographics and interests . we share the belief that increasing stem presence in the form of introducing career role models who share similar interests and demographics will inspire students to develop interests in stem related fields and emulate their models . our evaluation on 2,000 real college students demonstrated the accuracy of our ranking algorithm . we also design a novel implementation that recommends matched role models to the students . , recommendation systems , social media , text mining
at @xmath0 , a large fraction of all high mass stars form in dusty galaxies @xcite . most of the intrinsic uv emission from newly formed stars in these galaxies is obscured , or absorbed by dust grains that subsequently heat up and radiate at ir wavelengths . the ir luminosity resulting from this obscuration is usually much greater than the emergent uv luminosity . for galaxies in which the intrinsic uv emission from newly formed stars is less obscured , the ir luminosity is still greater than the emergent uv luminosity , but to a lesser degree @xcite . the relation between the ir and emergent uv emission from a @xmath0 galaxy depends on the interplay between star formation and dust obscuration . one of the many ways to select dusty galaxies at @xmath0 , without redshift determinations from spectroscopy , is to use the ratio of observed 24 to 0.65@xmath1 ( @xmath4-band ) flux densities @xcite . sources satisfying @xmath5 have been termed `` dust - obscured galaxies '' , or dogs ; their redshift distribution is approximately a gaussian that peaks at @xmath6 with @xmath7 @xcite . in the redshift range @xmath8 , 0.65@xmath1 observations are sensitive to rest - frame uv emission from newly formed massive stars , and 24@xmath1 observations are sensitive to mid - ir emission from hot dust and polycyclic aromatic hydrocarbons ( pahs ) . the dog criterion is thus unique in that it selects galaxies in a specific redshift range , that also exhibit extreme ratios between their rest - frame mid - ir and uv flux densities . we have yet to understand the physical mechanism driving the span of ratios exhibited by dusty galaxies at @xmath0 . the ir luminosities of dogs with @xmath9 are dominated by emission from active galactic nuclei ( agn ; @xcite ) . the dominant sources of the ir luminosities of less luminous dogs is a topic of debate . @xcite and @xcite conclude that the ir luminosities of many dogs with @xmath10 originate from agn , while @xcite conclude that many such dogs are powered by newly formed stars . in this paper , we pose the question `` what makes a dog a dog ? '' the primary goal of our study is determining why dogs have an extreme ratio between their rest - frame mid - ir and uv flux densities when compared to other dusty galaxies . unfortunately , the simple and singular selection criterion can not distinguish between a dog that is : * abnormally bright at rest - frame @xmath11 for its far - ir flux density , indicating its mid - ir luminosity may be dominated by agn emission , or abnormally strong emission from polycyclic aromatic hydrocarbons ( pahs ) ; * or , abnormally faint in the rest - frame uv for its optical flux density , indicating that dust more completely obscures the newly formed stars in the galaxy . we use _ herschel _ @xcite data in the great observatories origins deep survey - north ( goods - n ) region @xcite to show that , on average , dogs with @xmath12 are not abnormally bright at @xmath11 , but are more uv faint than other galaxies with similar ir luminosities . the ratio between rest - frame ir and uv flux densities is set by the amount of obscuration , which can vary with either : 1 ) the degree of alignment between the spatial distributions of dust and massive stars , or 2 ) the total dust content . this paper is organized as follows . we present the data and sample selection in [ sec : data ] ; in [ sec : results ] , we show the results . we discuss the implications of these results in [ sec : discuss ] , and conclude in [ sec : conclude ] . we assume a cosmology with @xmath13 km s@xmath14 mpc @xmath14 , @xmath15 , and @xmath16 . our study uses observations of the goods - n region , which is roughly 10 arcmin @xmath17 16.5 arcmin in extent . we cull the sample of dogs from a catalog of 24@xmath1 sources produced for the _ spitzer_/mips survey of the goods - n region ( m. dickinson , pi ; @xcite ) . a 24@xmath1 source is defined as a @xmath18 flux density measurement from psf fitting to _ 3.6@xmath1 source priors . the catalog is 99% complete at @xmath19jy , and contains 1603 sources . the 2.2@xmath1 ( @xmath20-band ) image we use to identify counterparts for the 24@xmath1 sources comes from observations using the canada - france - hawaii telescope ( cfht ) . the data are presented in @xcite ; we use our own reductions @xcite . the 0.65@xmath1 ( @xmath4-band ) subaru image we use to define the dog sample comes from @xcite . the 5@xmath21 depth of the 2.2@xmath1 image is @xmath220.60@xmath23jy ( 24.5 ab mag ) ; the 3@xmath21 depth of the 0.65@xmath1 image is @xmath220.05@xmath23jy ( 27.2 ab mag ) . to extract flux densities , we follow a modified version of the procedure used by @xcite . using sextractor @xcite , we place 3@xmath24 diameter apertures at the positions of sources detected ( @xmath25 ) in the 2.2@xmath1 image . if the 2.2@xmath1 flux density is detected with s / n @xmath25 but the 0.65@xmath1 flux density is not detected with s / n @xmath26 , we use a 3@xmath21 limit for the latter flux density . to determine rest - frame uv continuum power - law indices , we extract flux densities at 0.45 , 0.55 , 0.80 , and 0.90@xmath1 ( the @xmath27- , @xmath28- , @xmath29- , and @xmath30-bands ) from subaru images @xcite , using the same procedure . we use the 3.6 , 4.5 , 5.8 , and 8@xmath1 flux densities already associated with the 24@xmath1 sources to determine whether or not their spectral energy distributions ( seds ) at these wavelengths behave as power laws ; these flux densities come from a catalog produced for the _ spitzer_/irac survey of the goods - n region , and will be included in catalogs accompanying the goods-_herschel _ public data release . for the optical / near - ir photometry , we calculate aperture corrections , defined as the ratios of total flux density to flux density in a 3@xmath24 diameter aperture for point sources ( non - saturated stars ) . we take the sextractor parameter ` flux_auto ` as the total flux density . the corrections are factors of 1.086 , 1.225 , 1.247 , 1.057 , 1.086 , and 1.057 at 0.45 , 0.55 , 0.65 , 0.80 , 0.90 , and 2.2@xmath1 , respectively . to maintain the signal - to - noise ratios given by sextractor , both the flux densities and their uncertainties are multiplied by these factors . we associate each 24@xmath1 source with a 2.2@xmath1 source and its extracted optical flux densities if the 2.2@xmath1 source is a unique match within 0.76@xmath24 of the position of the 3.6@xmath1 prior . the match radius is chosen by maximizing the number of unique matches while minimizing the number of multiple matches . of the 1603 24@xmath1 sources , 87 either do not have a @xmath25 2.2@xmath1 counterpart within the match radius ( 85 of 87 ) or have multiple counterparts ( 2 of 87 ) . the far - ir flux densities come from a catalog produced for the goods-_herschel _ survey @xcite . we only use 100 and 160@xmath1 flux densities measured with pacs @xcite , and 250@xmath1 flux densities measured with spire @xcite , to avoid the complications of measuring flux densities for 24@xmath1 sources in the 350 and 500@xmath1 spire images that are affected by severe source confusion . we consider a @xmath18 measurement at 100 or 160@xmath1 to be a detection ; at 250@xmath31 , we require a @xmath25 measurement . we impose additional constraints on the 250@xmath1 flux densities ( and @xmath32 limits ) , similar to those defining the `` clean index '' @xcite . the 250@xmath33 flux densities of clean sources ( and @xmath32 limits of clean non - detections ) should not be affected by severe source confusion . we do not impose additional constraints on the 100 and 160@xmath1 flux densities because the 100@xmath1 images are not deep enough to be affected by source confusion , and the 160@xmath1 images are only deep enough to be moderately affected . the 3@xmath21 depths of the 100 and 160@xmath1 images are @xmath221100@xmath23jy and @xmath222700@xmath23jy , respectively ; the 5@xmath21 depth of the 250@xmath1 image is @xmath225700@xmath23jy . we attempt to match each 24@xmath1 source to a source with a spectroscopic redshift from the catalogs of ( * ? ? ? * which includes redshifts compiled from the literature ) and stern et al . ( in preparation ) . we find spectroscopic redshifts for 910 ( 57% ) of the 24@xmath1 sources . if no coincident sources with spectroscopic redshifts are found , we resort to the photometric redshift source catalog of pannella et al . ( in preparation ) to find a source match . we exclude sources with photometric redshifts derived from ill - fitting templates ( those with reduced @xmath34 ) . for an additional 510 ( 32% ) of the 24@xmath1 sources we have photometric redshift estimates . there are no redshift estimates for 183 ( 11% ) of the 24@xmath1 sources , and these sources are excluded from our samples . using the multi - wavelength information and redshifts , we define 2 samples from the superset of all 24@xmath1 sources with @xmath19jy : * _ dog sample _ : all 24@xmath1 sources with 2.2@xmath1 counterparts and @xmath35 ( fig . [ fig : sample ] ) . the 24@xmath1 flux density of the faintest dog is 53@xmath23jy , justifying our @xmath19jy cut for the control sample . the limiting quantity is the depth of the 0.65@xmath1 image ( 3@xmath21 = 0.05@xmath23jy * ? ? ? the redshift distribution of our sample of dogs is shown in fig . [ fig : z_dist ] , also motivating our @xmath8 cut for the control sample . six ( of 61 ; 10% ) dogs have spectroscopic redshifts . in the following analysis , we include only dogs with @xmath8 . our conclusions do not change if we include dogs without redshift estimates in the sample . * _ control sample _ : all @xmath19jy 24@xmath1 sources with 2.2@xmath1 counterparts that are at @xmath8 , and that do not satisfy the dog selection criterion . seventy four ( of 268 ; 28% ) control galaxies have spectroscopic redshifts . for each sample , table [ table : numbers ] characterizes the subset of sources with flux densities detected at 0.65 , 100 , 160 , and 250@xmath1 . more than 70% of these galaxies are undetected in optical spectra ( or are unobserved ) because their observed - frame optical flux densities are so faint . our sample contains dogs with fainter 24@xmath1 emission than does the @xcite sample . their sample is selected from the shallower _ spitzer_/mips survey of the botes region ; their limit is @xmath36jy . however , the goods - n region is much smaller than the botes region , so we have few dogs with @xmath36jy . @xcite also study a sample of dogs in goods - n . the main differences between the @xcite sample and ours are that : 1 ) they limit their sample to @xmath37jy ; and 2 ) they estimate a redshift for each dog using irac and mips photometry , whereas we match dogs to sources in a near - ir / optical catalog , with redshift estimates based on @xmath38 , 3.6 , and 4.5@xmath1 photometry . the fractions of 24@xmath1 sources at @xmath8 that meet the dog criterion increase with increasing 24@xmath1 flux density ( fig . [ fig : sample ] ) . of the sources with @xmath39jy , 4% are dogs . of the sources with @xmath37jy , 25% are dogs . @xcite find similar fractions ; they also find that of their sources with @xmath36jy , 60% are dogs . lccccccc dogs & 61 & 2.1 & 161 & 47 ( 77% ) & 29 ( 48% ) & 24 ( 39% ) & 9 ( 15% ) + control & 268 & 2.0 & 102 & 268 ( 100% ) & 81 ( 30% ) & 52 ( 19% ) & 15 ( 6% ) flux density vs. 0.65@xmath1 flux density for galaxies in the two samples . the dog sample is defined by @xmath40 . [ fig : sample ] ] sources ( before we impose redshift limits ) . for the analysis , we limit all samples to the redshift range @xmath8 ( the area between the vertical lines).[fig : z_dist ] ] several quantities are useful in analyzing the relation between ir and emergent uv emission from galaxies . in this section , we detail how we estimate the total ir and uv luminosities , uv continuum power - law indices , star formation rates , and stellar masses for the galaxies in our samples . we estimate a total ir luminosity ( 81000@xmath1 ; @xmath41 ) for each galaxy with detected emission at 100@xmath1 . we redshift the @xcite template spectral energy distributions ( seds ) to the distance of each galaxy , find the sed that most closely matches the observed 100@xmath1 flux density , and multiply the ir luminosity of that sed by the ratio between actual and predicted flux densities to get @xmath41 . we prefer this approach over estimating the ir luminosity directly , by summing several far - ir flux densities , because the latter procedure requires detected emission at 160 and 250@xmath1 . the goods-_herschel _ image at 160@xmath1 is moderately affected by blending due to source confusion , while the image at 250@xmath1 is so deep that blending is problematic . the drawback to our chosen approach is that all statements we make regarding ir luminosities assume that the low - redshift template seds accurately represent the seds of galaxies at @xmath0 . @xcite show that this assumption is valid when template matching is done to 100@xmath1 flux densities . for galaxies with uv emission from newly formed massive stars , the uv continuum can be approximated as a power law with an index @xmath42 : @xmath43 we use @xmath18 flux densities at 0.45 , 0.55 , 0.65 , 0.80 , and 0.90@xmath1 ( i.e. , in the @xmath27- , @xmath28- , @xmath4- , @xmath29- , and @xmath30-bands ) to fit for @xmath42 and the constant factor @xmath44 for each galaxy which has an estimate of @xmath41 . if only two flux densities have s / n @xmath18 , we calculate @xmath42 analytically . if only one flux density has s / n @xmath18 , we use an upper limit for the 0.45@xmath1 flux density to calculate a lower limit to @xmath42 . we estimate a uv luminosity @xmath45 at rest - frame 0.16@xmath1 for each galaxy using its redshift estimate and the power - law fit to the rest - frame uv flux densities . to estimate a stellar mass for each galaxy , we fit stellar population synthesis models to its @xmath38 , 3.6 , and 4.5@xmath1 flux densities @xcite . full details are in 4.2 of @xcite . we assume the stellar initial mass function in @xcite from 0.1 to 100 @xmath46 , as well as the dust attenuation law in @xcite . we use the equation in @xcite to calculate a star formation rate ( sfr ) , based on the ir luminosity , for each galaxy with detected 100@xmath1 emission . we make no correction for the emergent uv luminosity , since it is negligible for all galaxies in the dog and control samples ( see [ sec : discuss ] ) . in using the @xcite equation , we assume that the observed 100@xmath1 emission is due to star formation and not agn activity , and that the star formation episode lasts for @xmath47 years . @xcite find that the former assumption is valid for most agn with detected x - ray emission at @xmath48 . seventy six percent of the @xmath0 galaxies with detected 100@xmath1 emission have @xmath3 ( fig . [ fig : lir ] ) . the distributions of ir luminosities for the dogs and the control galaxies are statistically indistinguishable ( @xmath49 that the two samples are drawn from the same parent population , using a k - s test ) . . the two distributions are statistically indistinguishable . [ fig : lir ] ] fig . [ fig : mass_z ] shows the stellar masses for dogs and control galaxies with detected emission at 100@xmath1 . the distributions are not strongly different ( @xmath50 , using a k - s test ) , though some control galaxies have lower stellar masses than do the dogs . 17 ( of 72 ; 24% ) control galaxies have stellar masses @xmath51 , while only 1 ( of 25 ; 4% ) dog has a stellar mass below this threshold . . the distribution of stellar masses for dogs is not strongly different from that for control galaxies , though some control galaxies have lower stellar masses than do the dogs . twenty four percent of the control sample has @xmath52 , as opposed to 4% of the dog sample.[fig : mass_z ] ] the infrared seds of most low redshift dusty galaxies peak between rest - frame @xmath53 and 100@xmath1 . the rest - frame @xmath11 luminosity is only a fraction of the total ir luminosity in these galaxies . to address whether or not dogs are abnormally bright at rest - frame @xmath11 , we require a comparison of the rest - frame far - ir and @xmath11 flux densities between dogs and the control galaxies [ fig:100over24_z ] shows that dogs are statistically indistinguishable from the control galaxies when looking at the observed flux density ratio @xmath54 ( @xmath55 , using a k - s test ) . in this figure , we do not show the galaxies with limits for @xmath56 . we perform the gehan and logrank tests , which are conceptually similar to the two sample k - s test , but also allow for the inclusion of galaxies with limits at 100@xmath1 in the two samples . we find no statistically significant difference between the dogs and the control galaxies ( @xmath57 that both samples are drawn from the same parent population , from both tests ) . our conclusions are the same using @xmath58 ( not shown ; @xmath59 from both tests ) . if dogs were abnormally luminous at rest - frame @xmath11 for their far - ir luminosities , then we would expect them to have low values of observed @xmath54 and/or @xmath58 compared to those of the control galaxies ; they do not . however , 100 and 160@xmath1 observations are sensitive to rest - frame 33 and 53@xmath1 emission from galaxies at @xmath6 . these rest - frame wavelengths are still short of the presumed wavelength of the peak of the infrared sed ; the rest - frame luminosities are still only a fraction of the total ir luminosity in these galaxies . 250@xmath1 observations are sensitive to rest - frame @xmath60 emission from galaxies at @xmath6 . this rest - frame wavelength is generally close to the wavelength of the peak of the ir sed . [ fig:250over24_z ] shows that dogs are statistically indistinguishable from the control galaxies when looking at the observed flux density ratio @xmath61 . we reach the same conclusion when including galaxies with limits at 250@xmath1 in statistical tests ( @xmath62 that both samples are drawn from the same parent population , from both a gehan and logrank test ) . vs. redshift for galaxies in the two samples . galaxies with limits at 100@xmath1 are not shown in this figure , but we do include them in our statistical tests . dogs are statistically indistinguishable from the control galaxies in @xmath54 . [ fig:100over24_z ] ] vs. redshift for galaxies in the two samples . galaxies with limits at 250@xmath1 are not shown in this figure , but we do include them in our statistical tests . as in fig . [ fig:100over24_z ] , dogs span the same range of flux density ratios as do the control galaxies ; this leads us to conclude that the dog criterion does not select dusty galaxies that are abnormally bright at rest - frame @xmath11 . [ fig:250over24_z ] ] the dog criterion does not select galaxies that are abnormally bright at rest - frame @xmath11 for their far - ir flux densities . what makes a dog a dog must be that the galaxy s rest - frame uv emission is suppressed , compared to the uv emission from a control galaxy . therefore we expect a clear separation between dogs and control galaxies when looking at the ratios of rest - frame optical to uv flux densities . indeed , fig . [ fig : roverk_z ] shows that 92% of dogs have an observed @xmath63 , while 78% of control galaxies have @xmath64 . the gehan and logrank tests report a statistical difference between the dogs and control galaxies ( @xmath65 that both samples are drawn from the same parent population , from both tests ) . galaxies with @xmath66 are also known as `` extremely red objects '' , or eros ( @xcite ; @xcite ; @xcite ; @xcite ) . we note that even though a galaxy with detected 24@xmath1 emission may meet the ero criterion , it may not meet the dog criterion ( and vice versa , rarely ) . vs. redshift for galaxies in the two samples . a dog is distinguishable from a random galaxy with detected 24@xmath1 emission because it is abnormally faint in the rest - frame uv . [ fig : roverk_z ] ] the statistical difference between the dogs and control galaxies for @xmath67 , and the lack thereof for @xmath54 , is robust against photometric redshift errors . we offset photometric redshifts for the superset of all galaxies with detected 24@xmath1 emission , assuming the offsets obey a gaussian distribution with @xmath68 = 0.1 . we then select new dog and control samples . the differences in @xmath67 are always statistically significant , and the differences in @xmath54 are very rarely statistically significant ( from k - s tests ) . we have shown that the 100 , 160 , and 250 to 24@xmath1 flux density ratios for dogs with moderate ir luminosities ( @xmath69 ) are statistically indistinguishable from those flux density ratios for galaxies with detected 24@xmath23 m emission that lie at similar redshifts and have similar ir luminosities , but that do not meet the dog selection criterion . most dogs have higher 2.2 to 0.65@xmath1 flux density ratios than do the control galaxies . thus it seems clear that dogs occupy the tail of a distribution of uv obscuration in ir - luminous galaxies at @xmath0 . we select a sample of dogs using deep images of the goods - n region . these dogs have lower ir luminosities than do dogs selected using shallower images of wider regions . for example , the @xcite sample in the wide botes region contains many dogs with @xmath70 . our sample contains very few galaxies with @xmath70 , and our conclusions regarding obscuration may not apply to dogs with such high luminosities . at low redshift , agn emission dominates the ir luminosity for most galaxies with @xmath70 @xcite . @xcite find that many dogs at @xmath0 with @xmath70 have featureless seds from observed - frame 3.6 to 8@xmath1 , indicating the presence of agn - heated dust . several studies @xcite find that , for these dogs , the seds at rest - frame mid- and far - ir wavelengths are similar to those of low redshift galaxies with ir luminosities dominated by agn emission , such as markarian 231 . only 2 ( of 59 ; 3% ) dogs , and 8 ( of 253 ; 3% ) control galaxies , have seds that increase ( in @xmath71 ) from observed - frame 3.6 to 8@xmath1 , and indeed both are among the dogs with the highest @xmath41 ( at @xmath72 and @xmath73 ) . two galaxies are not enough to allow us to rule out the possibility that dogs with @xmath74 differ in their 100 or 250 to 24@xmath1 flux density ratios from control galaxies with similar luminosities . most galaxies in our sample have @xmath69 . whether their ir luminosities are dominated by emission from agn or newly formed stars is not clear . @xcite and @xcite find that the average dog in this @xmath41 range has an x - ray spectrum with a power - law index they interpret as indicative of heavily obscured agn emission . an obscured agn would presumably contribute to the ir luminosity as well . @xcite find the same average x - ray index for a similar sample of dogs , but also find pah emission in the mid - infrared spectra of the dogs ; they conclude that heavily obscured agn emission can not coexist with pah emission . @xcite attribute the x - ray emission of the dogs to x - ray binaries rather than agn . in our sample , 5 ( of 61 ; 8% ) dogs and 31 ( of 268 ; 12% ) control galaxies have detected x - ray emission @xcite . for galaxies with @xmath75 , we do not find any evidence that agn emission at rest - frame mid - ir wavelengths is more or less common in dogs than in the control galaxies : the fractions of dogs and control galaxies with increasing seds between observed - frame 3.6@xmath1 and 8@xmath1 are similar , as are the fractions with detected x - ray emission . no matter what powers their ir luminosities , dogs in this @xmath41 range are abnormally uv faint , and are probably the most obscured galaxies . we have shown that the brightness of the emergent uv emission in @xmath0 dusty galaxies can vary considerably from galaxy to galaxy , even though the emergent uv emission is only a fraction of the ir emission . in a dusty galaxy at either low or high redshift , most of the intrinsic uv emission from newly formed stars and/or agn is absorbed by dust grains , which emit at ir wavelengths . we now ask : `` are the uv - emitting regions that we see from a dog or a control galaxy spatially coincident with the ir - emitting regions ? '' if yes , the variation in emergent uv emission is due to uv obscuration . if no , the emergent uv emission may come from stars in a `` normal '' galactic disk , whereas the newly formed stars ultimately responsible for the ir emission are completely obscured , perhaps in a compact , nuclear star - forming region . in this scenario , the variation in the rest - frame optical to uv flux density ratios might be due as much to differences in the stellar populations from galaxy to galaxy as to differences in obscuration @xcite . we can not yet efficiently spatially resolve the ir - emitting regions for large samples of high redshift dusty galaxies . two pieces of circumstantial evidence support the statement that the uv- and ir - emitting regions in these dusty galaxies are spatially coincident : 1 ) a minority of dogs and control galaxies deviate from the median specific sfr for galaxies at @xmath0 ; and 2 ) dogs and control galaxies define a continuous relation between their median ir to uv luminosity ratios and median uv continuum power - law indices . the specific star formation rate of a galaxy is the ratio of its sfr to its stellar mass . most star - forming galaxies at the same redshift fall on a tight relation between sfr and stellar mass ; this relation is referred to as the `` main sequence '' @xcite . @xcite find that sfr and stellar mass are directly proportional at all redshifts , and present the evolution of the median specific sfr for galaxies at @xmath76 . @xcite also isolate the minority of galaxies with specific sfrs much higher than the median values ( `` starbursts '' ) . here , we define starbursts as galaxies with specific sfrs higher than 3 times the median value at their redshift . a minority of dogs ( 5 of 25 ; 20% ) and control galaxies ( 18 of 72 ; 25% ) are starbursts ( fig . [ fig : mainseq ] ) . @xcite find that individual low redshift starbursts have compact uv - emitting regions ; the correspondence between deviation from the median specific sfr and the compact size of the uv - emitting region also holds for the average @xmath0 starburst . if this correspondence holds for individual @xmath0 galaxies , few dogs and control galaxies would have concentrated uv - emitting regions . these regions in the typical dog or control galaxy would be more widely distributed , and it would be less plausible that the uv - emitting regions occupy parts of the galaxy not occupied by ir - emitting regions . that radio - emitting regions in high redshift galaxies are widely distributed indirectly supports this idea @xcite . emission . the solid line is the median specific sfr for star - forming galaxies , as a function of redshift @xcite . the dotted line is three times the median value , which we use as a threshold for identifying starbursts . 20% of dogs and 25% of control galaxies are starbursts . dogs deviate from the median specific sfr no more frequently than do the control galaxies . [ fig : mainseq ] ] for starburst galaxies at low redshift , @xcite ( @xcite , and more recently * ? ? ? * ) find a relation between the ratio of ir to uv luminosities ( denoted irx ) and the power - law index of the sed in the uv ( denoted @xmath42 ) . ( @xcite refer to low redshift galaxies with bright uv - emitting regions as `` starbursts '' ; the term does not necessarily refer to galaxies that deviate from the main sequence . ) this relation is generally interpreted to mean that the ir emission originates as uv emission from newly formed stars , which is partially absorbed by dust . the dependence between the emergent uv emission , the dust emission in the ir , and @xmath42 , in the local irx-@xmath42 relation is consistent with the dependence of dust absorption on @xmath42 in the @xcite dust attenuation law . in other words , we think that galaxies that lie on this relation are thought to have spatially coincident uv- and ir - emitting regions . @xcite show that most lyman break galaxies ( lbg ) at @xmath0 lie on the local irx-@xmath42 relation . however , both @xmath0 lbgs and the low redshift starbursts in @xcite are much less dusty than dogs . some sub - classes of dusty galaxies deviate from the local relation , perhaps because their ir emission is unrelated to their emergent uv emission @xcite . [ fig : irxbeta ] shows the local irx-@xmath42 relation from @xcite and the relevant quantities for the @xmath0 dusty galaxies in our samples . individual galaxies , in both the dog and control samples , are found on either side of the relation . because dogs are so faint in the uv , it is difficult to precisely determine the power - law indices of their uv seds ; the large uncertainties on @xmath42 values preclude us from reaching firm conclusions about individual galaxies . the median @xmath42 values for the dog and control samples tend to lie to the left of the local irx-@xmath42 relation ; dogs do not appear to deviate more strongly from the relation than do control galaxies . these @xmath0 dusty galaxies define their own irx-@xmath42 relation . as noted earlier , specific sub - populations of dusty galaxies deviate from the local relation , but the trend that we find between the averages for a broader @xmath0 population has not been discussed . because the median @xmath42 value increases with increasing irx , from the lbgs to the dogs , the ir emission is not completely independent of the emergent uv emission in these galaxies . this reinforces our initial conclusion that @xmath0 dusty galaxies populate a continuum of uv obscuration , and that dogs are simply the most heavily obscured galaxies . over 0.16@xmath1 luminosity ( irx ) vs. uv continuum power - law index ( @xmath42 ) for galaxies with detected 100@xmath1 emission . we only show errors on the uv power - law indices . filled hourglasses and circles show the median @xmath42 values , in bins of irx ; the vertical bars show the range of irx over which the @xmath42 values are binned . the `` average lbg '' black asterisk is from @xcite , and shows a mean and its dispersion . the solid line is the local irx-@xmath42 relation from @xcite . [ fig : irxbeta ] ] dogs in the luminosity range spanned by our sample are more obscured than the control galaxies for a reason unrelated to their far - ir luminosity , because the @xmath41 distributions of the two samples are statistically indistinguishable . in galaxies with dust heated by uv emission from newly formed stars , the amount of obscuration affecting that uv emission can vary due to either : 1 ) differences in the degree of alignment between the spatial distributions of dust and newly formed massive stars ; or 2 ) differences in the total dust content . @xcite derive the local irx-@xmath42 relation by assuming a uniform screen of dust in the line of sight between us and the newly formed stars ( see figure 8 in @xcite for an illustration ) . @xcite show that this assumption can be recovered when the dust is instead distributed in clumps around star - forming regions . thus , dogs might be more obscured than the control galaxies because dogs have more dust clumps surrounding star - forming regions . variations in uv obscuration might also be caused by differences in the physical properties responsible for the shape of the sub - millimeter sed . for example , increases in the mass of cold dust can result in greater uv obscuration . increases in the cold dust mass can also result in increases in sub - millimeter emission . since the sub - millimeter luminosity is usually a small fraction of the total ir luminosity , the increase in total ir luminosity due to a higher cold dust mass would not be easily measurable between 100 and 250@xmath1 . we can not fully test this hypothesis without sub - millimeter luminosities for galaxies in our samples . the sensitivities of current sub - millimeter facilities limit comparisons of the dust content between dogs and control galaxies to the most luminous ( @xmath70 ) galaxies @xcite . currently available data do not allow us to discriminate between any hypotheses for the physical mechanisms responsible for either the differences in the degree of alignment between dust and stars or differences in the dust content . for instance , two plausible hypotheses are that : 1 ) galaxy inclination might be responsible for the patchiness of dust in the line - of - sight , and 2 ) many dogs are merging galaxies , and some aspect of the merging process creates differences in the degree of alignment . to falsify either we require high resolution rest - frame optical images ; the cosmic assembly near - infrared deep extragalactic legacy survey ( candels ) will obtain such images using _ hst_/wfc3 in the near future @xcite . we note that @xcite and @xcite examine rest - frame optical images of dogs in goods - s and conclude that most are undisturbed disk galaxies . furthermore , @xcite simulate isolated galaxies that meet the dog selection criterion , so dogs are not necessarily merging galaxies on the basis of their extreme rest - frame mid - ir to uv flux density ratios . finally , uv obscuration is affected by the prominence of the `` dust bump '' at rest - frame 2175 in a galaxy s attenuation curve . at @xmath0 , rest - frame 2175 is redshifted into the @xmath4-band filter . the presence of this feature in the attenuation curves of high redshift galaxies has been controversial ; the bump is found , though , in the attenuation curves for several samples of galaxies at @xmath77 @xcite . however , doubling the amplitude of the average bump leads to a reduction in the 0.65@xmath1 flux density by a factor of 1.3 . this is not large enough to explain the spread of @xmath67 in fig . [ fig : roverk_z ] . 1 . we cull a sample of dust - obscured galaxies ( dogs ) , or galaxies with @xmath35 that are at @xmath8 , in the goods - n region . we use deep goods-_herschel _ data to compare the emission from dogs with that from other @xmath0 galaxies with detected 24@xmath1 emission . the dogs in our sample span @xmath3 . dogs and control galaxies , with detected 100@xmath1 emission , have similar distributions of ir luminosities and stellar masses . we compare the rest - frame far - ir and optical flux densities of dogs with those of the control galaxies . dogs have extreme ratios of @xmath78 not because they are abnormally bright at rest - frame @xmath11 for their far - ir flux densities , but because they are abnormally faint in the rest - frame uv . 4 . dogs and the control galaxies scatter around the median specific sfr established by @xmath0 galaxies falling on the `` main sequence '' ; 20% of dogs have specific sfrs greater than 3 times the median value , and are thus starbursts . if uv - emitting regions in high - redshift starbursts are distributed as they are in low - redshift starbursts , few dogs have compact uv - emitting regions . 5 . for both the dog and control samples , the median rest - frame uv continuum power - law index ( @xmath42 ) at a given ir to uv luminosity ratio ( irx ) is lower than the index predicted by the local irx-@xmath42 relation . dogs do not appear to deviate more from this relation than do the control galaxies . over more than a factor of 100 in irx , the median @xmath42 value for these galaxies systematically increases with increasing irx . these pieces of evidence suggest that , for most of these galaxies , the uv- and ir - emitting regions are spatially coincident . thus , the range in rest - frame mid - ir to uv flux density ratios spanned by dusty galaxies at @xmath0 is due to differing amounts of uv obscuration . dogs are the most heavily obscured galaxies . differences in the amount of obscuration between dogs and other dusty galaxies at @xmath0 can be due to : 1 ) differences in the degree of alignment between the spatial distributions of dust and massive stars , or 2 ) differences in the total dust content . our samples do not have many galaxies with @xmath9 . in dogs with these ir luminosities , where agn emission may be the dominant source of dust heating , our conclusions about obscuration may not be valid . further information about the nature of obscuration in these galaxies can come from measurements that spatially resolve uv- and ir - emitting regions . for instance , the h@xmath79 to h@xmath42 line ratio varies with the amount of dust absorption , so its variation across a galaxy traces the spatial dust distribution . @xcite measure the galaxy - wide balmer decrement for several very luminous dogs ( those most likely to have ir luminosities dominated by agn emission ) ; @xcite present spatially resolved h@xmath79 maps , but h@xmath42 is outside their spectral window . _ hst _ can resolve the rest - frame uv and optical emission from these galaxies @xcite , and alma will resolve the ir emission and allow us to measure sub - millimeter luminosities we need to determine cold dust mass . studies using such data will clearly establish the degree to which the spatial distributions of dust and massive stars align . this work is based on observations made with _ , a european space agency cornerstone mission with significant participation by nasa . support for this work was provided by nasa through an award issued by jpl / caltech . the research activities of m.d . and a.d . are supported by noao , which is operated by the association of universities for research in astronomy under a cooperative agreement with the national science foundation . we are grateful to n. drory for sharing the sed - fitting code we use to estimate galaxy stellar masses .	dusty galaxies at @xmath0 span a wide range of relative brightness between rest - frame mid - infrared ( 8@xmath1 ) and ultraviolet wavelengths . we attempt to determine the physical mechanism responsible for this diversity . dust - obscured galaxies ( dogs ) , which have rest - frame mid - ir to uv flux density ratios @xmath2 , might be abnormally bright in the mid - ir , perhaps due to prominent agn and/or pah emission , or abnormally faint in the uv . we use far - infrared data from the goods-_herschel _ survey to show that most dogs with @xmath3 are not abnormally bright in the mid - ir when compared to other dusty galaxies with similar ir ( 81000@xmath1 ) luminosities . we observe a relation between the median ir to uv luminosity ratios and the median uv continuum power - law indices for these galaxies , and we find that only 24% have specific star formation rates which indicate the dominance of compact star - forming regions . this circumstantial evidence supports the idea that the uv- and ir - emitting regions in these galaxies are spatially coincident , which implies a connection between the abnormal uv faintness of dogs and dust obscuration . we conclude that the range in rest - frame mid - ir to uv flux density ratios spanned by dusty galaxies at @xmath0 is due to differing amounts of uv obscuration . of galaxies with these ir luminosities , dogs are the most obscured . we attribute differences in uv obscuration to either : 1 ) differences in the degree of alignment between the spatial distributions of dust and massive stars , or 2 ) differences in the total dust content .
the algorithm for the calculation of the ground states bases on a special genetic algorithm @xcite and on _ cluster - exact approximation _ ( cea ) @xcite which is a sophisticated optimization method . next a short sketch of these algorithms is given . the genetic algorithm starts with an initial population of @xmath24 randomly initialized spin configurations (= _ individuals _ ) , which are linearly arranged in a ring . then @xmath25 times two neighbors from the population are taken ( called _ parents _ ) and two offsprings are created using a triadic crossover : a mask is used which is a third randomly chosen ( usually distant ) member of the population with a fraction of @xmath26 of its spins reversed . in a first step the offsprings are created as copies of the parents . then those spins are selected , where the orientations of the first parent and the mask agree @xcite . the values of these spins are swapped between the two offsprings . then a _ mutation _ with a rate of @xmath27 is applied to each offspring , i.e. a fraction @xmath27 of the spins is reversed . next for both offsprings the energy is reduced by applying cea : the method constructs iteratively and randomly a non - frustrated cluster of spins . spins adjacent to many unsatisfied bonds are more likely to be added to the cluster . during the construction of the cluster a local gauge - transformation of the spin variables is applied so that all interactions between cluster spins become ferromagnetic . [ 0.5 ] to make all adjacent bonds positive . please note that many other combinations of spins can be used to build a cluster without frustration.,title="fig : " ] [ 0.5 ] to make all adjacent bonds positive . please note that many other combinations of spins can be used to build a cluster without frustration.,title="fig : " ] fig . [ fig_cea_example ] shows an example of how the construction of the cluster works using a small spin - glass system . for 3d @xmath0 spin glasses each cluster contains typically 58 percent of all spins . the non - cluster spins act like local magnetic fields on the cluster spins , so the ground state of the cluster is not trivial . since the cluster has only ferromagnetic interactions , an energetic minimum state for its spins can be calculated in polynomial time by using graph theoretical methods @xcite : an equivalent network is constructed @xcite , the maximum flow is calculated @xcite with level(@xmath28 ) = level(@xmath29)+1 , but also all edges @xmath30 where @xmath31 is the sink . for this measure , we observed an additional speed - up of roughly factor 2 for the systems we calculated . ] and the spins of the cluster are set to their orientations leading to a minimum in energy . this minimization step is performed @xmath32 times for each offspring . afterwards each offspring is compared with one of its parents . the pairs are chosen in the way that the sum of the phenotypic differences between them is minimal . the phenotypic difference is defined here as the number of spins where the two configurations differ . each parent is replaced if its energy is not lower ( i.e. not better ) than the corresponding offspring . after this whole step is done @xmath25 times , the population is halved : from each pair of neighbors the configuration which has the higher energy is eliminated . if more than 4 individuals remain the process is continued otherwise it is stopped and the best individual is taken as result of the calculation . the representation in fig . [ fig_algo ] summarizes the algorithm . the whole algorithm is performed @xmath33 times and all configurations which exhibit the lowest energy are stored , resulting in @xmath34 statistically independent ground - state configurations . this algorithm was already applied to examine the ground state structure of 3d spin glasses @xcite . = = = = = = genetic cea(@xmath35 , @xmath24 , @xmath33 , @xmath27 , @xmath32 ) + * begin * + create @xmath24 configurations randomly + ( @xmath36 ) * do * + + @xmath37 * to * @xmath25 * do * + + select two neighbors + create two offsprings using triadic crossover + do mutations with rate @xmath27 + both offsprings * do * + + @xmath38 * to * @xmath32 * do * + + construct unfrustrated cluster of spins + construct equivalent network + calculate maximum flow + construct minimum cut + set new orientations of cluster spins + + offspring is not worse than related parent + + replace parent with offspring + + + half population ; @xmath39 + + one configuration with lowest energy + * end * we used the simulation parameters determined in former calculations for @xmath15 : for each system size many different combinations of the simulation parameters @xmath40 were tried for some sample systems . the final parameters where determined in a way , that by using four times the numerical effort no reduction in energy was obtained . here @xmath41 and @xmath42 were used for all system sizes . for smaller concentrations @xmath2 the ground states are easier to find , because the number of frustrated plaquettes is smaller . but it was not possible to reduce the computational effort substantially in order to get still ground states . so we used the parameters of @xmath15 for all concentrations @xmath2 . table 1 summarizes the parameters . also the typical computer time @xmath43 per ground state computation on a 80 mhz ppc601 is given . using these parameters on average @xmath44 ground states were obtained for every system size @xmath11 using @xmath42 runs per realization . [ cols="^,^,^,^,^",options="header " , ] tab 1 . simulation parameters : @xmath11 = system size , @xmath24 = initial size of population , @xmath33 = average number of offsprings per configuration , @xmath32 = number of cea minimization steps per offspring , @xmath43 = average computer time per ground state on a 80mhz ppc601 . we compared our results for 180 sample systems of @xmath45 with exact ground states which were obtained using a branch - and - cut program @xcite . the genetic cea algorithm found the true ground states for all systems ! the same result was obtained for @xmath46 as well . the branch - and - cut program needs to much computer time because of the exponential time complexity . ] a more detailed analysis is presented in @xcite . so we can be sure that genetic cea and our method of choosing the parameters lead to true ground states or at least to states very close to true ground states . we performed ground state calculations for @xmath47 $ ] for lattice sizes @xmath48 . the number of independent realizations of the bond - disorder ranged from @xmath49 for @xmath50 to @xmath51 for smaller systems . most of the systems have af - bond concentrations around @xmath52 [ 0.45 ] per spin as function of af - bond concentration @xmath2 for system sizes @xmath53 . lines are guides for the eyes only.,title="fig : " ] the ground state energy @xmath54 as function of system size for different system sizes is shown in fig [ fig_energy ] . to keep the figure clear only the sizes @xmath53 are presented . for small concentrations @xmath2 the ground state is mainly ferromagnetic . it follows that all af bonds are not satisfied , so the ground state energy increases linearly like @xmath55 with @xmath2 . for larger concentrations the ground state energy approaches the @xmath15 limit , because the spins can arrange so that not all af - bonds are broken . with increasing @xmath11 the ground state energy decreases , because the periodic boundary conditions impose less constraints on the system . for @xmath56 and @xmath15 a ground state energy of @xmath57 is found in @xcite . [ 0.45 ] for system size @xmath58 . two typical error bars are given . lines are guides for the eyes only.,title="fig : " ] from fig . [ fig_energy ] is clear that the energy as function of concentration is not well suited for determining the critical concentration @xmath19 where the ferromagnetic behavior disappears . for this purpose the binder cumulant @xcite @xmath59 for the magnetization @xmath60 is used . the average @xmath61 denotes both average over different ground states of a realization and over the disorder . in fig . [ fig_binder ] the binder cumulant is shown for @xmath58 . l=@xmath62 are omitted in this figure to keep it clear . for the same reason only typical error bars for two sample points are shown . all curves intersect at @xmath63 . only @xmath46 ( not shown ) is little worse , because it meets the others in the interval @xmath64 $ ] . so we conclude that the critical concentration for the ferromagnetic order is @xmath65 . the value for @xmath19 is comparable to results from high - temperature series expansions : @xmath66@xcite , @xmath67 @xcite , from monte - carlo renormalization - group results @xmath68 @xcite and from monte - carlo simulations : @xmath69 @xcite . our value is much larger than a result from a zero - temperature expansion : @xmath70 @xcite and much lower than a recent result from a renormalization - group study : @xmath71 @xcite . since the intersection of the curves of the binder cumulant is very sharp , we believe that our result is very reliable , although the systems investigated here are rather small . [ 0.45 ] for the binder cumulant the following finite - size scaling relation is assumed @xcite @xmath72 by plotting @xmath73 against @xmath74 with correct parameter @xmath75 the datapoints for different system sizes should collapse onto a single curve . the best results were obtained for @xmath76 and @xmath77 . in fig . [ fig_binder_scale ] the resulting scaling plot is shown . it is possible to change the value of @xmath75 in a wide range without large effects on the scaling plot . so we estimate @xmath78 . this is consistent with @xmath79 which was found using monte - carlo simulations of spin glasses ( @xmath15 ) at finite temperature @xcite . the average magnetization @xmath80 has the standard finite - size scaling form @xcite @xmath81 by plotting @xmath82 against @xmath74 with correct parameters @xmath83 the datapoints for different system sizes should collapse onto a single curve . the best result was obtained using @xmath77 and @xmath84 . it is shown in fig . [ fig_magn_scale ] for @xmath58 . from variations of the value @xmath85 we estimate the value of the exponent @xmath86 . lines are guides for the eyes only.,title="fig : " ] to characterize the spin - glass behavior the overlap @xmath87 is used . it compares two different states @xmath88 of the same realization of the random bonds @xmath89 for ferromagnetic / antiferromagnetic order two independently calculated ground states are identical or related by a global flip of all spins , i.e. @xmath90 . for spin - glass order many different ground states exist @xcite , so @xmath87 can take also intermediate values @xmath91 $ ] . since the system has no external field , each state is equivalent to the state where all spins are reversed , so only the absolute value @xmath92 is considered here . [ 0.45 ] value as function of af - bond concentration @xmath2 for @xmath93 . with increasing concentration more and more spins belong to clusters which contribute to the degeneracy of the ground state , so the average overlap value decreases . where the onset of the spin glass behavior is located exactly can not be seen from this figure . lines are guides for the eyes only.,title="fig : " ] in fig . [ fig_av_q ] the average value of the overlap @xmath94 is shown for the lattice sizes @xmath93 . the decrease of @xmath95 with increasing concentration @xmath2 is clearly visible . but this quantity has much larger fluctuations than the magnetization , so it is difficult to use this data for further analysis . also the value is not equal to @xmath96 for a large interval , so it is not possible to extract at which concentration the onset of the spin - glass behavior is located . [ 0.45 ] of overlaps for af - bond concentrations about @xmath97 for @xmath93 . a finite fraction of spins is contained in small clusters which can take two orientations in the ground state . for @xmath56 a delta - function is obtained similar to the distribution found for the ground states of random - field ising systems . lines are guides for the eyes only.,title="fig : " ] more information can be obtained , if one calculates not only the average of @xmath92 , but its distribution @xcite @xmath98 which describes the ground - state structure . in fig . [ fig_p_q_p0.18]the distributions for sizes @xmath93 at @xmath97 are presented have exactly the same number of ferromagnetic bonds , for each system size only a finite number of different concentrations is possible . here for each lattice size the value of @xmath2 is chosen which is nearest to @xmath99 . with increasing size the distributions becomes narrower . the reason is , that due to the discrete bond distribution @xmath12 there are always some small clusters of spins , which can take two orientations in the ground state . with increasing system size @xmath11 these effects cancel out and @xmath100 converges to a delta - function @xmath101 , where @xmath102 is just the fraction of spins contained in such free clusters . this ground - state structure is similar to that of random - field ising systems @xcite . [ 0.45 ] of overlaps for af - bond concentrations about @xmath103 for @xmath93 . the distribution is broad and extends to @xmath7 for all sizes @xmath11 , indicating a spin - glass behavior . lines are guides for the eyes only.,title="fig : " ] in fig . [ fig_p_q_p0.23 ] the distributions of overlaps is displayed for a concentration slightly larger than @xmath19 . here the behavior is completely different . the distributions extend over a large interval down to @xmath7 . with increasing lattice size @xmath11 only the shape of the large - q part changes a little bit while for small values no systematic modification is visible . the peak at large @xmath87-values , which raises with increasing concentration @xmath2 , results from small clusters of spins which can take two orientations in the ground state . this is the same as for smaller concentrations of the af - bonds . but the large extent down to @xmath7 can not be explained in this way . it shows that spin - glass ground states have a very rich structure , similar to the behavior found for the sk - model @xcite , where each spin interacts with every other spin ( for a detailed discussion of the ground - state structure see @xcite ) . to investigate this behavior quantitatively the variance @xmath104 of the distributions as function of af - bond concentration @xmath2 and system size @xmath11 is calculated @xmath105 [ 0.45 ] of overlap distribution as function of af - bond concentration @xmath2 . for small concentrations the width of the distribution shrinks to zero with increasing size @xmath11 . for larger values the spin - glass phase is characterized by broad distributions of overlaps . lines are guides for the eyes only.,title="fig : " ] in fig . [ fig_sigma_q ] the result for @xmath93 is shown . for small concentration @xmath2 the width of the distributions is small and shrinks with increasing size @xmath11 . for larger concentrations the width increases and remains nonzero even for larger lattice sizes . in @xcite it was shown , that the spin glass ground state with @xmath15 is likely to have a broad distribution even for @xmath56 . for larger sizes the statistics is too bad to extract for example a critical concentration @xmath106 by a finite - size scaling analysis similar to that presented above . the onset of the spin glass behavior can even better be observed by calculating the fraction @xmath107 of the distribution of overlaps below a fixed value @xmath108 : @xmath109 [ 0.45 ] of overlap distribution below @xmath110 as function of af - bond concentration @xmath2 for sizes @xmath93 . in the spin - glass phase the distribution of overlaps extends to @xmath7 for all lattice sizes . lines are guides for the eyes only.,title="fig : " ] in fig . [ fig_x05 ] the value of @xmath111 is shown as function of @xmath2 for the lattice sizes @xmath93 . for the limiting case @xmath15 the value of @xmath111 converges to a nonzero value for @xmath56 @xcite . using a combination of a genetic algorithm and cluster - exact approximation ground states of the three - dimensional @xmath0 random - bond ising model were calculated for different concentrations of the antiferromagnetic bonds . a former comparison with exact ground states calculated using a branch - and cut program shows that genetic cea is able to calculate true ground states . for small concentrations the ground state is mainly ferromagnetic . the critical concentration where the ferromagnetic order disappears was determined using the binder - cumulant @xmath73 of the magnetization : @xmath112 this is the first time @xmath19 is calculated by direct investigations of ground states . using a finite - size scaling analysis of the magnetization and the binder - cumulant critical exponents were obtained : @xmath113 these values are consistent with results from monte - carlo simulations of the @xmath15 case at finite temperature , where the same analysis presented here was performed for the spin - glass order parameter @xmath87 instead of the magnetization @xmath114 . the onset of the spin - glass behavior with increasing lattice size was investigated the first time by calculating the distributions of overlaps as function of @xmath2 . the spin glass phase is characterized by a broad distribution of overlaps which extends down to @xmath7 and does not change substantially with increasing system size . for this quantity the bad statistics for larger values of the concentration @xmath115 does not allow to determine a second critical concentration @xmath106 , so it is yet not possible to check whether the ferromagnetic order disappears at the same concentration where the spin - glass phase appears . here much more data are needed . the author thanks h. horner and g. reinelt for manifold support . he is grateful to m. jnger , m. diehl and t. christof who put a branch - and - cut program for the exact calculation of spin - glass ground states of small systems at his disposal . he thanks r. khn for critical reading of the manuscript and for giving many helpful hints . this work was engendered by interesting discussions with j. bendisch at the `` algorithmic techniques in physics '' seminar held at the _ international conference and research center for computer science schloss dagstuhl _ in wadern / germany . the author took much benefit from discussions with s. kobe , h. rieger and a.p . he thanks f. hucht for supplying the program _ fsscale _ for performing the finite - size scaling analysis . he is also grateful to the _ paderborn center for parallel computing _ for the allocation of computer time . this work was supported by the graduiertenkolleg `` modellierung und wissenschaftliches rechnen in mathematik und naturwissenschaften '' at the _ interdisziplinres zentrum fr wissenschaftliches rechnen _ in heidelberg .	large numbers of ground states of the three - dimensional @xmath0 random - bond ising model are calculated for sizes up to @xmath1 using a combination of a genetic algorithm and cluster - exact approximation . several quantities are calculated as function of the concentration @xmath2 of the antiferromagnetic bonds . the critical concentration where the ferromagnetic order disappears is determined using the binder cumulant of the magnetization . a value of @xmath3 is obtained . from the finite - size behavior of the binder cumulant and the magnetization critical exponents @xmath4 and @xmath5 are calculated . the behavior of the distribution of overlaps @xmath6 is used to investigate how the spin - glass phase evolves with increasing concentration @xmath2 . the spin - glass order is characterized by a broad distribution of overlaps which extends down to @xmath7 . * keywords ( pacs - codes ) * : spin glasses and other random models ( 75.10.nr ) , numerical simulation studies ( 75.40.mg ) , general mathematical systems ( 02.10.jf ) . [ [ introduction ] ] introduction + + + + + + + + + + + + in this work systems of @xmath8 spins @xmath9 , described by the hamiltonian @xmath10 are investigated . the spins are placed on a three - dimensional ( d=3 ) cubic lattice of linear size @xmath11 with periodic boundary conditions in all directions . systems with quenched disorder of the nearest - neighbor interactions ( bonds ) are investigated . their possible values are @xmath12 . the concentration of the antiferromagnetic ( af ) bonds ( @xmath13 ) is denoted with @xmath2 , all other @xmath14 interactions are ferromagnetic . a constrained disorder is used , so that the fraction of the antiferromagnetic bonds is exactly @xmath2 for all realizations of the disorder . the model shows a complex behavior for low temperatures . for large concentrations of the ferromagnetic bonds it is energetically favorable for two interacting spins to have the same orientation . so the system shows ferromagnetic order , which means that most of the spins have the same value . for large concentrations @xmath2 the system is antiferromagnetically ordered : the system can be divided into to penetrating sublattices and each sublattice has ferromagnetic order , but the sign of the ordering of the two sublattices is different . for intermediate concentrations of the antiferromagnetic bonds neither ferromagnetic nor antiferromagnetic order exists . the system is called a spin glass @xcite . for finite - dimensional spin glasses no final agreement about their behavior exists . recent results from simulations@xcite of small systems with @xmath15 up to size @xmath16 indicate that the three - dimensional spin glass has a complex behavior : the ( free ) energy landscape has many stable configurations which differ strongly from each other . whether the onset of this spin - glass behavior takes place at the same concentration where the ferromagnetic order disappears is unclear . here the ground - state , i.e. zero temperature ( @xmath17 ) , behavior of the model as function of the concentration @xmath2 is investigated . since the phase diagram of the system is symmetrical to @xmath15 if one identifies the ferromagnetic with the antiferromagnetic regime , only values @xmath18 are used . of special interest is the critical concentration @xmath19 where the ferromagnetic order disappears . in the past the @xmath0 random - bond ising model has been studied using monte - carlo simulations@xcite , zero - temperature series expansions @xcite , high - temperature series expansions @xcite , monte carlo renormalization - group calculations @xcite and renormalization - group theory @xcite . all of these results are qualitatively consistent with the phase - diagram described above , but no agreement is found on the value of the critical concentration @xmath19 or its temperature dependence . in all of these publications the detailed structure of the states was not investigated , which can be done for example by calculating the distribution of overlaps . a study of this model using true ground states has not been performed before . only for special two - dimensional systems , where exact ground states can be calculated efficiently , some results @xcite are known . there the critical concentration of a square lattice is estimated to be @xmath20 . the behavior of the @xmath0 random bond ising model is determined by the occurrence of _ frustration _ @xcite . the simplest example of a frustrated system is a triple of spins where all pairs are connected by antiferromagnetic bonds . it is not possible to find a spin - configuration were all bonds contribute with a negative value to the energy . one says that it is not possible to _ satisfy _ all bonds . in general a system is frustrated , if closed loops of bonds exists , where the product of these bond - values is negative . for square and cubic systems the smallest closed loops consist of four bonds . they are called ( elementary ) _ plaquettes_. the presence of frustration makes the calculation of exact ground states of such systems computationally hard . only for the special case of the two - dimensional system with periodic boundary conditions in no more than one direction and without external field a polynomial - time algorithm is known @xcite . for the general case the simplest method works by enumerating all @xmath21 possible states and has obviously an exponential time complexity . even a system size of @xmath22 is too large . the basic idea of an algorithm called _ branch - and - bound _ @xcite is to exclude the branches of the search - tree , where no low - lying states can be found , so that the complete low - energy landscape of systems of size @xmath22 can be calculated @xcite . a more sophisticated method called _ branch - and - cut _ @xcite works by rewriting the problem as a linear optimization problem with an additional set of inequalities which must hold for the solution . since not all inequalities are known a priori the method iteratively solves the linear problem , looks for inequalities which are violated , and adds them to the set until the solution is found . since the number of inequalities grows exponentially with the system size the same holds for the computation time of the algorithm . with branch - and - cut anyway small systems up to @xmath23 are feasible . the method used here is able to calculate true ground states up to size @xmath1 . by studying ground states one does not encounter ergodicity problems or critical slowing down like in monte - carlo simulations . since it is possible to compare with exact results from branch - and - cut calculations no uncontrolled approximations are used . the only uncertainty comes from the fact that one is restricted to relatively small systems . in the next section the algorithm is explained . then all results are presented . in the last section a conclusion is driven .
the laser interferometer space antenna ( lisa ) is a planned nasa esa gravitational - wave ( gw ) observatory sensitive in the @xmath3@xmath4 hz range @xcite . lisa s data will contain superposed signals from millions of sources , including all the binaries in the galaxies with orbital periods below five hours and massive black - hole ( mbh ) binary coalescences out to @xmath5 @xcite . thousands of sources will be resolvable individually . the potential for source confusion and the very complex dynamics and waveforms of sources such as extreme mass - ratio inspirals ( emris ) , suggested the need for a coordinated effort to develop and demonstrate lisa data - analysis capabilities . the mock lisa data challenges ( mldcs ) began in early 2006 with this very purpose . the complexity and ambition of the challenges has risen with each round : mldc 1 @xcite focused on simple sources , isolated or moderately interfering ; mldc 2 @xcite introduced a galactic _ ensemble _ of 26 million binaries ( 20,000 of which were successfully recovered ) , as well as the problem of detecting mbh binaries over the galactic cacophony ; mldc 1b @xcite reprised the first challenge for new research collaborations joining the effort , and saw the first successful detections of emri signals . mldc 3 , released in april 2008 and due in april 2009 , consisted of five subchallenges that featured more realistic models of previously examined sources ( chirping galactic binaries in mldc 3.1 , spinning mbh binary inspirals in 3.2 , superposed emris in 3.3 ) and entirely new sources ( gw bursts from cosmic - string cusps in 3.4 , an isotropic stochastic background in 3.5 ) ; see @xcite for a detailed discussion of the source models and gw content of each subchallenge . fifteen collaborations , comprising all the participants listed in the byline and most task - force members , submitted a total of seventeen entries ( all can be found at www.tapir.caltech.edu/~mldc/results3 together with technical notes about search implementation ) . in this paper we briefly report on the detection and parameter - estimation performance demonstrated by each entry . altogether , mldc 3 showed substantial progress in tackling increasingly difficult data - analysis problems , and introduced new search methods such as nested sampling and sophisticated genetic optimization markov chain hybrids . however , there is certainly room for improvement and further work : fewer galactic binaries were recovered by the searches employed here than by the multi - source mcmc demonstrated in mldc 2 ; mbh binaries and emris were detected with high confidence , but the accurate estimation of their parameters ( beyond the dominant ones ) was stymied by the complex global structure and the many local maxima of likelihood surfaces . on the bright side , searches for cosmic - string bursts and stochastic backgrounds ( admittedly simple problems in the absence of nonstationary or non - gaussian instrument noise ) met no roadblocks . section [ sec : mldc4 ] of this article introduces mldc 4 , which is being released as we write ( november 2009 ) , with entries due at the end of 2010 . challenge dataset 3.1 contained signals from over 60 million chirping galactic binaries . the vast majority of these are too weak to be isolated , and the unresolved component forms a nonstationary confusion noise that adds to the overall noise level . estimates based on self - consistent removal schemes @xcite and bayesian model selection @xcite suggest that it should be possible to recover between 20,000 and 30,000 binaries . three groups submitted source catalogs for mldc 3.1 : * * bhamuib * ( a collaboration between the universities of birmingham and the balearic islands ) implemented a delayed - rejection mcmc algorithm @xcite to search three narrow frequency windows , @xmath6 , @xmath7 , and @xmath8 , using the mldc waveform generator @xcite . a total of 494 sources were reported . * * aeirit * ( researchers at the albert einstein institute in hannover , germany , and the rochester institute of technology ) set up a ligo - style hierarchical search based on the @xmath9-statistic and on frequency - domain rigid - adiabatic templates @xcite . triggers are generated for the individual tdi channels ; those found in coincidence are analyzed coherently using noise - orthogonal tdi observables . a total of 1940 sources were reported . * * powrwa * ( a collaboration between the albert einstein institute , the university of wrocaw , and the polish academy of sciences ) adopted an iterative matched - filtering search that used the @xmath9-statistic and rigid - adiabatic templates @xcite , and analyzed a few 0.1-mhz wide frequency bands . the brightest source in each band is identified and removed , and the process repeated until a pre - set snr threshold is reached . a total of 14,838 sources were reported . the entries for this round fell short of the theoretical target for a variety of reasons . bhamuib analyzed only a small fraction of the data , while aeirit and powrwa used single - pass or iterative search schemes , which are limited in how deep they can dig before source confusion degrades signal recovery . previous studies @xcite indicate that it should be possible to recover approximately 99% of the resolvable sources to an accuracy of better than 90% , as measured by the overlap between injected and recovered waveforms . we therefore adopt correlation as the metric by which the entries are measured . for this we need to identify the injected signal that corresponds most closely to each recovered signal . to do this , we consider all injected signals with @xmath10 within six frequency bins of the recovered signal , and we select the injected signal with the minimum @xmath11 , where @xmath12 and @xmath13 are the injected and recovered waveforms , and @xmath14 denotes the standard noise - weighted inner product , summed over the noise - orthogonal @xmath15 , @xmath16 , and @xmath17 tdi channels . the correlation is then computed as @xmath18 . using as a figure of merit the percentage of recovered sources with @xmath19 , we find 30% for bhamuib , 95% for aeirit , 33% for powrwa . powrwa have reported a bug in their code that affects sources with frequencies above 3 mhz . results are better after correcting it @xcite . if we consider only the 6,955 sources they report below 3 mhz , their figure of merit improves to 58% . challenge dataset 3.2 contained five signals from mbh inspirals , embedded in instrument noise and a partially subtracted galactic background ( see @xcite for a synopsis of the waveform model and for the random selection of source parameters ) . multi - tdi snrs ranged from @xmath20 to @xmath21 , with the two weakest signals ( mbh-2 and mbh-6 ) corresponding to mergers after the end of the dataset . five groups submitted entries : * * aei * ( researchers at the albert einstein institute ) used a genetic matched - filtering algorithm extended to a multi - modal search ; as a merit function they took the @xmath15-statistic , a geometrical mean of the log - likelihood for the signal and for a low - frequency signal subset @xcite . * * cambaei * ( a collaboration between the cambridge cavendish laboratory and institute of astronomy with the the albert einstein institute ) adopted the multinest implementation of nested - sampling integration @xcite to compute the evidence and produce posterior distributions . this method is intrinsically multimodal , and the @xmath15-statistic was used to identify modes . * * gsfc * ( researchers at goddard space flight center ) employed the tempered metropolis - hastings mcmc algorithm found in xspec @xcite . * * jplcitnwu * ( a collaboration between jpl / caltech and northwestern university ) employed a two - stage method whereby they followed up a nonspinning mbh search @xcite with multinest . * * mtgwagapc * ( researchers at montana state university and at the astro - particle and cosmology institute , apc , in paris ) adopted a parallel - tempering mcmc algorithm , evolved from the thermostated , frequency - annealed algorithm employed in previous challenges @xcite . all five groups recovered parameters for the loud signals ; mbh-2 ( snr @xmath22 19 ) was recovered by aei , cambaei , jplcitnwu and mtgwagapc ; mbh-6 ( snr @xmath22 13 ) by aei , cambaei and mtgwagapc . participants were encouraged to submit multiple modes of comparable probability , if present . for the best ( highest - snr ) mode in each entry , table [ tab : smbh_err ] lists the fractional parameter errors , as well as the recovered snr and individual tdi - channel fitting factors ( ffs ) . both the recovered snr and the @xmath23 of the true waveform were computed by filtering the noisy dataset ( in its lisa simulator version ) with the appropriate template , while the ffs were computed between noiseless signals . note that the reported modes had often snr differences of less than one , and the highest - snr mode did not always have the best parameters . l@r@\|@r@r@r@r@r@r@r@\|@r@r@r source & group & @xmath24 & @xmath25 & @xmath26 & @xmath27 & @xmath28 & @xmath29 & @xmath30 & snr & @xmath31 & @xmath32 + @xmath33 & & @xmath34 & @xmath35 & ( sec ) & ( deg ) & @xmath36 & @xmath36 & @xmath37 + * mbh-1 * & aei & 2.4 & 6.1 & 62.9 & 11.6 & 7.6 & 47.4 & 8.0 & 1657.71 & 0.9936 & 0.9914 + ( 1670.58 ) & cambaei & 3.4 & 40.7 & 24.8 & 2.0 & 8.5 & 79.6 & 0.7 & 1657.19 & 0.9925 & 0.9917 + & mtapc & 24.8 & 41.2 & 619.2 & 171.0 & 13.3 & 28.7 & 4.0 & 1669.97 & 0.9996 & 0.9997 + & jpl & 40.5 & 186.6 & 23.0 & 26.9 & 39.4 & 66.1 & 6.9 & 1664.87 & 0.9972 & 0.9981 + & gsfc & 1904.0 & 593.2 & 183.9 & 82.5 & 5.7 & 124.3 & 94.9 & 267.04 & 0.1827 & 0.1426 + * mbh-3 * & aei & 9.0 & 5.2 & 100.8 & 175.9 & 6.2 & 18.6 & 2.7 & 846.96 & 0.9995 & 0.9989 + ( 847.61 ) & cambaei & 13.5 & 57.4 & 138.9 & 179.0 & 21.3 & 7.2 & 1.5 & 847.04 & 0.9993 & 0.9993 + & mtapc & 333.0 & 234.1 & 615.7 & 80.2 & 71.6 & 177.2 & 16.1 & 842.96 & 0.9943 & 0.9945 + & jpl & 153.0 & 51.4 & 356.8 & 11.2 & 187.7 & 414.9 & 2.7 & 835.73 & 0.9826 & 0.9898 + & gsfc & 8168.4 & 2489.9 & 3276.9 & 77.9 & 316.3 & 69.9 & 95.6 & 218.05 & 0.2815 & 0.2314 + * mbh-4 * & aei & 4.5 & 75.2 & 31.4 & 0.1 & 47.1 & 173.6 & 9.1 & 160.05 & 0.9989 & 0.9994 + ( 160.05 ) & cambaei & 3.2 & 171.9 & 30.7 & 0.2 & 52.9 & 346.1 & 21.6 & 160.02 & 0.9991 & 0.9992 + & mtapc & 48.6 & 2861.0 & 5.8 & 7.3 & 33.1 & 321.1 & 33.0 & 149.98 & 0.8766 & 0.9352 + & jpl & 302.6 & 262.0 & 289.3 & 4.0 & 47.6 & 184.5 & 28.3 & 158.34 & 0.8895 & 0.9925 + & gsfc & 831.3 & 1589.2 & 1597.6 & 94.4 & 59.8 & 566.7 & 95.4 & @xmath38 & @xmath39 & @xmath40 + * mbh-2 * & aei & 1114.1 & 952.2 & 38160.8 & 171.1 & 331.7 & 409.0 & 15.3 & 20.54 & 0.9399 & 0.9469 + ( 18.95 ) & cambaei & 88.7 & 386.6 & 6139.7 & 172.4 & 210.8 & 130.7 & 24.4 & 20.36 & 0.9592 & 0.9697 + & mtapc & 128.6 & 45.8 & 16612.0 & 8.9 & 321.4 & 242.4 & 13.1 & 20.27 & 0.9228 & 0.9260 + & jpl & 287.0 & 597.7 & 11015.7 & 11.8 & 375.3 & 146.3 & 9.9 & 18.69 & 0.9661 & 0.9709 + * mbh-6 * & aei & 1042.3 & 1235.6 & 82343.2 & 2.1 & 258.2 & 191.6 & 26.0 & 13.69 & 0.9288 & 0.9293 + ( 12.82 ) & cambaei & 5253.2 & 1598.8 & 953108.0 & 158.3 & 350.8 & 215.4 & 29.4 & 10.17 & 0.4018 & 0.4399 + & mtapc & 56608.7 & 296.7 & 180458.8 & 119.7 & 369.2 & 297.6 & 25.1 & 11.34 & -0.0004 & 0.0016 + the binaries that merge within the duration of the dataset ( mbh-1 , 3 , and 4 ) have significantly larger snr and reach higher frequencies , and thus allow a much more accurate recovery of source parameters . for these binaries , the mass , time - of - coalescence , distance , and sky - position errors are comparable , if not better , to those of nonspinning - binary searches ( see tables 1 and 2 of @xcite ) . the errors in the spin amplitudes are consistent with the fisher - matrix results of @xcite . however , despite the high snr , the estimation of the initial direction of the spins and of the orbital angular momentum has proved to be very difficult . we observe a large number of the local likelihood maxima that have comparable , high snr ( and @xmath41 ) , yet very different values of these parameters , as illustrated in fig . [ fig : smbh_spinldeg ] for the mbh-3 entries . parameter estimation was not as successful for the weaker signals with mergers beyond the end of the dataset . still , for mbh-2 the errors in the masses and time of coalescence are comparable to the fisher - matrix predictions . the errors in sky position are @xmath0 deg , with strong local likelihood maxima at the antipodal sky position . spin amplitudes are determined very poorly ; this reflects the fact that the spins are nearly degenerate with other parameters in the low - frequency part of the waveforms . lang and hughes @xcite report that spin - induced modulations remove correlations between parameters in fisher - matrix computations , improving overall parameter determination . however , here we observe that spin interactions also cause nonlocal degeneracies in parameters space , especially so for spin and orbital angular - momentum angles . further investigations are needed to determine which phenomenon is stronger . nevertheless , the entries to mldc 3.2 demonstrate a solid detection capability for spinning - binary inspirals , and a good recovery of most source parameters . : true value ; @xmath42 : aei ; @xmath43 : cambaei ; @xmath44 : mtgwagapc ; @xmath45 : jplcitnwu.[fig : smbh_spinldeg],title="fig : " ] + challenge dataset 3.3 contained five barack cutler @xcite emri signals immersed in instrument noise ( see @xcite for details about the waveforms and the random choice of parameters ) . in comparison to previous emri challenges , here participants had to contend with multiple simultaneous sources , as well as weaker signals the injected snrs varied between @xmath46 and @xmath47 . three groups submitted entries : * * babakgair * ( a collaboration between the albert einstein institute and cambridge university ) used stochastic sampling and mcmc to identify emri harmonics , then carried out an @xmath48-statistic search in the space of harmonics , and performed a final mcmc fit in source parameter space . this search improved on the method described in @xcite , adding more sophisticated harmonic identification . * * etfag * ( researchers at cambridge and northwestern universities ) searched for harmonics in the time frequency spectrogram using the chirp - based algorithm for track search ( cats , @xcite ) developed for earlier mldcs @xcite , and improved for this search to deal with intersecting tracks from multiple sources . * * mtapcioa * ( a collaboration between montana state university , apc paris , and cambridge university ) improved the mldc algorithm used in previous rounds @xcite to include parallel tempering as described in @xcite , and to enhance the implementation of `` harmonic jumps '' between secondary likelihood maxima . parameter - estimation errors are presented in table [ tab : emri_err ] . altogether , all emris were found by at least one group , and masses were estimated accurately ; more work remains to be done on improving the estimates of emri parameters for relatively weak and overlapping signals . mtapcioa recovered all five emri signals , generally with very good parameter - estimation accuracies : errors of a few tenths % in the masses of both bodies , and sky - position errors of a few deg ( @xmath49 for emri-4 ) . their second solution for emri-1 was particularly impressive , with fractional errors of a few @xmath3 in masses , initial eccentricity , and spin . however , their other solution exemplifies the difficulty of resolving secondary maxima : the snr is almost the same ( @xmath50 vs. @xmath51 ) , but parameter errors are two orders of magnitude greater . l@l\|r@r@r@r@r@r@r@r@r@r source & group & snr & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58spin & @xmath58sky & @xmath59 + ( @xmath23 ) & & & @xmath36 & @xmath36 & @xmath34 & @xmath36 & @xmath36 & @xmath36 & ( deg ) & ( deg ) & + * emri-1 * & mtapcioa & 21.794 & 5.05 & 3.29 & 1.61 & @xmath60 & @xmath61 & @xmath62 & 23 & 2.0 & 0.07 + ( 21.673 ) & mtapcioa & 21.804 & @xmath63 & @xmath64 & @xmath65 & @xmath66 & 0.02 & 0.54 & 3.5 & 1.0 & 0.13 + * emri-2 * & mtapcioa & 32.387 & @xmath67 & @xmath68 & @xmath69 & 3.8 & 0.87 & 12 & 11 & 3.7 & @xmath70 + ( 32.935 ) & babakgair & 22.790 & 33.1 & @xmath71 & 10.1 & @xmath72 & @xmath73 & 250 & 47 & 3.5 & @xmath74 + & babakgair & 22.850 & 32.7 & @xmath75 & 9.94 & @xmath76 & @xmath77 & 250 & 58 & 3.5 & @xmath78 + & babakgair & 22.801 & 33.5 & @xmath79 & 10.5 & @xmath72 & @xmath80 & 240 & 40 & 3.5 & @xmath74 + * emri-3 * & mtapcioa & 19.598 & 1.62 & 0.38 & @xmath81 & @xmath82 & @xmath83 & @xmath84 & 5.0 & 3.0 & @xmath85 + ( 19.507 ) & babakgair & 21.392 & 1.77 & 1.01 & 1.95 & @xmath86 & @xmath87 & @xmath88 & 116 & 4.5 & 0.13 + & babakgair & 21.364 & 2.26 & 1.88 & 2.71 & @xmath89 & @xmath90 & @xmath91 & 65 & 6.1 & 0.14 + & babakgair & 21.362 & 1.51 & 1.01 & 2.09 & @xmath92 & @xmath93 & @xmath94 & 7.6 & 6.2 & 0.14 + & etfag & & 54.0 & 4.88 & @xmath95 & 26 & 17 & & & 32 & 0.83 + * emri-4 * & mtapcioa & @xmath96 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & 144 & 950 & 99 & 13 & @xmath88 + ( 26.650 ) & + * emri-5 * & mtapcioa & 17.480 & @xmath101 & 5.00 & @xmath102 & 0.22 & 55 & 62 & 43 & 1.8 & @xmath92 + ( 36.173 ) & + the time frequency analysis carried out by etfag was particularly hard - hit by the simultaneous lowering of the snr and the presence of multiple overlapping signals in mldc 3.3 . the group was unable to find the low - frequency , low - snr emri-1 ; while etfag did find a medium - frequency source , the relatively large parameter - estimation errors suggest that the time frequency approach did not adequately resolve between the overlapping harmonics of emri-2 and emri-3 . babakgair submitted three ( relatively close ) solutions for each of emri-2 and emri-3 . for emri-3 , their estimates are better than those of etfag and comparable to those of mtapcioa , albeit somewhat less accurate for the initial parameter values and distance . for emri-2 , babakgair had errors of a few % in the masses and initial eccentricity , and had significant errors in spin - orientation angle and in distance , although sky location was still found correctly , within @xmath103 deg . the high - frequency emri-4 and emri-5 presented a challenge for all groups . although mtapcioa found them and estimated their masses fairly accurately , the errors in spin magnitude and orientation were significantly larger than for other sources , and the distance to both sources was overestimated by factors of @xmath104 or @xmath105 ( vs. errors @xmath106 for the other emris ) . furthermore , the negative snr for claimed emri-4 and the low ffs between the recovered and injected noiseless waveforms indicate that the mtapcioa search could not resolve these sources individually , but converged on two parameter sets that jointly fit the combination of the two injected sources . l@r@\|@r@r@r@r@\|@r@r@r source & group & @xmath58sky & @xmath107 & @xmath108 & @xmath109 & snr & @xmath31 & @xmath32 + @xmath33 & & ( deg ) & ( sec ) & ( rad ) & & & & + * string-1 * & cam & 106.9 & 1.462 & 0.501 & 0.904 & 43.706 & 0.99947 & 0.99797 + ( 43.46 ) & cam & 49.4 & 2.331 & 1.065 & 1.128 & 43.520 & 0.99964 & 0.99591 + & jplcit & 34.2 & 1.585 & 3.726 & 0.413 & 43.506 & 0.99986 & 0.99844 + & jplcit & 113.7 & 1.574 & 3.739 & 0.431 & 43.497 & 0.99988 & 0.99847 + & mtgwag & 106.6 & 2.071 & 2.600 & 0.745 & 43.287 & 0.99975 & 0.99565 + * string-2 * & cam & 82.0 & 3.683 & 4.846 & 0.062 & 33.690 & 0.99945 & 0.99986 + ( 33.6 ) & jplcit & 90.5 & 4.005 & 4.268 & 0.282 & 33.689 & 0.99949 & 0.99929 + & jplcit & 45.2 & 3.847 & 6.364 & 0.231 & 33.694 & 0.99939 & 0.99960 + & mtgwag & 53.1 & 3.223 & 0.158 & 0.011 & 33.696 & 0.99926 & 0.99978 + * string-3 * & cam & 80.8 & 1.249 & 3.785 & 0.338 & 41.326 & 0.99073 & 0.99923 + ( 41.42 ) & cam & 133.3 & 1.715 & 3.257 & 0.238 & 41.456 & 0.99388 & 0.99869 + & cam & 44.5 & 0.763 & 3.202 & 0.066 & 41.142 & 0.99700 & 0.99883 + & jplcit & 59.0 & 1.546 & 3.129 & 0.317 & 41.315 & 0.99554 & 0.99848 + & jplcit & 157.7 & 1.226 & 5.614 & 0.220 & 41.316 & 0.99717 & 0.99864 + & mtgwag & 137.9 & 0.980 & 0.110 & 0.161 & 41.418 & 0.99327 & 0.99948 + challenge dataset 3.4 contained three burst signals from cosmic - string cusps , immersed in instrument noise with slightly randomized levels for each individual noise ( i.e. , from the six proof masses and photodetectors ) . the dataset was less than a month long ( @xmath110 s ) , with a higher sampling rate ( 1 s ) than the others , to accommodate the potential high - frequency content of these signals , which have power - law spectrum up to an @xmath111 determined by the characteristic length scale of the string and the viewing angle ( see @xcite for more details about the waveforms and the random choice of their parameters ) . four collaborations submitted entries : * * cam * ( a collaboration between cambridge u. and apc paris ) used multinest . * * canoe * ( researchers at cambridge and northwestern universities ) implemented a time frequency algorithm , a modified version of cats @xcite . * * jplcit * ( caltech / jpl ) experimented with mcmc and multinest , but only submitted entries based on the latter @xcite . * * mtgwag * ( montana state university ) used a parallel - tempering mcmc @xcite . all groups successfully recovered all three bursts . table [ tab : parerrs ] shows the parameter - estimation errors , snrs , and ffs for all entries . although the accuracy of parameters is poor , the ffs are very high ; this suggests that these results are not due to shortcomings in the search methods , but to the very character of the waveforms . as a matter of fact , for relatively short signals such as these bursts , lisa can be considered as a static detector , and its response is not imprinted with any modulations from the lisa orbit or from the rotation of its constellation . hence the sky position of burst sources can only be determined by triangulation between the spacecraft a weaker effect , and one that vanishes in the limit of long wavelengths . thus , the determination of sky position is intrinsically harder for bursts , and it is further complicated by the presence of degeneracies @xcite such as the reflection of sky position across the instantaneous lisa plane . the bursts barycentric central time @xmath112 and its polarization are strongly coupled with sky position , and therefore are also determined poorly . to compensate for this fact , we calculate the error in the arrival time of the burst at the center of the detector constellation @xmath113 . this parameter has a weaker correlation with the sky position and constrained better by observations . challenge dataset 3.5 contained an isotropic stochastic background signal immersed in instrument noise , with noise levels slightly randomized as in mldc 3.4 . to facilitate the use of the gw - canceling tdi combination @xmath17 , the lisa orbits were approximated as those traced by a rigidly rotating triangle , with equal and constant arm lengths ( except for the sagnac effect ) . the background was realized by placing @xmath114 linearly polarized , stochastic pseudosources at positions uniformly distributed across the sky . see @xcite for more details . now , isotropic stochastic backgrounds can be characterized by a single dimensionless quantity , @xmath115 where @xmath116 is the energy density in gws , and @xmath117 is the energy density required to close the universe . in this mldc round , @xmath118 was taken to be constant across frequencies , and in this dataset it was set equal to @xmath119 ( with @xmath120 ) . two groups submitted entries for this challenge , and both analyzed the same version of the dataset , generated by synthetic lisa @xcite : * * aeibham * ( a collaboration between the albert einstein institute and the university of birmingham ) submitted multiple entries using different applications of the same mcmc algorithm @xcite . here we focus on the aeibham analysis of the tdi @xmath15 and @xmath16 observables , which was run separately over frequency bands of 0.11 mhz ( a ) and 0.15 mhz ( b ) . * * mtgwag * ( montana state university ) analyzed the tdi @xmath15 , @xmath16 , and @xmath17 spectra with parallel - tempering mcmc . they estimated the background level as well as the levels of select linear combinations of the individual lisa noises ( other combinations are left essentially undetermined by tdi observations ) . figure [ fig : stochastic_pdf ] shows the reported posterior pdfs and the best - fit values for @xmath118 , along with their fractional error ; both groups recovered the injected value within less than @xmath121 . figure [ fig : stochastic_pdf ] shows also the best - fit levels reported by mtgwag for the determinable proof - mass and photodetector noise combinations . the aeibham analyses used theoretical expressions for tdi @xmath15 and @xmath16 noise spectra , which left only their overall levels as unknowns . however , this treatment assumed that the secondary noises were symmetric , which was not the case in this dataset . in addition , because of limitations in the mldc stochastic - background generation code , the actual gw strain spectrum differed from the nominal @xmath122 power law at higher frequencies . this may explain why the aeibham ( b ) result was further from the injected value that either the aeibham ( a ) result ( in which band the gw spectrum was actually @xmath122 ) or the mtgwag result , which used all three tdi channel , solved for their unknown correlations , and calibrated the gw spectrum using training datasets . while the third round of the mldcs was focused on increasing the complexity and variety of gw sources , we are devoting the next iteration to the _ global - fit _ problem of detecting and analyzing sources of different types superposed in the lisa data . thus , mdlc 4 consists of a single , `` whole enchilada '' challenge that includes all the sources of mldcs 3.13.5 _ in the same dataset _ , albeit with larger source numbers ( for emris and cosmic - string bursts ) and parameter ranges ( for mbh binaries and emris ) . see table [ table : mldc4 ] for details . the duration of the dataset is again approximately two years ( @xmath123 s ) , but the data are released both as a high - cadence time series of @xmath124 samples with @xmath125 s , and as a downsampled time series of @xmath126 samples with @xmath127 s ( however , only the cosmic - string bursts and instrument noise contribute spectral content above 33 mhz ) . reported by each analysis , their modes @xmath128 , and the corresponding fractional error w.r.t . the true @xmath118 . the horizontal - axis range corresponds to the `` prior '' mldc range . table on right : estimated lisa noise levels ( modes ) in the mtgwag analysis . the @xmath129 and @xmath130 are the proof mass noises introduced in the left and right optical assemblies on lisa spacecraft @xmath131 ; likewise for photodetector noises @xmath132 and @xmath133 . the combinations of noises shown in the table are the only ones that are constrained effectively by the data after laser phase noise has been removed with tdi.[fig : stochastic_pdf ] ] l@+l\|ll & mtgwag / true & frac . error + pm@xmath134 & pm@xmath135 & @xmath136 & @xmath137 + pm@xmath138 & pm@xmath139 & @xmath140 & @xmath141 + pm@xmath142 & pm@xmath143 & @xmath144 & @xmath145 + pd@xmath134 & pd@xmath135 & @xmath146 & @xmath147 + pd@xmath138 & pd@xmath139 & @xmath148 & @xmath149 + pd@xmath142 & pd@xmath143 & @xmath150 & @xmath151 + three versions of both datasets are released : fractional - frequency fluctuation data generated by synthetic lisa @xcite and lisacode @xcite , and equivalent - strain data generated by the lisa simulator @xcite . in contrast to mldcs 3.4 and 3.5 , all the lisa instrument noises are set to symmetric , nominal levels , and the lisa armlength are allowed to `` breathe . '' see @xcite for more details about the mldc models of lisa orbits , noises , and interferometric observables . the waveform models of mldc 3 @xcite are used unchanged in mldc 4 , except for the following : the spectra of cosmic - string bursts are truncated below @xmath3 hz ; the stochastic - background spectrum is @xmath122 between @xmath3 hz and 33 mhz , and drops lower below and above that range ; last , when mbh binaries end within the duration of the dataset , they are terminated at the time @xmath152 when the frequency derivative changes sign , indicating that the pn expansion is failing . to reduce spectral leakage , a half hann window @xmath153 $ ] is applied between @xmath154 and @xmath152 , with @xmath155 , @xmath111 the frequency at time @xmath152 , and @xmath156 the quality factor ( lisaxml parameter ` q ` ) set to 3 for mldc 4 ( setting ` q`=0 yields the old @xmath157 termination condition ) . l@l _ galactic - binary background _ & @xmath158 interacting , @xmath159 detached systems + 46 _ mbh binaries _ & @xmath160 , @xmath161 , @xmath162 , + & @xmath163 , with @xmath164 and snrs as in mldc 3.2 + an average of 6 _ emris _ & @xmath165 , @xmath166 , @xmath167 , + & @xmath168 s , snr @xmath169 + & * poisson(2 ) * systems with @xmath170 + & * poisson(2 ) * systems with @xmath171 + & * poisson(2 ) * systems with @xmath172 + * poisson(20 ) * _ cosmic - string bursts _ & @xmath173 , @xmath174 s , @xmath175 + _ isotropic stochastic background _ & @xmath176 + challenges beyond mldc 4 will feature ever more realistic datasets , including more sophisticated waveform models , such as mbh coalescences with merger and ringdown phases , and `` real - mission '' noise models and instrument events drawn from the ongoing experimental investigations of the lisa architecture and subsystem . to obtain more information and to participate in the mldcs , see the official mldc website , astrogravs.nasa.gov/docs/mldc , the task force wiki , www.tapir.caltech.edu/listwg1b , and the ` lisatools ` software repository @xcite . the authors acknowledge support from the german aerospace center ( sb , ap , jtw ) ; the max - planck society ( sb , jtw , ap , rp , elr ) ; the lisa mission science office at jpl ( cc , mv ) ; mnisw grant n n203 387237 ( ab , ak ) ; nasa grants nng04gd52 g ( mb ) , nng05gi69 g and nnx07aj61 g ( njc ) , 06-befs06 - 31 ( mv and cc ) , nnx07ah22 g ( i m ) , and the nasa center for gw astronomy at the u. of texas , brownsville ; nsf grant phy-0855494 ( jtw ) , and the nsf a&a postdoctoral fellowship under ast-0901985 ( i m ) ; the science and technology facilities council ( av ) ; st catharine s college , cambridge ( jg ) ; e.u . feder funds , spanish ministerio de ciencia e innovacin projects fpa2007 - 60220 , ha2007 - 0042 and csd2007 - 00042 , govern de les illes balears , conselleria deconomia , hisenda i innovaci ( mt ) . cc s and mv s work was carried out at the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . copyright 2009 . all rights reserved . 99 bender p , danzmann p and the lisa study team 1998 `` laser interferometer space antenna for the detection of gravitational waves , pre - phase a report '' * mpq 233 * ( garching : max - planck - institt fr quantenoptik ) baker j g et al . 2007 _ phys . d _ * 75 * 124024 arnaud k a et al . ( the mldc task force ) 2006 _ laser interferometer space antenna : 6th international lisa symp . ( greenbelt , md , 1923 jun 2006 ) _ ed merkowitz s m and livas j c ( melville , ny : aip ) p 619 ; _ ibid _ p 625 ( long version with lisaxml description : gr - qc/0609106 ) arnaud k a et al . ( the mldc task force and challenge 1 participants ) 2007 _ class . * 24 * s529 arnaud k a et al . ( the mldc task force ) 2007 _ class . grav . _ * 24 * s551 babak s et al . ( the mldc task force and challenge 2 participants ) 2008 _ class . grav . _ * 25 * 114037 babak et al . 2008 _ class . grav . _ * 25 * 184026 timpano s e , rubbo l j and cornish n j 2006 _ phys . d _ * 73 * 122001 crowder j and cornish n j 2007 _ class . * 24 * s575 trias m , vecchio a and veitch j 2009 arxiv:0904.2207 cornish n j and littenberg t b 2007 _ phys . rev . d _ * 76 * 083006 whelan j t , prix r and khurana d 2009 arxiv:0908.3766 ; 2008 _ class . quantum grav . _ * 25 * 184029 ; prix r and whelan j t 2007 _ class . * 24 * s565 baut a , babak s and krlak a 2009 arxiv:0911.3020 petiteau a , shang y , babak b and feroz f _ in preparation _ bridges m , feroz f and hobson m p 2009 _ mnras _ * 398 * 1601 arnaud k a et al . , xspec website , http://heasarc.gsfc.nasa.gov/xanadu/xspec brown d a et al . 2007 _ class . * 24 * s595 cornish n j and porter e k 2006 _ class . * 24 * 5729 lang r n and hughes s a 2006 phys . d * 74 * 122001 barack l and cutler c 2004 _ phys . _ d * 69 * 082005 babak s , gair j r and porter e 2009 _ class . * 26 * 135004 . mandel i and gair j r _ in preparation _ gair j r , mandel i and wen l 2008 _ class . grav . _ * 25 * 184031 cornish n j 2008 arxiv:0804.3323 shapiro key j and cornish n j 2009 _ phys . d _ * 79 * 043014 cornish n j and rubbo l j 2003 _ phys . d _ * 67 * 022001 ; erratum - ibid . 029905 ; lisa simulator website , www.physics.montana.edu/lisa ; included in @xcite vallisneri m 2005 _ phys . d _ * 71 * ; synthetic lisa website , www.vallis.org/syntheticlisa ; included in @xcite cohen m , cutler c and vallisneri m _ in preparation _ robinson e l , romano j d and vecchio a 2008 _ class . grav . _ * 25 * 184019 petiteau a et al . 2008 _ phys . d _ * 77 * 023002 ; lisacode website , www.apc.univ-paris7.fr/lisa-france/analyse.phtml ; included in @xcite lisatools website and svn repository , lisatools.googlecode.com	the mock lisa data challenges are a program to demonstrate lisa data - analysis capabilities and to encourage their development . each round of challenges consists of one or more datasets containing simulated instrument noise and gravitational waves from sources of undisclosed parameters . participants analyze the datasets and report best - fit solutions for the source parameters . here we present the results of the third challenge , issued in apr 2008 , which demonstrated the positive recovery of signals from chirping galactic binaries , from spinning supermassive black - hole binaries ( with optimal snrs between @xmath0 and @xmath1 ) , from simultaneous extreme mass - ratio inspirals ( snrs of 1050 ) , from cosmic - string cusp bursts ( snrs of 10100 ) , and from a relatively loud isotropic background with @xmath2 , slightly below the lisa instrument noise .
radial abundance gradients in the milky way disk are among the main constraints of models of the chemical evolution of the galaxy . the study of the gradients comprises the determination of their magnitudes along the disk , space variations and their time evolution ( see for example henry & worthey 1999 , maciel & costa 2003 ) . probably the most interesting property of the gradients is their time evolution , which is a distinctive constraint of recent chemical evolution models . maciel et al . ( 2003 ) suggested that the o / h gradient has been flattening during the last few gyr , on the basis of a large sample of planetary nebulae ( pn ) for which accurate abundances are available , and for which the ages of the progenitor stars have been individually estimated . this work has been recently extended ( maciel et al . 2005 ) to include the s / h ratio in planetary nebulae , [ fe / h ] metallicities from open clusters and cepheid variables , as well as some young objects , such as ob associations and hii regions . in this work , we review the main characteristics of the work by maciel et al . ( 2005 ) and analyze the uncertainties involved in the determination of the gradients . in particular , we investigate whether the derived uncertainties support either a systematic variation of the abundances with the galactocentric distance , as assumed by our work , or simply a dispersion of the abundances around some average value . the main results for the time variation of the gradients as derived from planetary nebulae , open clusters , and cepheids are shown in tables 1 and 2 . adopting average linear gradients , which can be taken as representative of the whole galactic disk , the abundances can be written in the form where @xmath2(o / h ) + 12 or @xmath2(s / h ) + 12 for pn , hii regions and ob stars , and @xmath3 [ fe / h ] for open clusters and cepheids . for planetary nebulae , we have taken into account both o / h and s / h determinations and evaluated the gradient in the galactic disk according to the ages of the progenitor stars . for comparison purposes , we can also derive the [ fe / h ] metallicities from the o / h abundances , on the basis of a [ fe / h ] @xmath4 o / h correlation derived for disk stars ( see maciel 2002 and maciel et al . 2005 for details ) . the ages follow from the age - metallicity relation by edvardsson et al . ( 1993 ) , which also depends on the galactocentric distance . in this way , we can divide the sample of pn into different age groups , each one having a characteristic gradient . table 1 shows representative examples of 3 age groups for o / h and 2 age groups for s / h . the table gives the gradient @xmath5 ( dex / kpc ) as defined by equation ( 1 ) . all gradients in this paper have been calculated assuming @xmath6 kpc for the galactocentric distance of the lsr . for detailed references on the pn data the reader is referred to maciel et al . ( 2003 , 2005 ) . it should be mentioned that the pn age groups shown in table 1 are typical groups , arbitrarily defined . in fact , we have extended this procedure by taking into account a variety of definitions of the age groups , with similar results . column 2 of table 4 shows the estimated values of @xmath7 and @xmath8 [ within brackets ] assuming average values , that is , no linear variations . the results for pn show that the probability is very low in all cases , so that the data points are probably not distributed according to a gaussian distribution around some average value . however , it is interesting to note that , if we restrain the galactocentric distances to a smaller range , such as from @xmath9 kpc to 8 kpc , or @xmath10 kpc to 10 kpc , the probability @xmath8 increases , showing that , for a given galactocentric bin , the abundances show a better agreement with the gaussian distribution around some average value . for the open clusters , the table shows a generally better agreement with the gaussian distribution around a mean value , both for the friel and chen samples , in agreement with our conclusions in sect . however , for cepheid variables we have the same results as for the pn , that is , the cepheid data are apparently not consistent with a gaussian distribution around a mean value . we can also estimate @xmath8 in each case taking into account the derived linear correlations which are displayed in tables 1 and 2 . here we have @xmath11 for the number of degrees of freedom , so that we can estimate @xmath7 and @xmath8 provided we have a reliable estimate of the uncertainty of the data . for planetary nebulae , recent discussions by pottasch et al . ( 2005 ) of objects with iso data suggest that the abundances of the beststudied elements are probably correct within 20% , which corresponds to 0.10 dex for oxygen . this is probably a lower limit for other nebulae for which no infrared data is available , so that their abundances depend more heavily on ionization correction factors . we may then adopt @xmath12 dex for o / h and @xmath13 dex for s / h as realistic estimates for planetary nebulae . the latter can also be attributed to the open clusters , in view of the heterogeneity of the data and the use of photometric abundances . for cepheid variables , which have the best determinations , an average uncertainty @xmath14 seems appropriate . the results are shown in column 3 of table 4 , under the heading linear . again the probabiliy is given within brackets . we can see that in all cases the @xmath7 values are lower than the corresponding values for the averages , so that the probability @xmath8 is higher for the linear correlation than for the simple averages . in fact , these probabilities are very close to unity in most cases , especially if we consider the more realistic , higher uncertainties . it can also be seen that for cepheid variables the probability given in column 3 is essentially unity , reinforcing our conclusion about systematic abundance variations with the galactocentric distance .	we have studied the time variation of the radial abundance gradients using samples of planetary nebulae , open clusters , cepheids and other young objects . based on the analysis of o / h and s / h abundances for planetary nebulae and metallicities of the remaining objects , we concluded that the gradients have been flattening out in the last 8 gyr with an average rate of the order of 0.005 0.010 dex kpc@xmath0 gyr@xmath0 . we have estimated the errors involved in the determination of the gradients , and concluded that the existence of systematic abundance variations is more likely than a simple statistical dispersion around a mean value . address = iag / usp , so paulo , brazil
the author thanks prof . j. katriel for helpful comments on the manuscript . j. katriel , through correspondence , showed that eqns [ eq : evenzeros ] and [ eq : oddzeros ] agree with the first asymptotic term from dominici for @xmath51 for low @xmath40 ( not for maximal @xmath40 ) . the latter result makes sense from the point of view that the maximal @xmath8 is always close to the edge of the hilbert space where the wavefunction goes to zero for any finite @xmath12 whereas that of the harmonic oscillator decays forever . [ eq : evenzeros ] and [ eq : oddzeros ] do not agree with higher order terms ( w.r.t . @xmath52 ) in dominici s asymptotic expansion .	the holstein - primakoff representation for spin systems is used to derive expressions with solutions that are conjectured to be the zeros of hermite polynomials @xmath0 as @xmath1 . this establishes a correspondence between the zeros of the hermite polynomials and the boundaries of the position basis of finite - dimensional hilbert spaces . the hermite polynomials are prevalent in many fields . they can be defined as @xmath2 in the physics community , they are perhaps best recognized as the gaussian - weighted eigenfunctions ( in position representation ) of the quantum harmonic oscillator ( with @xmath3 , a convention that will be used for the rest of the paper ) : @xmath4 as such , they are orthogonal over the gaussian - weighted whole domain , @xmath5 . this last property allows their use in gaussian quadrature , a useful and popular numerical integration technique where @xmath6 is approximated as @xmath7 where @xmath8 are the zeros of @xmath0 and @xmath9 is a well - behaved function . for this and many other reasons , an analytic formula for the asymptotic zeros of hermite and other orthogonal polynomials has been a subject of much interest@xcite , especially in the applied mathematics community and the field of approximation theory . in this paper , i examine the position state representation of the eigenstates of finite dimensional @xmath10-spin systems , as expressed in the holstein - primakoff transformation . as @xmath11 , the system becomes the infinite dimensional harmonic oscillator . this association allows me to derive the simple main results presented in eqs [ eq : evenzeros ] and [ eq : oddzeros ] , with solutions that i conjecture become the asymptotic zeros of the hermite polynomials ( as @xmath1 ) . furthermore , i numerically show that this convergence is rather quick and so the expressions can frequently be used , in many instances of finite - precision application , as the effective zeros of @xmath0 with finite @xmath12 , such as in applications of gaussian quadrature . in a more aesthetic sense , these results also establish a beautiful correspondence between the boundaries of equal area partitions of circles with radii that are increasing in a certain manner and the hermite polynomial zeros . spin systems are defined by the fundamental commutation relations between operators @xmath13 , @xmath14 and @xmath15 : @xmath16 = \hat s^+ , \hskip 5pt \left [ \hat s^z , \hat s^- \right ] = \hat s^- , \hskip 5pt \left [ \hat s^+ , \hat s^- \right ] = 2 \hat s^z.\ ] ] associating a spin with a boson @xmath17 , holstein and primakoff showed that to satisfy these commutation relations , the operators can be expressed as@xcite @xmath18 @xmath19 this is a very useful association and has found many applications in the condensed matter field s study of many - body spin systems . each boson excitation represents the `` ladder up '' finitesimal excitation away from the spin s extremal @xmath10 state . the hilbert space is finite - dimensional and possesses @xmath20 states @xmath21 . in fact , considering eq . [ eq : hkladder ] it is clear that the hilbert space outside this defined space is not even hermitian . transforming from the holstein - primakoff bosonic representation to position ( and its conjugate momentum ) space ( using the relations @xmath22 and @xmath23 ) reveals that the trivial hamiltonian is the harmonic oscillator : @xmath24 . moreover , transformation of the @xmath14 and @xmath15 in eq . [ eq : hkladder ] reveals that the hilbert space spans the domain @xmath25 . just as in the @xmath26 representation , @xmath27 states all with the same area must exist within this domain . fig . [ fig : qbasis ] sketches out what they look like for the @xmath28-spin systems . -basis representation of a ) @xmath29 , b ) @xmath30 and c ) @xmath31 systems is shown . the radius of the hilbert space s domain is equal to @xmath32 and so grows along with the number of allowed basis elements . ] for a particular @xmath10-spin system , the lowest eigenstate must have the same sign at all @xmath33-basis elements since it must be nodeless . on the other hand , the highest eigenstate must have @xmath34 nodes and so the @xmath33-basis elements must alternate in sign such that the eigenfunction passes through zero between them . this latter behavior is sketched in fig . [ fig : qbasis ] in red by the hermite polynomial @xmath0 denoting the value of the overlying @xmath33-basis element for the highest eigenstate . for @xmath11 , the hilbert space becomes infinite - dimensional and the hamiltonian becomes that of the harmonic oscillator defined over @xmath35 with the associated eigenfunctions proportional to @xmath36 . it therefore follows that as @xmath11 , the boundaries of the @xmath33-basis elements become the zeros of the hermite polynomial @xmath0 where @xmath37 since the highest eigenstate must still have alternating sign with each @xmath33-basis element . hermite polynomial zeros @xmath8 are real and symmetric around @xmath38 . to determine these boundary points , the @xmath39-dimensional hilbert space s circular shape in position space can be exploited . for even @xmath39 , the area of the all the @xmath33-basis elements up until the @xmath40th boundary ( measuring from the origin ) is @xmath41 . for odd @xmath39 , the area is @xmath42 . this is illustrated in fig . [ fig : areas ] . or b ) @xmath43 @xmath33-basis elements that approximately determine the @xmath40th zero of the hermite polynomial @xmath0 for @xmath12 even and odd respectively is shaded in blue . the approximate @xmath40th zero is at the right boundary of these regions . ] using simple relations for the area of circle sectors and rectangles , it is possible to relate these @xmath33-basis element areas to @xmath8 ; the equation involving the approximate zeros of hermite polynomials @xmath44 with @xmath12 even is : @xmath45 while for odd @xmath12 it is : @xmath46 where @xmath47 and @xmath48 . solving these equations for @xmath8 yields the approximate @xmath40th zero for the @xmath12th hermite polynomial . the results for the zeros of the first @xmath49 hermite polynomials are compared to the exact zeros in fig . [ fig : zeros ] . in both cases , eqs . [ eq : evenzeros ] and [ eq : oddzeros ] converge to the zeros of the hermite functions quite quickly and [ eq : oddzeros ] agree with the first asymptotic term from dominici@xcite for @xmath1 for low @xmath40 ( not for maximal @xmath40 ) . the latter result makes sense from the point of view that the maximal @xmath8 is always close to the edge of the hilbert space where the wavefunction goes to zero for any finite @xmath12 whereas that of the harmonic oscillator decays forever . eqs . [ eq : evenzeros ] and [ eq : oddzeros ] do not agree with higher order terms ( w.r.t . @xmath50 ) in dominici s asymptotic expansion . ] . th zeros of the hermite polynomials @xmath0 for @xmath12 a ) even and b ) odd compared to those obtained from solving eqs . [ eq : evenzeros ] and [ eq : oddzeros].,title="fig : " ] th zeros of the hermite polynomials @xmath0 for @xmath12 a ) even and b ) odd compared to those obtained from solving eqs . [ eq : evenzeros ] and [ eq : oddzeros].,title="fig : " ] the finding that the boundaries of equal area partitions of growing circles correspond to the asymptotic zeros of the hermite functions appears to be a novel one from a search of the literature . it is all the more surprising that the origin of this one - to - one correspondance stems from the holstein - primakoff representations for finite - dimensional spin systems . furthermore , on a practical level , the apparently rapid convergence of these solutions suggests that they may be useful for more efficient determination of hermite polynomial zeros for large - dimensional implementations of gaussian quadrature .
since cabibbo first noted the difference between the decays of neutron ( @xmath3 ) and muon ( @xmath4 ) and suggested there is a mixing angle between d and s quarks@xcite , and the mixing perfectly explained the data of the @xmath5decays of @xmath6 and @xmath7 @xcite . later the ckm matrix @xcite has been proposed to mix the three generation quarks . in the @xmath8 matrix there are three mixing angles and a cp phase which seem to be completely independent of each other . the mixing is understood as that the eigen - basis of weak interaction is not the same as that of mass , so matching them , a unitary transformation matrix must be introduced@xcite . from then on , the research field about quark mixing has been thoroughly investigated and exploration of its source has never ceased . in analog to the quark sector , the pontecorvo - maki - nakawaga - sakata ( pmns ) matrix @xcite relates the lepton flavor eigenstates with the mass eigenstates . thus it is natural to consider that there might be an underlying symmetry which results in the practical ckm and pmns matrices after symmetry breaking . recent studies on these matrices indicate that there exist the quark - lepton complementarity and self - complementarity@xcite which hint an existence of a higher symmetry . all the progress in this area inspires a trend of searching for whether such a hidden symmetry indeed exists and moreover investigation of its phenomenological implication is also needed . the key point is to investigate whether there exist some relations among the matrix elements of the ckm and/or pmns matrices which seem to be completely independent if there is no such a symmetry to make an arrangement . there are nine parametrization schemes which manifest the mixing in different ways . therefore , we expect that some relations among the parameters of these nine schemes might hint a hidden symmetry if they indeed exist . listing the parameters ( the mixing angles and cp phase ) of the nine schemes in a table and staring on them , we notice that there are relations among the sine values of the cp phases in these schemes . namely , for the nine parametrization schemes , we find equalities among those @xmath9 i.e. @xmath10 and @xmath11 where the the subscripts @xmath12 through @xmath13 refer to the nine parametrization schemes . considering the relations not to be accidental , we would be tempted to believe there is a hidden symmetry . then associating the expressions of jarlskog invariant which include sine values of the cp phases in all schemes with the experimentally measurable ckm matrix elements @xmath14 and using the above relations , we establish several equalities which do not depend on any concrete parametrization scheme at all . in principle when we apply the equalities to a special scheme , the solutions of those equalities would give the values of the parameters of the concerned scheme up to a universal dimensionless constant . what is the universal dimensionless constant ? surely it must be a scheme - independent quantity and has clear physics significance . naturally , one can conjecture that the jarlskog invariant@xcite which is related to the cp violation of hadrons is the best choice . thus we would accept the allegation . however , we also notice that all the equalities only approximately hold , even though the coincidence is very high . this can be well understood that the hidden symmetry is slightly broken , but some characteristics of the original symmetry is partly retained after the breaking . the approximate equalities are listed in the context of this work . we generalize the relations among the ckm matrix to the pmns case and find that all the aforementioned equalities also hold for the lepton sector , even though the accuracy is not as high as for the quark sector . along the line , we further investigate the induced phenomenological implication which may be tested in more accurate neutrino experiments . the paper is organized as follows . after the introduction we present those relations in section ii . in section iii , we check these relations numerically . in section iv we will discuss the implications about the possible hidden symmetry and draw our conclusion . in this section we show how to obtain the relations among the elements of the ckm and as well the pmns matrices . mixing among different flavors of quarks ( leptons ) via the ckm ( pmns ) matrix has been firmly recognized and widely applied to phenomenological studies of hadronic processes . the lagrangian of the weak interaction reads @xmath15 where @xmath16 , @xmath17 , @xmath18and @xmath19 . @xmath20 and @xmath21 are the ckm and pmns matrices respectively . the @xmath8 mixing matrices are written as @xmath22 generally , for a @xmath8 unitary matrix there are four independent parameters , namely three mixing angles and one cp - phase . there can be various schemes to parameterize the matrix and in literature , nine different schemes are presented and widely applied . they are clearly listed in ref.@xcite . here we try to collect those elements for various parametrization schemes in tab . [ tab1 ] where @xmath23 and @xmath24 are the mixing angles and cp - phase . for clarity , we use @xmath25 and @xmath26 to denote the corresponding quantities in the pmns matrix . [ cols="^,^,^,^",options="header " , ] + ( a ) ( b ) then we extend the same calculation to the pmns case . with the mixing angles @xmath27 and @xmath28 in p@xmath2 parametrization which are experimentally measured with certain errors , we compute corresponding @xmath29 which depends on @xmath30 and the result is presented in fig . [ fig : dep](b ) . it is also noted , the value of @xmath29 does not vary much within a relatively wide range @xmath31 to @xmath32 centered at @xmath1 . therefore , it is hard to firmly fix the cp phase at the lepton sector with this @xmath29 method , and one needs to wait for the more accurate long - baseline neutrino experiments in the future to determine its value . our estimate merely indicates its possible range . as is well known , the ckm and pmns matrices emerge due to the mismatch between the mass eigenstates and flavor eigenstates in the quark and lepton sectors respectively . the masses of different generations are not the same and compose the hierarchy problem . that is one of the unknown parts in particle physics . there must be sort of association between the mass hierarchy and mixing matrix . in a recent paper xing@xcite reviews the situation and extends the discussion from the quark sector to the lepton sector . merlo also discusses the neutrino masses and mixing from continuous symmetries@xcite . especially , the mass hierarchy problem of neutrinos i.e. it is either normal or inverted , is seriously considered and the juno experiment which will operate in china in a few years , will determine it . there are still more mysteries for the mixing , for example , the complementarity relations between elements of ckm and pmns matrices and the self - complementarities among the matrix elements of pmns @xcite . it was studied that the obvious symmetry in pmns can be traced to the possible symmetric textures for lepton masses@xcite , thus whether a fundamental symmetry really exists and what it is , is still an unsolved problem in our theoretical prospect . therefore , we are tempted to investigate the ckm and pmns matrices from a new angle and hope to gain a better understanding of the possible underlying theories . we note several relations among the ckm matrix parameters of the nine parametrization schemes i.e. @xmath33 , by which we establish several scheme - independent equalities . those relations may imply existence of something which is independent of any concrete parametrization . the jarlskog invariant is a physically measurable quantity , so must be the same for all the schemes which include the sine values of the cp phase @xmath24 . by the observation , it was suggested that the mixing matrices originate from a higher symmetry@xcite , and the nine parametrization schemes are merely various representations of the unique physical mixing matrix , therefore the relations among the sine values of the cp phases of the nine schemes actually reveal the inherent essence of the symmetry . we check these equalities numerically and find the first four hold with very high precision . moreover , if the underlying symmetry really exists , the elements of the ckm matrix which seem to be in an anarchy state at first glimpse , might be well arranged according to a certain rule . thus we would see whether the ckm matrix elements can be completely determined from those equalities up to a universal constant , which is the jarlskog invariant . on the other aspect , most of the symmetries in the nature are somewhat broken . mostly , the breaking might not be too large , so people still can trace back to the original symmetry . for example , we proposed that breaking of the complementarity and self - complementarity between ckm and pmns matrices is due to the involvement of a sterile neutrino@xcite . here we can observe that the relations of the ckm matrix elements hold with high precision , but an approximation degree is also noted as @xmath34 . especially , the fifth relation listed in eq.([rl4 ] ) deviates from equality by @xmath35 . it is naturally conjectured that the relations obtained for the ckm matrix can also be applied to the pmns matrix . because the pmns matrix elements are not well measured as that of the ckm matrix , we can only expect that the relations among the pmns matrix elements only approximately hold . for the lepton case we can also estimate @xmath30 in terms of the @xmath29 method , the resultant @xmath36 . but as shown in the fig . [ fig : dep](b ) , the value @xmath29 does not vary much from @xmath31 to @xmath32 . therefore , we can not firmly determine the cp phase for the lepton sector as much as we do for the quark sector from the relations yet . our observation on the relations indicates that a symmetry at higher scale does exist and after running down to the low energy scale its essential characters remain . therefore , even though it is slightly broken , we still can trace back to the original symmetry . further studies on the higher symmetry is definitely needed to eventually understand these relations . this work is supported by the national natural science foundation of china ( nnsfc ) under the contract no . 11375128 and 11135009 . m. kobayashi and t. maskawa , prog . phys . * 49 * , 652 ( 1973 ) . h. fritzsch , nucl . b * 155 * , 189 ( 1979 ) . l. f. li , phys . b * 84 * , 461 ( 1979 ) . h. fritzsch and z. z. xing , phys . b * 413 * , 396 ( 1997 ) [ hep - ph/9707215 ] . m. raidal , phys . lett . * 93 * , 161801 ( 2004 ) [ hep - ph/0404046 ] . g. altarelli , f. feruglio and l. merlo , jhep * 0905 * , 020 ( 2009 ) [ arxiv:0903.1940 [ hep - ph ] ] ; g. altarelli and d. meloni , j. phys . g * 36 * , 085005 ( 2009 ) [ arxiv:0905.0620 [ hep - ph ] ] ; r. de adelhart toorop , f. bazzocchi and l. merlo , jhep * 1008 * , 001 ( 2010 ) [ arxiv:1003.4502 [ hep - ph ] ] ; g. altarelli , f. feruglio , l. merlo and e. stamou , jhep * 1208 * , 021 ( 2012 ) [ arxiv:1205.4670 [ hep - ph ] ] . y. -j . zheng and b. -q . ma , eur . j. plus * 127 * , 7 ( 2012 ) [ arxiv:1106.4040 [ hep - ph ] ] ; x. zhang and b. -q . ma , phys . d * 86 * , 093002 ( 2012 ) [ arxiv:1206.0519 [ hep - ph ] ] ; h. qu and b. -q . ma , phys . d * 88 * , 037301 ( 2013 ) [ arxiv:1305.4916 [ hep - ph ] ] . n. haba , k. kaneta and r. takahashi , europhys . * 101 * , 11001 ( 2013 ) [ arxiv:1209.1522 [ hep - ph ] ] . c. jarlskog , phys . lett . * 55 * , 1039 ( 1985 ) ; d. d. wu , phys . d * 33 * , 860 ( 1986 ) ; o. w. greenberg , phys . d * 32 * , 1841 ( 1985 ) . f. vissani , hep - ph/9708483 . v. d. barger , s. pakvasa , t. j. weiler and k. whisnant , phys . b * 437 * , 107 ( 1998 ) [ hep - ph/9806387 ] . a. j. baltz , a. s. goldhaber and m. goldhaber , phys . * 81 * , 5730 ( 1998 ) [ hep - ph/9806540 ] . j. beringer _ [ particle data group collaboration ] , phys . d * 86 * , 010001 ( 2012 ) . c. -w . chiang , m. gronau , j. l. rosner and d. a. suprun , phys . 034020 ( 2004 ) [ hep - ph/0404073 ] . h. -w . ke , j. tang , x. -q . hao and x. -q . li , phys . d * 76 * , 074035 ( 2007 ) [ arxiv:0706.2074 [ hep - ph ] ] .	observing the ckm matrix elements written in different parametrization schemes , one can notice obvious relations among the sine - values of the cp phases in those schemes . using the relations , we establish a few parametrization - independent equations , by which the matrix elements of the ckm matrix can be completely fixed up to a universal parameter . if it is true , we expect that there should exist a hidden symmetry in the nature which determines the relations . moreover , it requires a universal parameter , naturally it would be the famous jarlskog invariant which is also parametrization independent . thus the four parameters ( three mixing angles and one cp phase ) of the ckm matrix are not free , but determined by the symmetry and the universal parameter . as we generalize the rules to the pmns matrix for neutrino mixing , the cp phase of the lepton sector is predicted to be within a range of @xmath0 centered at @xmath1 ( in the p@xmath2 parametrization ) which will be tested in the future experiments .
in this section , we justify remark 3 above and outline the derivation of a result analogous to theorem [ kesten - ipc-1 ] for the random walk on h. kesten s _ incipient infinite cluster _ ( iic ) . for cylinder events @xmath74 , the iic measure is defined by @xmath716 it was shown in @xcite that the limit ( [ eq : iic - def ] ) exists and that the resulting set function extends to a measure . note that the connected cluster of the origin , @xmath717 , is @xmath718-almost surely unbounded . we will refer to this cluster as the iic . we have the following result : let @xmath719 denote a simple random walk on the incipient infinite cluster started at @xmath7 . let @xmath0 denote the first exit time of @xmath720 from @xmath35 . there exists @xmath36 such that , for @xmath718-almost every @xmath38 and almost - every realization of @xmath721 , there is a ( random ) @xmath15 such that @xmath39 for @xmath1 greater than @xmath15 . we can proceed along the lines of the proof of estimate ( [ eq : w1bound ] ) , and consider a suitable modification of the random walk whose distribution coincides with that of @xmath162 from the first hitting time @xmath722 of @xmath723 to the first hitting time of @xmath517 after time @xmath722 , @xmath724 . to use the argument leading to ( [ eq : w1bound ] ) in our case , we merely need to show that we can prove an estimate equivalent to the one obtained for @xmath725 in section [ section - comparison ] . we will show that there are constants @xmath726 and @xmath34 such that @xmath727 by the argument given in the proof of lemma [ pisztoraslemma ] , there exists @xmath726 and @xmath34 such that @xmath728 let us denote the event on the left by @xmath729 . clearly @xmath730 @xmath729 depends only on the status of edges inside @xmath731 . write the conditional probability in the definition of @xmath718 as a ratio : @xmath732 for @xmath733 , we have , by independence and monotonicity @xmath734 now @xmath735 and by quasi - multiplicativity @xmath736 using this in ( [ eq : iic3prod ] ) , we have , by ( [ eq : pcbound ] ) : @xmath737 from which ( [ eq : piicbound ] ) follows at once . * acknowledgements . * we thank t. kumagai for suggesting the problem of proving a quenched analogue of kesten s subdiffusivity theorem and for comments on a previous verion . we are very grateful to a. fribergh for comments that led to a substantial reorganization of the presentation . j. h. and p. s. thank m. aizenman for advising and thank the organizers of the workshop `` current topics in mathematical physics '' at the erwin schrdinger institute , where some of this work was done . pisztora , a. , _ scaling inequalities for shortest paths in regular and invasion percolation _ , carnegie - mellon cna preprint , available at http://www.math.cmu.edu/cna/publications/publications2000/001abs/00-cna-001.pdf	we derive quenched subdiffusive lower bounds for the exit time @xmath0 from a box of size @xmath1 for the simple random walk on the planar invasion percolation cluster . the first part of the paper is devoted to proving an almost sure analog of h. kesten s subdiffusivity theorem for the random walk on the incipient infinite cluster and the invasion percolation cluster using ideas of m. aizenman , a. burchard and a. pisztora . the proof combines lower bounds on the instrinsic distance in these graphs and general inequalities for reversible markov chains . in the second part of the paper , we present a sharpening of kesten s original argument , leading to an explicit almost sure lower bound for @xmath0 in terms of percolation arm exponents . the methods give @xmath2 , where @xmath3 depends on the instrinsic distance and @xmath4 can be taken to be @xmath5 on the hexagonal lattice . [ [ section ] ] h. kesten has proved @xcite that the simple random walk @xmath6 started at @xmath7 on the incipient infinite cluster ( iic ) @xcite in two - dimensional bernoulli bond percolation is subdiffusive in the sense that there exists @xmath8 such that the family @xmath9 is tight . the purpose of the current work is to explain how a `` quenched '' version of this result can be obtained and extended to the random walk in an environment generated by a related two - dimensional model , invasion percolation . ( the model is defined in the next section ) . we present some refinements of kesten s method , which provides a general framework for proving subdiffusivity of random walks in stochastic geometric models . in the case of two - dimensional invasion percolation ( as well as the incipient infinite cluster ) , the ideas in @xcite can be used to give explicit bounds on @xmath10 from in terms of known critical exponents ( see and below ) . our main result is the following : [ main - theorem ] let @xmath11 be a simple random walk on the invasion percolation cluster ( ipc ) , and @xmath0 the first time @xmath12 exits the box @xmath13 ^ 2 $ ] : @xmath14 there exists @xmath3 such that , for almost every realization of the random walk , and almost every realization of the ipc , there is a ( random ) @xmath15 such that @xmath16 @xmath4 is a constant that can be estimated in terms of the behaviour of the one - arm and two - arm probabilities in critical percolation ( with measure @xmath17 ) : @xmath18 where @xmath19 are exponents such that @xmath20 and @xmath21 for some constants @xmath22 and @xmath23 . if one repeats the arguments of this paper in the setting of random walk on the iic or ipc of the hexagonal lattice , one can use the exact values of the one - arm and two - arm exponents to give a stronger bound on @xmath4 . indeed , it is not necessary in that case to use the van den berg - kesten inequality @xcite in , therefore giving @xmath24 using the conjectured value @xmath25 @xcite , we get a lower bound @xmath26 . without using this value , but using @xmath27 @xcite , we get @xmath28 . this result is stronger than the corresponding theorem for the random walk on the iic stated in ( * ? ? ? * theorem 1.27 ) , but it is derived by a modification of the strategy used there . in particular , kesten proves that @xmath29 for the `` averaged '' measure @xmath30 , which incorporates averaging with respect to the iic measure constructed in @xcite . closer examination of his proof reveals that one can take @xmath31 , and that the estimates in @xcite are sufficient to establish a `` quenched '' result by a simple application of the borel - cantelli lemma . a substantial part of the present paper is concerned with presenting arguments to overcome the difficulties in adapting kesten s proof to the invasion percolation cluster . the second result of the paper concerns a simple derivation of subdiffusivity of random walk on the ipc using results in @xcite and @xcite concerning the length of the shortest path from the origin to @xmath32 ( the _ chemical distance _ ) in near - critical percolation . the work of these authors implies that for large @xmath1 this length is of order at least @xmath33 , where @xmath34 . although theorem [ kesten - ipc-1 ] is contained in theorem [ main - theorem ] , it is of interest because its proof represents a significant reduction in complexity from the original argument of kesten . [ kesten - ipc-1 ] let @xmath0 be the time for a random walker on the invasion percolation cluster to exit @xmath35 . there exists @xmath36 such that , for @xmath37-almost every @xmath38 and almost - every realization of the random walk , @xmath39 for @xmath1 greater than some random @xmath15 . a similar , but simpler , argument applies to the incipient infinite cluster and gives an alternative proof that the random walk on the iic is almost surely subdiffusive . see the appendix for details . @xmath8 in the statement of theorem [ kesten - ipc-1 ] depends on the value of @xmath40 obtained by the methods of aizenman - burchard and pisztora . @xmath40 is both very small and difficult to calculate explicitly . kesten s comparison argument ( explained in section [ section - comparison ] ) yields an improvement of the estimate for @xmath0 in the previous theorem by a factor of the form @xmath41 , which leads to theorem [ main - theorem ] . we note that any explicit bound on @xmath40 would be directly reflected in that theorem . indeed , if one has upper and lower bounds ( with high enough probability ) @xmath42 then one can get the lower bound @xmath43 for any @xmath44 satisfying @xmath45 on the hexagonal lattice , this can be improved as above to @xmath46 one can actually show @xmath47 can be taken to be @xmath48 , which yields the improved bound ( assuming again the exact value of @xmath49 ) @xmath50 the improvement due to @xmath51 and @xmath47 in the previous remark comes from choosing @xmath52 larger in . it is actually a common misconception that kesten s original `` lost in the bushes '' argument gives a lower bound for @xmath0 proportional to the ratio of volume of the iic to the volume of its backbone . the reason this is false is that it is not clear how to increase @xmath52 to order @xmath1 . the parameter @xmath52 gives the scale at which volume estimates can be applied . there has been little success with rigorous results for random walks on low - dimensional critical models ( for instance , the iic and ipc ) . one notable example is the work of d. shiraishi @xcite on random walk on non - intersecting two - sided random walk trace . for results in high dimensions , we mention the recent work of g. kozma and a. nachmias @xcite on the iic in dimensions @xmath53 and of m. barlow , a. jrai , t. kumagai and g. slade on the iic for oriented percolation @xcite . on a critical galton - watson tree , kesten @xcite found the asymptotics of @xmath0 and constructed a scaling limit for random walk on the iic ( see also @xcite and @xcite ) . later , o. angel , j. goodman , f. den hollander and g. slade @xcite found similar results for random walk on the ipc on a regular tree . after setting some notation below , we give the definition of the invasion percolation model in section [ section - invasion ] , and recall some useful properties of the ipc derived in previous literature . we then prove theorem [ kesten - ipc-1 ] in section [ section - backbone ] , and explain how kesten s volume comparison argument is used to obtain theorem [ main - theorem ] in section [ section - comparison ] . section [ section - volume ] contains the derivation of estimates used in the proof of theorem 1 . for convenience , we work on the square lattice @xmath54 , but our results extend to planar lattices for which the russo - seymour - welsh estimates hold true . in this section , we give notation used throughout the paper for future reference . for any vertex ( lattice point ) @xmath55 , @xmath56 is the box @xmath57\times[v_2-n , v_2+n])\cap \mathbb{z}^2\\ & = \{x\in \mathbb{z}^2 : \|x - v\|_\infty \le n\}.\\ the box @xmath58 , centred at the origin . @xmath59 refers to the internal vertex boundary of @xmath56 : @xmath60 we also define @xmath61 where ipc is defined in the next section . for a graph @xmath62 , the set of edges is denoted by @xmath63 . for each @xmath64 $ ] , the independent bond percolation measure @xmath65 is an infinite product of bernoulli measures with parameter @xmath66 indexed by the edges of @xmath54 . for a finite set @xmath67 , and a vector @xmath68 we have @xmath69 a _ configuration _ @xmath70 is an element of @xmath71 . an edge @xmath72 is said to be _ open _ in the configuration @xmath70 if @xmath73 , and _ closed _ otherwise . if @xmath74 and @xmath75 are subsets of @xmath54 we denote by @xmath76 the probability of the event that @xmath74 and @xmath75 are connected by a path of open edges . the notation @xmath77 is defined analogously . we write @xmath78 to denote the probability that @xmath74 and @xmath75 are connected by @xmath66-open edges . we will use the connection probabilities @xmath79 and @xmath80 defined as @xmath81 these probabilities refer to independent bond percolation with parameter @xmath66 . when no parameter is specified , it is understood that @xmath82 ; that is , @xmath83 we denote by @xmath37 the invasion percolation measure on bond configurations in @xmath54 . throughout , @xmath38 will denote a realization of the ipc ; that is , a subgraph of @xmath54 sampled from @xmath37 . for each such @xmath38 , we denote by @xmath84 the probability measure associated with the simple random walk on the invasion cluster in the realization @xmath38 ( which by definition contains the origin @xmath85 ) . for @xmath86 , we denote by @xmath87 the logarithm of @xmath88 in base 2 . throughout the paper , @xmath89 will denote constants chosen independent of @xmath1 . we use the notation @xmath90 if there exists a constant @xmath91 such that @xmath92 this notation is only used if the implicit constant @xmath91 is deterministic ; that is , it does not depend on the realization of the ipc or of the random walk . the notation @xmath93 is used to emphasize that the implicit constant depends on the parameter @xmath94 . the notation @xmath95 denotes the existence of two positive constants @xmath96 and @xmath97 such that @xmath98 if @xmath99 and @xmath100 are two positive sequences , we use the notation @xmath101 to mean @xmath102 [ [ section - invasion ] ] the planar invasion percolation cluster is a random subgraph of the lattice @xmath54 which can be constructed from the familiar coupling of the independent bond percolation measures @xmath65 , @xmath103 . to every edge @xmath72 of @xmath54 , viewed as a graph , associate a random variable @xmath104 , uniformly distributed in @xmath105 $ ] , @xmath106 and @xmath107 being independent for @xmath108 . an edge is called @xmath66-_open _ if @xmath109 and is @xmath66-_closed _ otherwise . the distribution of the set of @xmath66-open edges is that of a bernoulli bond - percolation process at density @xmath66 . the distribution of @xmath110 is a product measure which will be denoted by @xmath30 . the ipc consists of a union of subgraphs of @xmath54 constructed by an iterative process : we start at the origin @xmath111 . at every stage , we form @xmath112 by adding to the current ( finite ) graph @xmath113 the edge @xmath72 with the least weight @xmath104 among @xmath114 as well as the endpoints of @xmath72 . the ipc is defined to be the union @xmath115 . since the percolation probability @xmath116 at @xmath117 is zero @xcite , the ipc contains infinitely many edges @xmath72 with @xmath118 . on the other hand , for any @xmath119 , by the russo - seymour - welsh theorem , the ipc will intersect the ( unique ) @xmath66-open infinite cluster almost surely ( see @xcite for general @xmath120 ) . by construction , once an edge @xmath72 in the @xmath66-open infinite cluster has been added , all edges added to the ipc after @xmath72 have weight no bigger than @xmath66 . we will later require bounds for the probability that the ipc intersects the @xmath66-open infinite cluster , for some fixed @xmath66 , by the time it reaches an annulus of size @xmath1 . such estimates can be found in @xcite , @xcite . an important notion in this context is the _ finite - size scaling length _ @xmath121 . to define it , consider for @xmath119 the probability @xmath122\times[0,m]).\ ] ] then @xmath121 is defined to be @xmath123 from @xcite , it is known that @xmath124 for @xmath125 , so we shall fix @xmath126 and henceforth simply refer to @xmath127 . we note the following properties of @xmath128 : 1 . @xmath129 is right - continuous , non - increasing in @xmath130 and @xmath131 as @xmath132 . 2 . taking @xmath133 small enough , there exists @xmath134 such that ( * ? ? ? * ( 2.8 ) ) : @xmath135 3 . again from ( * ? ? ? * eq . 2.10 ) , there exists @xmath136 independent of @xmath66 such that @xmath137 let @xmath138 be the @xmath139-th iterate of @xmath140 , and @xmath141 define , for @xmath142 and @xmath143 @xmath144 @xmath145 is a constant to be determined later . note that if @xmath146 , then @xmath147 when @xmath148 is sufficiently large . by ( 3 ) above , there exists @xmath149 such that @xmath150 item ( 2 ) in the list above implies ( * ? ? ? * ( 2.21 ) ) @xmath151 the measure @xmath30 refers to the coupling of the @xmath66-bernoulli measures described earlier . if the event @xmath152 occurs , the ipc intersects the @xmath153-open infinite cluster by the time it reaches @xmath154 . the bound ( [ eq : jarai - bound ] ) plays a role in estimates derived in section [ section - volume ] . [ [ section - backbone ] ] we begin by giving a brief sketch of the main idea . the first step is to consider a restriction of the random walk to a certain subset of the ipc , the backbone . the exit time for this walk from a box of size @xmath1 is controlled using the varopolous - carne inequality . this inequality implies that the exit time is at least of order @xmath155 , where @xmath120 is the chemical ( instrinsic ) distance to the boundary of the box of size @xmath1 through the ipc . in lemma [ pisztoraslemma ] , we outline an argument of a. pisztora that proves that @xmath120 grows superlinearly with @xmath1 . all of these estimates are tight enough to apply borel - cantelli and close the proof of subdiffusivity . the simple random walk started at @xmath7 on the ipc is the markov chain @xmath11 with the set of sites in the ipc as its state space , such that @xmath156 , and with transition probabilities given by @xmath157}{\operatorname{deg}(x,\mathrm{ipc})}.\ ] ] the random variable @xmath158 denotes the number of sites @xmath159 such that the edge @xmath160 belongs to the @xmath161 . below , it will be convenient to work with a modification of @xmath162 that is reversible on @xmath163 . thus , we let @xmath164 be the markov chain started at the origin and defined by the transition probabilities @xmath165}{\operatorname{deg}(x,\lambda(n))}.\ ] ] note that the distribution of @xmath166 coincides with that of @xmath12 for @xmath167 , where @xmath168 moreover , the distribution of @xmath169 is equal to the distribution of the exit time @xmath170 defined in terms of the `` full '' random walk @xmath162 on @xmath163 . thus , it will suffice to obtain bounds on @xmath169 . the `` backbone '' @xmath171 of @xmath163 is the set of sites in @xmath163 connected in the invasion cluster to @xmath7 and to @xmath32 by two disjoint paths . a simple argument ( see ( * ? ? ? * lemma 3.13 ) ) shows that whenever @xmath172 leaves the backbone , it must return at the site where it left before it reaches @xmath173 . thus the random walk @xmath172 on @xmath163 induces a random walk @xmath174 on @xmath171 which moves only when @xmath172 is in @xmath171 . that is , if we define @xmath175 then @xmath174 is a random walk on the backbone @xmath171 , with transition probabilities given by @xmath176}{\operatorname{deg}(x,\lambda(n ) ) } , & y\neq x\\ \frac{\operatorname{deg}(x,\lambda(n ) ) -\operatorname{deg}(x , b(n))}{\operatorname{deg}(x,\lambda(n ) ) } , & x = y . \end{cases}\ ] ] here , @xmath177 is defined as the number of edges @xmath160 in @xmath163 such that @xmath178 . irrespective of the geometry of @xmath171 , @xmath174 must travel at least @xmath1 steps in @xmath171 to reach @xmath173 , because the distance between any two points in @xmath179 is no less than the corresponding chemical distance in @xmath54 . this fact was used by kesten to conclude that the time spent by the walker on the backbone is of order at least @xmath180 with high probability . the carne - varopoulos bound ( @xcite , @xcite ; see also ( * ? ? ? * theorem 13.4 ) ) allows us to obtain a better estimate by considering the chemical distance on @xmath171 . it implies that the reversible markov chain @xmath174 has at most diffusive speed in the intrinsic metric of the backbone . if @xmath181 is the stationary measure for the walk @xmath174 ( @xmath181 depends on @xmath38 ) , then @xmath182 the right side of this expression refers to the chemical distance in the backbone @xmath171 . the ratio appearing on the right can be bounded independently of the realization @xmath38 of the invasion percolation , since the stationary measure @xmath181 satisfies @xmath183 for any @xmath184 . since @xmath179 , we have the inequality of graph distances : @xmath185 summing this bound over @xmath186 , we find @xmath187 suppose we restrict our attention to realizations @xmath38 of the environment such that the chemical distance in @xmath171 satisfies @xmath188 for some @xmath189 and some deterministic constants @xmath34 , @xmath190 . for such @xmath38 , @xmath191 and @xmath1 sufficiently large , we have : @xmath192 it follows from work of aizenman and burchard @xcite that the chemical distance inside a large box in independent bond percolation with parameter @xmath117 is bounded below by a power @xmath34 of the euclidean distance in @xmath54 with high probability . pisztora @xcite showed how to extend this result to @xmath119 suitably close to @xmath193 , and to the invasion percolation cluster . we reproduce the argument leading to his result , in a form that suits our needs , in the lemma below . theorem [ kesten - ipc-1 ] follows from these results and the considerations above . [ pisztoraslemma ] there exist @xmath194 and @xmath34 such that @xmath195 the models considered in @xcite are defined by families @xmath196 of probability measures on collections of curves in a compact region @xmath197 . for each @xmath198 , @xmath199 is supported on unions of polygonal curves with step size @xmath198 . the realizations in the support of @xmath199 are denoted by @xmath200 . a truncated version of capacity is used to obtain lower bounds on the minimal number @xmath201 of sets of diameter @xmath198 required to cover a given set @xmath202 : @xmath203 where @xmath204 the infimum is over borel probability measures supported on @xmath74 . under the assumption `` hypothesis h2 , '' the authors of @xcite obtain uniform bounds for @xmath205 : if there exist some @xmath206 , and @xmath207 such that for every @xmath208 and collection of @xmath208 rectangles @xmath209 of lengths @xmath210 and cross - section @xmath70 , and satisfying @xmath211 for all @xmath139 , we have @xmath212 then the capacity @xmath205 of macroscopic curves is bounded below for some @xmath34 ( * ? ? ? * theorem 1.3 ) : all curves @xmath213 in @xmath200 with @xmath214 satisfy @xmath215 @xmath216 is a random variable which is _ stochastically bounded below _ in the sense that @xmath217 uniformly in @xmath198 as @xmath218 . we will apply the results in @xcite , with @xmath219 to bond percolation on the rescaled lattice @xmath220 ^ 2.\ ] ] for @xmath64 $ ] , let @xmath221 denote the independent bond percolation measure with parameter @xmath66 on the edges of @xmath222 . @xmath221 induces a probability measure on configurations @xmath223 of curves in @xmath224 ^ 2 $ ] : the percolation configuration is a union of connected paths of @xmath66-open edges , each edge being identified with a line segment of length @xmath225 . in the case of independent percolation , hypothesis h2 reduces to the existence of a cross - section @xmath70 and @xmath226 such that the probability that there exists an open - crossing of a rectangle of cross - section @xmath70 is less than @xmath80 . by the russo - seymour - welsh estimates , hypothesis h2 is satisfied for @xmath227 . the lower bound ( [ eq : lowerbd ] ) gives an estimate for the chemical distance in @xmath223 between any two sets in @xmath228 ^ 2 $ ] . any @xmath229-open path in @xmath222 connecting subsets @xmath74 and @xmath75 of @xmath222 at euclidean distance @xmath230 contains at least @xmath231 bonds . denote by @xmath232 the ( random ) number of bonds in the shortest @xmath229-open path connecting @xmath74 and @xmath75 in @xmath222 . by ( [ eq : straight - runs ] ) , given any @xmath8 , we can choose @xmath233 such that for all @xmath1 , @xmath234 the scaling @xmath235 defines a measure - preserving bijection between @xmath236 and @xmath237 . it follows that for each @xmath8 , there exists a constant @xmath233 such that for all subsets @xmath74 , @xmath75 of @xmath35 at euclidean distance @xmath238 from each other , @xmath239 note that if @xmath75 cuts @xmath74 from @xmath154 in @xmath54 , the restriction that the path be contained in @xmath35 is superfluous . this point will be relevant below . the observation in @xcite is that the aizenman - burchard bounds remain valid for @xmath119 as long as @xmath1 is smaller than the correlation length @xmath129 . the estimate used to obtain ( [ eq : straight - runs ] ) depends only on @xmath70 and @xmath80 @xcite . it follows from the definition of @xmath129 and the russo - seymour - welsh estimates that there exists @xmath226 such that for rectangles of cross - section ratio @xmath240 , say , with long side @xmath241 , @xmath242\times[0,n/3])\le \rho.\ ] ] thus ( [ eq : straight - runs ] ) remains true uniformly for @xmath243 . repeating the argument above , we see that we can choose @xmath233 independent of @xmath244 to make the probability @xmath245 smaller than an arbitrary @xmath8 . the distance refers to the chemical distance in the union of all percolation clusters in the box @xmath246 . since @xmath131 as @xmath132 , for any fixed @xmath10 , @xmath129 is much greater than @xmath233 , and so the estimate on the distance is not vacuous . more precisely , we find @xmath247 a block argument with blocks of size @xmath248 converts the initial estimate ( [ eq : c - epsilon ] ) into an exponential bound for the macroscopic chemical distance in near - critical percolation ( see the proof of ( * ? ? ? * theorem 1.3 , pp . 12 - 14 ) ) . there exist constants @xmath249 , @xmath250 such that if @xmath66 is sufficiently close to @xmath229 : @xmath251 with this in hand , ( [ eq : invasion - chemical ] ) follows from the construction described in section [ section - invasion ] . we outline the argument . the occurrence of the event @xmath252 implies that all edges of the ipc in @xmath253 are @xmath254-open . @xmath255 the final inequality follows from ( [ eq : renormalized - ab ] ) and ( [ eq : jarai - bound ] ) . recalling ( [ eq : l_p ] ) : @xmath256 and choosing @xmath257 suitably large in the definition of @xmath258 , we find : @xmath259 by slightly lowering @xmath40 to absorb the logarithm , the probability on the left of ( [ eq : finaldistance ] ) can be made to match the form of the left side of ( [ eq : invasion - chemical ] ) . the final part of the proof of lemma [ pisztoraslemma ] shows that for any @xmath260 and any @xmath261 , one can find constants ( depending on @xmath262 and @xmath208 ) such that @xmath263 here @xmath34 is the constant appearing in ( [ eq : invasion - chemical ] ) . such a statement will be used in section [ section - comparison ] below . for @xmath34 , let @xmath264 be the event @xmath265 by ( [ eq : invasion - chemical ] ) in the previous lemma , we have @xmath266 for some @xmath34 . applying the borel - cantelli lemma and choosing @xmath267 we can use ( [ eq - strong - cv ] ) with @xmath268 and @xmath269 ; for some @xmath270 we have : @xmath271 a second application of the borel - cantelli lemma leads to theorem [ kesten - ipc-1 ] . note that for the argument above it was not necessary to consider @xmath174 . however , the decomposition of the ipc into a backbone and `` dangling ends '' will be central in the derivation of theorem [ main - theorem ] below . the proof of theorem [ kesten - ipc-1 ] shows that @xmath174 alone already contributes at least @xmath272 steps to @xmath0 . [ [ section - comparison ] ] our modification of kesten s argument compares the volume of sites in the invasion percolation cluster ( ipc ) to the volume of sites on the backbone to conclude that the walk must be subdiffusive . we assume for simplicity of notation that @xmath273 , @xmath274 . we introduce two stopping times : @xmath275 by definition , we clearly have : @xmath276 hence , it will suffice to obtain a lower bound on the right side of the previous expression . @xmath277 is a simple random walk on ipc ; now define @xmath278 to be the simple random walk on ( the possibly disconnected ) @xmath279 with initial point @xmath280 . letting @xmath281 be the hitting time of @xmath282 by the walk @xmath278 , we note that @xmath281 has the same distribution as @xmath283 . a key tool in kesten s argument is the following result from @xcite , expressing the spatial `` smoothness '' of the local times for a reversible markov chain . [ continuity ] let x , y be two sites in @xmath284 , and let @xmath285 be the local time at a site @xmath88 of the walk @xmath278 . then , for some @xmath286 and any @xmath287 : @xmath288 in @xcite , lemma [ continuity ] is stated in terms of the intrinsic distance on the incipient infinite cluster . replacing @xmath289 , @xmath290 and @xmath291 in the proof of lemma 3.18 in @xcite by @xmath292 , @xmath293 and @xmath294 , respectively , we obtain lemma [ continuity ] above . we also modify our definition of the backbone . @xmath295 is defined to be the set of sites in @xmath284 connected by two disjoint paths ( in @xmath284 ) to @xmath32 and @xmath296 . @xmath297 is the induced walk on @xmath298 , defined analogously to @xmath174 in section [ section - backbone ] . we let @xmath299 be the number of steps @xmath297 takes between @xmath300 and @xmath281 ; @xmath299 is the time spent by @xmath278 on @xmath295 . kesten s comparison argument will be applied to @xmath278 . the idea is to consider a `` thickening '' of the backbone of size @xmath52 . by lemma [ continuity ] , if a box @xmath301 of size @xmath52 contains a site @xmath302 with @xmath303 , the random walk visits all accessible sites of @xmath284 inside @xmath301 at least @xmath304 times , with high probability . if it is traversed by a portion of the random walk , the box @xmath301 typically contains @xmath305 sites of @xmath284 , and at most @xmath306 sites of @xmath295 . thus the time spent by @xmath278 in @xmath307 up to @xmath281 is larger than the time @xmath297 spends there by a factor of at least @xmath308 . by choosing @xmath52 appropriately , the set of sites @xmath159 on the backbone which do not satisfy the lower bound of order @xmath309 on @xmath310 will make a contribution bounded by a fraction of the total time spent on the backbone . to realize the strategy just described , we tile @xmath311 by squares of size @xmath312 for a constant @xmath313 to be determined . here @xmath49 is the exponent appearing in ( [ eq : twoarmprob ] ) . we note for future reference that @xmath314 this bound on @xmath49 can be proved using the method of ( * ? ? ? * cor . 3.15 ) . for the details , the reader can see a standard sketch of a similar inequality ( for crossings of an annulus ) under equation later in the paper . for @xmath315 , define @xmath316\times [ q ( j_2 - 1 ) , q(j_2 + 2)].\end{aligned}\ ] ] given a realization @xmath38 of ipc and a realization of the walk , we follow the path of @xmath317 until @xmath281 by introducing two sequences @xmath318 and @xmath319 first setting @xmath320 and @xmath321 to be the index such that @xmath322 and then defining @xmath323 by @xmath324 @xmath278 may reach @xmath282 before leaving @xmath325 , in which case @xmath323 ends the sequence . we let @xmath326 denote the component of @xmath327 containing @xmath328 . @xmath278 may return several times to the same square , so @xmath329 may be equal to @xmath326 for @xmath330 . enumerating the @xmath326 without repetition as @xmath331 , with @xmath332 the component of @xmath325 where @xmath333 is such that @xmath334 we define @xmath335 since any @xmath88 belongs to at most @xmath336 different @xmath337 squares , we have : @xmath338 we now state volume estimates analogous to those obtained in @xcite for the incipient infinite cluster ; they will be derived in the next section . we will only be concerned with those indices in the set @xmath339 the first estimate is for the number of backbone sites in any @xmath337 square ; for any @xmath340 , we have : @xmath341 the second provides , with high probability , a lower bound for the number of sites of the ipc in a box @xmath342 , @xmath343 , given that there is a crossing of @xmath344 : @xmath345 here @xmath94 is arbitrary but the implicit constant depends on the choice of @xmath94 . we now define the events @xmath346 , @xmath348 and @xmath349 , @xmath350 . the ratio @xmath351 will be bounded below by @xmath308 on the event @xmath352 . 1 . @xmath353 2 . @xmath354 3 . @xmath355 4 . @xmath356 5 . [ item - w1 ] @xmath357 6 . @xmath358 7 . @xmath359 by the remark following the proof of lemma [ pisztoraslemma ] , there exists @xmath360 such that @xmath361 with the same constant @xmath40 as in ( [ eq : invasion - chemical ] ) . for any @xmath362 and @xmath363 , we use the carne - varopoulos estimate ( [ eq - strong - cv ] ) , applied to the symmetric chain @xmath297 , to show as in the proof of theorem [ kesten - ipc-1 ] : @xmath364 giving the bound ( [ kesten - ipc-1 ] ) . recall the definition of @xmath52 in . we have @xmath365 noting that there are , up to a constant , at most @xmath366 indices @xmath367 in @xmath368 , and choosing @xmath369 in ( [ eq - backbone - bound ] ) , and accordingly in the definition of @xmath370 , we find @xmath371 by the estimate ( [ eq - big - backbone - bound ] ) in section [ section - volume ] : @xmath372 by ( [ eq - crossing - bound ] ) ( for some @xmath94 large enough ) , we have @xmath373 finally , we have @xmath374 uniformly in @xmath38 . indeed , suppose @xmath88 and @xmath159 are two sites as in the description of @xmath375 , then , for @xmath1 sufficiently large , @xmath376 for any @xmath38 . using @xmath377 for any @xmath378 in the ipc , we find that on @xmath379 , for some pair @xmath378 either @xmath380 and @xmath381 or @xmath382 and @xmath383 the first case is contained in the event appearing in ( [ eq : localtimes ] ) . so is the second case , after reversing the roles of @xmath88 and @xmath159 in that event . using @xmath384 applying lemma [ continuity ] , and taking the union over all pairs , @xmath385 , we find that , whatever @xmath38 in the support of @xmath37 : @xmath386 a similar argument applies to @xmath387 . applying the borel - cantelli lemma to @xmath388 we find that for @xmath37-almost every @xmath38 , there exists @xmath270 such that @xmath389 holds when @xmath390 . for any such @xmath38 , a further application of the borel - cantelli lemma shows that , @xmath84-almost surely , @xmath391 holds for @xmath1 large enough . it remains to show that whenever all the events above hold , we have the subdiffusive bound of theorem [ main - theorem ] . first , on @xmath392 , if we denote by @xmath393 the sum over indices @xmath394 such that @xmath395 contains a site @xmath396 with @xmath397 then , assuming @xmath398 also occurs , adjusting the constant @xmath313 in the definition of @xmath52 ( see ): @xmath399 it follows that @xmath400 on @xmath375 , letting @xmath401 be the lexicographically earliest point of @xmath295 in @xmath395 , we have for those indices @xmath402 occurring in @xmath393 : @xmath403 for each @xmath404 , @xmath405 contains an invaded crossing in @xmath395 . thus , on @xmath406 , we can write : @xmath407 bounding every @xmath408 term below individually in ( [ eq : sumcompare ] ) and using the bk inequality , we find : @xmath409 on @xmath398 , we have @xmath410 . recalling the definition of @xmath52 from , we have the following bound for @xmath411 . @xmath412 choosing @xmath3 such that @xmath413 , we obtain : @xmath414 the desired result . [ [ section - volume ] ] in this section we prove the volume estimates ( [ eq - backbone - bound ] ) and ( [ eq - crossing - bound ] ) . we show that the following moment bounds hold for @xmath415 : @xmath416 for @xmath417 and constants @xmath418 , @xmath419 . the estimate ( [ eq - small - moments ] ) implies the existence , for @xmath420 small enough , of the exponential moment : @xmath421 applying chebyshev s inequality with @xmath422 yields ( @xmath423 ) . from ( [ eq - big - moments ] ) , we obtain the finiteness , for sufficiently small @xmath424 , of the stretched exponential moment : @xmath425 using chebyshev s inequality with @xmath426 , we obtain , for each @xmath427 : @xmath428 to derive ( [ eq - small - moments ] ) and ( [ eq - big - moments ] ) , we follow the method introduced by jrai @xcite to estimate the moments of the volume @xmath429 of the ipc in a box . we will instead apply this argument to the volume of a backbone , and then combine it with an inductive argument of nguyen @xcite . we begin with the first moment ( @xmath430 ) in ( [ eq - small - moments ] ) . if @xmath431 , the number of sites of @xmath342 with two disjoint connections in the ipc to @xmath432 provides an upper bound for the volume of @xmath433 . let @xmath434 denote the set of sites in @xmath342 with two @xmath435-open connections to @xmath432 . note that @xmath436 on @xmath437 ( defined in ) , every edge of the ipc in @xmath438 is @xmath435-open , as noted at the end of section [ section - invasion ] , and thus : @xmath439 the first term is bounded up to a constant factor by : @xmath440 the terms of the sum are estimated using the harris - fkg inequality : @xmath441 by decomposing @xmath342 according to the distance @xmath442 to @xmath432 , we find : @xmath443 . \nonumber\end{aligned}\ ] ] by the same argument used for ( 7 ) in @xcite ( see remark ( 37 ) there ) , the sum up to @xmath444 in ( [ eq : smoothsum ] ) is bounded up to a constant by @xmath445 the proof in @xcite is carried out for @xmath446 , but the implicit constants that appear are due to applications of rsw theory and thus are uniformly bounded in @xmath119 . by comparability of the arm exponents below @xmath129 @xcite ( see also ( * ? ? ? * theorem 26 ) ) , we have @xmath447 thus , finally , in ( [ eq : nsquared ] ) , we have ( since @xmath448 ) @xmath449 where in the first step we have used the inequality @xmath450 for @xmath451 . a similar inequality for @xmath452 was used in @xcite , where the author indicates that it can be proved by the argument in ( * ? ? ? * corollary 3.15 ) . the proof of ( [ eq : twoarms ] ) follows the same general strategy , but does not use the van den berg - kesten inequality : let @xmath453 , @xmath454 and consider the annuli @xmath455 . the inner squares of these @xmath456 annuli are adjacent . the event that there exists a @xmath229-open left - right crossing of @xmath457 has probability bounded below uniformly in @xmath458 , and implies that one of the annuli is crossed by two disjoint @xmath229-open paths . by quasi - multiplicativity , this probability is comparable to @xmath459 , and ( [ eq : twoarms ] ) follows by a union bound . inserting ( [ eq : rholessthanlog ] ) into and then into ( [ eq : small - moment - sum ] ) , we find @xmath460 the final term corresponds to @xmath461 , which is @xmath462 by . using , we may choose @xmath257 large enough to make the first term @xmath463 an important point made in @xcite is that choosing @xmath257 possibly larger , we may bound the contribution from the sum in the parentheses by a constant . indeed , we have @xmath464 this establishes ( [ eq - small - moments ] ) for @xmath430 . to deal with the higher moments , we use the following general lemma : [ nguyenslemma ] let @xmath465 , @xmath466 , and @xmath467 be the set of sites of @xmath35 with two disjoint @xmath66-open connections to @xmath32 . there exists a constant @xmath468 independent of @xmath1 and @xmath66 such that , for any @xmath469 , the following inductive bound holds : @xmath470 the result is essentially due to nguyen @xcite , who proved that for @xmath471 and @xmath472 , @xmath473 where @xmath474 is the set of sites in @xmath475 connected to @xmath476 by a @xmath66-open path . @xmath477 is a constant uniform in @xmath478 and @xmath66 . when @xmath479 , the proof in @xcite is easily adapted to the variables @xmath480 . we define the event @xmath481 the idea is to write @xmath482 where we have set @xmath483 letting @xmath484 we have @xmath485 uniformly in @xmath442 and @xmath1 ( even for @xmath486 ) . by the fkg inequality : @xmath487 @xmath488 is the event that @xmath489 is connected to @xmath32 by two disjoint open paths outside of @xmath490 . by independence , the last quantity on the right is bounded , up to a constant , by @xmath491 for any @xmath442 , we have : @xmath492 returning to ( [ eq : nguyen - triple - sum ] ) , we find @xmath493 if @xmath479 , we have @xmath494 and the estimate ( see the remark concerning ( [ eq : smoothsum ] ) above ) @xmath495 leads to the inductive estimate claimed above . if @xmath496 , we split the sum as we did in the treatment of the first moment of @xmath497 : @xmath498 here we have used that for @xmath499 , @xmath500 . this follows by a variant of the argument presented in ( * ? ? ? * section 7.4 ) . this establishes the lemma . using lemma [ nguyenslemma ] , induction and the fact that @xmath501 for large @xmath148 , we obtain @xmath502 thus , arguing as for ( [ eq : small - moment - sum ] ) : @xmath503 using the value of @xmath52 from and choosing @xmath504 , we use ( [ eq : logsconverge ] ) to get ( [ eq - small - moments ] ) in the case where @xmath367 is such that @xmath505 . for a general @xmath415 , the intersection @xmath506 is a union of at most two rectangles with side lengths @xmath507 and @xmath508 , @xmath509 . repeating the arguments above , we see that the size of the intersection of each of these rectangles with the backbone @xmath295 enjoys the moment bounds ( [ eq - small - moments ] ) , with @xmath510 replaced by @xmath511 , with @xmath512 . using ( [ eq : twoarms ] ) , we obtain the upper bound @xmath513 . for the higher moments , moment bounds of the form ( [ eq - small - moments ] ) with a larger ( but still uniform in @xmath514 ) constant @xmath418 are valid . ( [ eq - backbone - bound ] ) now follows by a union bound . the proof of ( [ eq - big - moments ] ) follows a very similar pattern to the above . instead of @xmath434 , we consider the sets @xmath515 of points of @xmath516 with two disjoint @xmath435-open connections to @xmath517 . repeating the steps for the case @xmath430 gives for the first moment . in the case of higher moments , we need to modify inequality , replacing it with @xmath518 for some @xmath519 . decomposing as before over the events @xmath520 and choosing @xmath521 leads to . in its general outline , the proof is similar to that of ( 3.24 ) in @xcite , with some parameters chosen differently because we wish to bound only logarithmic deviations from the mean . however , the estimates in @xcite are carried out for critical percolation ( @xmath522 ) , and the proof of the initial estimate ( [ eq : pre - peierls - bound ] ) below in the supercritical case introduces an additional technical difficulty . as explained in section [ section - invasion ] , the entire ( finite ) @xmath229-open cluster of any site in the ipc also belongs to the ipc . thus , for any crossing @xmath458 ( in the ipc ) of @xmath344 , the number of sites in @xmath344 connected to @xmath458 by @xmath229-open paths provides a lower bound for the quantity in ( [ eq - crossing - bound ] ) . the starting point is the following : let @xmath458 be a deterministic path crossing @xmath523 . let @xmath524 be the set of sites @xmath229-connected to @xmath458 inside this annulus . we have the lower bound : @xmath525 for some constants @xmath526 independent of @xmath1 . the proof is essentially that of ( 56 ) in @xcite . kesten s idea is to compute the first and second moments of the number @xmath527 of sites in @xmath528 connected to open circuits in @xmath529 and @xmath530 ( and thus to @xmath458 ) and use the harris - fkg inequality and the second moment method . fix some @xmath531 ( to be chosen later ) . we first show that , for any @xmath532 ( entailing in particular @xmath533 by ) , and any coordinates @xmath534 such that @xmath535\times [ -t+v_2,4t+v_2]\subset s(3m)\setminus ( s(m))^\circ,\ ] ] we have , for some constants @xmath536 , @xmath537 , @xmath538\times [ v_2,3t+v_2 ] \\\text{such that } \sharp \{x \in t(\mathbf{v } ) :x \xrightarrow{p_c } r \text { in } t(\mathbf{v})\ } \le { \mathbf{c}}_{24 } t^2\pi(t)/(\log t)^\delta \end{gathered}\right)\\ \lesssim \frac{1}{(\log t)^{{\mathbf{c}}_{25}}}.\end{gathered}\ ] ] for any crossing @xmath458 of @xmath539 , let @xmath540 the probability on the left of ( [ eq : pre - peierls - bound ] ) equals @xmath541 the precise meaning of `` @xmath458 intersects no @xmath229-open crossing of @xmath539 '' is that no site in @xmath539 is a common endpoint of an edge in @xmath458 and an edge in some horizontal @xmath229-open crossing of @xmath539 . in particular , @xmath458 is edge - disjoint from all @xmath229-open crossings . both terms on the right in ( [ eq : crossings - union - bound ] ) will be bounded , up to a constant factor , by @xmath542 . we begin by estimating the first term in ( [ eq : crossings - union - bound ] ) . for any crossing lattice path @xmath543 of @xmath539 , let @xmath544 be the set of edges with an endpoint that can be connected to @xmath545 \times \{v_2\}$ ] by a path in @xmath539 that does not touch @xmath543 ( below @xmath543 ) . note that @xmath544 may include edges not entirely contained in @xmath539 . the lowest @xmath229-open crossing @xmath546 of @xmath547 is defined as the horizontal crossing of the rectangle by @xmath229-open edges such that the component @xmath548 , is minimal . @xmath549 is defined inductively as the lowest crossing of @xmath550 ( defined analogously see ( * ? ? ? * prop . 2.3 ) for the existence of @xmath549 and precise definitions ) . for a given ( lattice path ) crossing @xmath543 of @xmath539 , write @xmath551 for the sigma algebra generated by the status of edges in @xmath552 . we define @xmath553 to be the maximal @xmath208 such that @xmath549 exists . the veracity of the following string of inequalities is then evident : @xmath554 on @xmath555 , we have the following uniform estimate for the conditional probability given @xmath556 : @xmath557 to see this , consider the left endpoint @xmath558 of the crossing @xmath559 , and annuli @xmath560 for @xmath561 , the existence of circuits @xmath562 around @xmath558 in @xmath563 and @xmath564 in @xmath565 , all of whose edges outside @xmath566 are @xmath229-open implies that any site in @xmath567 connected to @xmath568 is @xmath229-connected to the crossing @xmath559 . thus , using the harris - fkg inequality , independence of the edge configurations in @xmath566 and @xmath569 and the second moment method as in the discussion preceding ( [ eq : pre - peierls - bound ] ) , there exist constants @xmath536 , @xmath570 , such that for each @xmath139 with @xmath571 : @xmath572 there are @xmath573 admissible indices @xmath139 , and so by independence of the configuration in the different annuli , we find @xmath574 which is the same as ( [ eq : oneoverlogdelta ] ) . returning to the double sum of : @xmath575 by the russo - seymour - welsh method , the @xmath17 probability of a dual vertical crossing of @xmath539 is bounded below by some @xmath576 . thus , by disjointness of the @xmath549 s and the bk inequality , @xmath577 this allows us to bound the sum in ( [ eq : k - sum ] ) by @xmath578 . we now estimate the second term on the right in ( [ eq : crossings - union - bound ] ) . denote by @xmath579 the event that there exists a @xmath258-open crossing @xmath458 of @xmath539 such that @xmath458 intersects no @xmath229-open crossing of @xmath539 . for any @xmath580 , we have @xmath581 as previously , @xmath553 denotes the maximal number of disjoint @xmath229-open crossings of @xmath539 . we will choose @xmath582 , so as to give the following bound for the second term above : @xmath583 where the constant @xmath584 is a constant such that @xmath585 . for the first term , we have the union bound @xmath586 it will be shown below ( see lemma [ crossing - lemma ] ) that there is a constant @xmath587 such that , for any @xmath588 with @xmath589 , @xmath590 where @xmath591 is the `` alternating 4-arm probability , '' associated to the event that @xmath592 is connected to @xmath593 by two disjoint @xmath229-open paths and its dual edge is connected to @xmath593 by two disjoint @xmath229-closed dual paths . thus @xmath594 for a constant @xmath595 . the factor @xmath596 is @xmath597 ) . indeed , it was shown in @xcite that , uniformly for @xmath119 sufficiently close to @xmath229 : @xmath598 applying this to @xmath599 , and using ( [ eq : l_p ] ) and @xmath600 ( * ? ? ? * proposition 12 ) , we find , for @xmath601 large enough : @xmath602 thus we have @xmath603 here we have used @xmath604 . using ( [ eq : nearcritical ] ) again , we have : @xmath605 by quasimultiplicativity ( * ? ? ? * proposition 12 ) : @xmath606 where @xmath607 is the probability that there are four arms of alternating occupation status connecting @xmath32 to @xmath608 in @xmath609 . using reimer s inequality @xcite and the ( exact ) scaling of the 5-arm exponent ( see ( * ? ? ? * theorem 23 ) or @xcite ) , we have : @xmath610 here , @xmath611 is the one - arm probability , that @xmath32 is connected to @xmath608 by an open path . since the one - arm probability satisfies the power - type upper bound @xmath612 for some @xmath613 ( apply the bk inequality to the bound on @xmath49 in ) , we find that @xmath596 is bounded , up to a constant , by @xmath614 since we assume @xmath615 , and @xmath616 ( see ( [ eq : twoarmbound ] ) ) , we find @xmath617 for some @xmath427 . returning to ( [ eq : bound - for - pgek ] ) , we have the bound : @xmath618 it remains to prove ( [ eq : fourarms ] ) . this is done in lemma [ summation - lemma ] below . before proceeding , let us introduce a definition : a _ @xmath229-closed arm with @xmath208 defects _ is a path of dual edges , all of which except for @xmath208 of them are @xmath229-closed . the proof of lemma [ summation - lemma ] depends on the following : [ crossing - lemma ] let @xmath579 be the event that there exists a @xmath258-open crossing @xmath458 of @xmath539 such that @xmath458 intersects no @xmath229-open crossing of @xmath539 , and @xmath553 be the maximal number of horizontal @xmath229-open crossings of @xmath539 . suppose @xmath619 ; then there exists an edge @xmath620 such that 1 . @xmath72 has two disjoint @xmath258-open arms to @xmath621 $ ] ( the left side of @xmath539 ) and @xmath622 $ ] ( the right side of @xmath539 ) , respectively . 2 . @xmath623 , the dual edge to @xmath72 , has two disjoint @xmath229-closed arms , each with at most @xmath208 defects to @xmath545\times \{v_2\}$ ] ( the bottom side of @xmath539 ) and to @xmath545\times \{v_2 + 3t\}$ ] ( the top side of @xmath539 ) , respectively . 3 . @xmath624 $ ] . on the event @xmath625 , menger s theorem ( * ? ? ? * section 3.3 ) implies that there is a dual path @xmath40 joining the top of @xmath539 to the bottom , all of whose edges , with exactly @xmath208 exceptions , are closed and which moreover does not intersect itself . this path must intersect the horizontal @xmath258-open crossing @xmath458 ( * ? ? ? * prop . 2.2 ) along a @xmath258-open edge @xmath72 . this edge must then be @xmath229-closed . the dual edge @xmath623 , being part of the non - self - intersecting @xmath40 with @xmath208 defects , has two dual arms joining it to the top and bottom of @xmath539 . ( see figure [ fig : menger ] . ) moreover , the total number of defects on these two arms is @xmath208 . this establishes the lemma . . the dotted path has @xmath208 defects , shown as empty circles . the solid black path represents a @xmath258-open crossing and the grey paths represent disjoint @xmath229-open crossings . ] the proof of ( [ eq : pre - peierls - bound ] ) is concluded by the following lemma , which establishes the estimate ( [ eq : fourarms ] ) : [ summation - lemma ] there is a constant @xmath587 such that , for each @xmath469 , the following bound holds : @xmath626 it suffices to estimate the probability that there is an edge in @xmath539 satisfying the two conditions in lemma [ crossing - lemma ] . to that end , we will show that the expected number of such edges in @xmath539 is bounded by the quantity on the right side of equation ( [ eq : xibound ] ) . for @xmath620 , let @xmath627 be the event that @xmath72 satisfies the conditions of lemma [ crossing - lemma ] . the key step is the existence of a constant @xmath595 such that @xmath628 where @xmath629 is the event that @xmath72 has two disjoint @xmath229-open arms joining it to the left and right sides of @xmath539 respectively , and @xmath623 has two disjoint @xmath229-closed dual arms to the top and bottom sides of @xmath539 . the effect of the arms with defects is to produce the logarithmic factor indicated in the equations above : @xmath630 where @xmath631 is defined analogously to @xmath629 above , except that the open connections are required to be @xmath258-open rather than @xmath66-open . this follows from the argument in ( * ? ? ? * prop . 17 ) , where it is shown that if @xmath632 denotes the probability that the origin is connected to @xmath32 by @xmath139 paths , with @xmath120 defects in total , whose occupation status is specified by the sequence @xmath633 , then @xmath634 @xmath635 is the event that there are @xmath139 arms ( without defects ) to @xmath32 ( with occupation status as in @xmath70 ) . inspection of the proof in @xcite reveals that the constant implicit in ( [ eq : defects ] ) is of the form @xmath636 . separating the four arms as in @xcite verifies that ( [ eq : defectbound ] ) holds . it now remains to show that @xmath637 ; that is , that for @xmath1 sufficiently large , we can change the @xmath258-open arms in the definition of @xmath631 to be @xmath229-open at the cost of a constant probability factor . for edges @xmath72 at distance @xmath638 from the boundary , this follows immediately from ( * ? ? ? * lemma 6.3 ) . we briefly sketch how the proof given there can be adapted to the case where @xmath72 is close to the boundary . we write @xmath639 as @xmath640 , where for @xmath641 , @xmath642 denotes the event that @xmath72 has two disjoint @xmath66-open arms to opposite vertical sides of @xmath539 and @xmath623 has two disjoint @xmath52-closed dual arms to the top and bottom of @xmath539 . using russo s formula as in ( * ? ? ? * ( 39 ) ) , we find @xmath643 @xmath644 is the event that @xmath623 has two disjoint @xmath229-closed dual connections to the top and bottom of @xmath539 , and @xmath645 is the event that there exist three disjoint @xmath66-open paths joining , respectively , one vertical side of @xmath539 to one endpoint of @xmath72 , the other endpoint of @xmath72 to an endpoint of @xmath646 , and the other endpoint of @xmath646 to the remaining vertical side of @xmath539 . note that our notation differs somewhat from the one in @xcite . for the purposes of illustration , we will henceforth suppose that @xmath647 for some @xmath648 ; that is , @xmath72 is close to the left side of @xmath539 . the sum on the right of ( [ eq : damronetalsum ] ) can be rewritten as : @xmath649 @xmath650 denotes the left endpoint of the edge @xmath72 , if @xmath72 is a horizontal edge , and its bottom endpoint if @xmath72 is a vertical edge . the first sum is bounded by @xmath651 @xmath652 denotes the probability that there are four arms of alternating occupation status joining @xmath32 to @xmath608 ; @xmath653 is the event that there are two @xmath229-closed arms , as well as a @xmath66-open arm connecting @xmath654 to @xmath593 . using gluing constructions similar to those in proofs of quasi - multiplicativity , and the fact that we may change the length of any connections involved by constant factors at the cost of constant factors in the probabilities , we have : @xmath655 for @xmath656 , we can use ( * ? ? ? * theorem 27 ) to assert @xmath657 we can now follow @xcite exactly ( see equations ( 42 ) and ( 43 ) and the surrounding discussion ) to show that the sum in ( [ eq : fourfactors ] ) is bounded by : @xmath658 to deal with the second sum in ( [ eq : triplesum ] ) , we note that when @xmath659 the conjunction of the events @xmath660 and @xmath645 appearing in the probability on the right of the equation implies that @xmath72 has 2 @xmath66-open , and @xmath623 two @xmath229-closed arms to distance @xmath661 , that @xmath646 has four alternating arms with parameter @xmath66 to the boundary of the intersection of @xmath662 with @xmath539 , three of which reach to distance @xmath661 , and finally that @xmath663 has two @xmath229-closed arms to the top and bottom of @xmath539 and a @xmath66-open arm to the right side of @xmath539 , and all these connections occur inside @xmath539 . using these observations , an argument similar to the previous case and a summation analogous to that in the proof of ( * ? ? ? * lemma 6.2 ) , shows that we can estimate ( in addition , using the remarks in ( * ? ? ? * section 4.6 ) to change the @xmath66-open and closed arms in a half - plane to @xmath229-open and closed arms ) @xmath664 is at distance @xmath442 from the left boundary of @xmath539 and the distance between @xmath646 and @xmath72 is @xmath139 , a number between @xmath665 and @xmath442 . the dark dotted curve represents a @xmath229-closed dual path ( given by menger s theorem ) and the dark solid curve represents a @xmath258-open path , connecting @xmath646 and @xmath72 to each other and to the left and right sides of @xmath539 . the grey dotted curve represents a @xmath258-closed dual path connecting the edge dual to @xmath646 with the top and bottom of @xmath539 . ] turning to the final sum on the left in ( [ eq : triplesum ] ) , we can again closely follow @xcite to bound this term by @xmath666 the estimates outlined above for the left side of ( [ eq : damronetalsum ] ) imply @xmath667 integrating this from @xmath229 to @xmath258 and using ( [ eq : nearcritical ] ) , we find @xmath668 which is what we wanted to prove . we have thus established ( [ eq : ecomp ] ) ; that is , we have shown @xmath669 \lesssim ( { \mathbf{c}}_{30}\log t)^{2k}\cdot(p_n(1)-p_c ) \sum_{e\in j(\mathbf{v})}\mathbb{p}(b_e),\ ] ] where @xmath670 is the number of edges satisfying the conditions in lemma [ crossing - lemma ] . note that @xmath629 is equal to the event that the edge @xmath72 is _ pivotal _ for the existence of a left - right @xmath229-open crossing of @xmath539 . following ( * ? ? ? * lemma 6.2 ) , we can show @xmath671 this concludes the proof of the lemma . we use a block argument and a peierls argument to upgrade ( [ eq : pre - peierls - bound ] ) to ( [ eq - crossing - bound ] ) . the annulus @xmath344 , centred at @xmath672 is tiled with smaller squares of side length @xmath673 the existence of a @xmath258-open crossing of @xmath344 implies that of a crossing @xmath674 of @xmath675 along edges of @xmath54 , with @xmath676 and @xmath677 . the reason for considering this smaller annulus will become clear below . we can now introduce sequences @xmath678 , and @xmath679 relative to the sequence @xmath680 and squares of size @xmath601 ; that is , @xmath681 @xmath682 the first observation is that we have a lower bound on @xmath424 due to the difference in scales : @xmath683 implying @xmath684 the second observation is that @xmath685 where @xmath686 denotes the @xmath208-th coordinate of the vector @xmath367 . from this , for each fixed @xmath424 , given @xmath687 , there are at most 16 choices for @xmath688 and so at most @xmath689 choices for the sequence @xmath690 . the first factor is an estimate for the number of choices of squares @xmath691 with @xmath692 . the third observation is that @xmath693 must contain , between @xmath694 and @xmath695 , a `` short '' crossing @xmath696 of a @xmath697 or @xmath698 rectangle @xmath262 ( that is , the crossing is between the long sides ) . denote by @xmath699 the @xmath700 or @xmath701 rectangle around @xmath262 , as in ( [ eq : pre - peierls - bound ] ) . then @xmath702 and so @xmath703 where @xmath704 is the number of points in @xmath699 connected to @xmath696 by a @xmath229-open path in @xmath699 . it follows that @xmath705 the sum is over all possible finite sequences of squares @xmath706 , for all @xmath707 . this quantity is controlled by choosing a subsequence of @xmath708 disjoint @xmath699 : each rectangle intersects a fixed number of other such rectangles . the events appearing in the last probability are independent for disjoint @xmath699 s . their probability can be bounded using ( [ eq : pre - peierls - bound ] ) ( with @xmath709 ) , since our choice of @xmath601 implies @xmath710 for large @xmath52 . moreover , one can use the bound on the number of sequences of @xmath367 s ( there are at most 4 choices of @xmath262 for a given @xmath687 ) to control the entire sum : the last line in ( [ eq : peierls - sum ] ) is bounded up to a constant factor by : @xmath711 for @xmath52 large enough , this sum is bounded ( again up to a constant ) by : @xmath712 for any @xmath427 . on @xmath713 , any crossing @xmath458 in the portion of the ipc @xmath284 consists of @xmath714-open edges . since any site @xmath229-connected to a site in the ipc also belongs to the ipc , we find that the probability in ( [ eq - crossing - bound ] ) is bounded by : @xmath715 choosing @xmath257 appropriately in the definition of @xmath258 ( depending on the parameter @xmath94 in ( [ eq - crossing - bound ] ) ) establishes the claim .
there exists an open discussion on the validity of online interactions as indicators of real social activity @xcite . most of the online social networks incorporate several types of user - user interactions that satisfy the need for different level of involvement or relation intensity between users @xcite . the cost of establishing the cheapest relation is usually very low , and it requires the acceptation or simply the notification to the targeted user . these connections can accumulate due to the asymmetric social cost of cutting and creating them , and pile up to the astronomic numbers that capture popular imagination @xcite . if the number of connections increases to the thousands or the millions , the amount of effort that a user can invest into the relation that each link represents must fall to near zero . does this mean that online networks are irrelevant for understanding social relations , or for predicting where higher quality activity ( e.g. , personal communications , information transmission events ) is taking place ? by analyzing the clusters of the network formed by the cheapest connections between users of twitter , we show that even this network bears valuable information on the localization of more personal interactions between users . furthermore , we are able to identify some users that act as brokers of information between groups . the theory known as _ the strength of weak ties _ proposed by granovetter @xcite deals with the relation between structure , intensity of social ties and diffusion of information in offline social networks . it has raised some interest in the last decades @xcite and its predictions have been checked in a mobile phone calls dataset @xcite . on one hand , a tie can be characterized by its strength , which is related to the time spend together , intimacy and emotional intensity of a relation . strong ties refer to relations with close friends or relatives , while weak ties represent links with distant acquaintances . on the other hand , a tie can be characterized by its position in the network . social networks are usually composed of groups of close connected individuals , called communities , connected among them by long range ties known as bridges . a tie can thus be internal to a group or a bridge . grannoveter s theory predicts that weak ties act as bridges between groups and are important for the diffusion of new information across the network , while strong ties are usually located at the interior of the groups . burt s work @xcite later emphasizes the advantage of connecting different groups ( bridging structural holes ) to access novel information due to the diversity in the sources . more recent works , however , point out that information propagation may be dependent on the type of content transmitted @xcite and on a _ diversity - bandwidth tradeoff _ @xcite . the bandwidth of a tie is defined as the rate of information transmission per unit of time . aral et al . @xcite note that weak ties interact infrequently , therefore have low bandwidth , whereas strong ties interact more often and have high bandwidth . the authors claim that both diversity and bandwidth are relevant for the diffusion of novel information . since both are anticorrelated , there has to be a tradeoff to reach an optimal point in the propagation of new information . they also suggest that strong ties may be important to propagate information depending on the structural diversity , the number of topics and the dynamic of the information . due to the different nature of online and offline interactions , it is not clear whether online networks organize following the previous principles . our aim in this work is to test if these theories apply also to online social networks . online networks are promising for such studies because of the wide data availability and the fact that different type of interactions are explicitly separated : e.g. , information diffusion events are distinguished from more personal communications . diffusion events are implemented as a system option in the form of _ share _ or _ repost _ buttons with which it is enough to single - click on a piece of information to rebroadcast it to all the users contacts . this is in contrast to personal communications and information creation for which more effort has to be invested to write a short message and ( for personal communication ) to select the recipient . all these features are present in twitter , which is a micro - blogging social site . the users , identified with a username , can write short messages of up to @xmath0 characters ( tweets ) that are then broadcasted to their followers . when a new follower relation is established , the targeted user is notified although his or her explicit permission is not required . this is the basic type of relation in the system @xcite , which generates a directed graph connecting the users : the follower network . after some time of functioning , some peculiar behaviors started to extend among twitter users leading to the emergence of particular types of interactions . these different types of interactions have been later implemented as part of twitter s system @xcite . _ mentions _ ( tweets containing @username ) are messages which are either directed only to the corresponding user or mentioning the targeted user as relevant to the information expressed to a broader audience . a _ retweet _ ( rt @username ) corresponds to content forward with the specified user as the nominal source . in contrast to the normal tweets , mentions usually include personal conversations or references @xcite while retweets are highly relevant for the viral propagation of information @xcite . this particular distinction between different types of interactions qualifies twitter as a perfect system to analyze the relation between topology , strength of social relation and information diffusion in online social networks . the properties of the follower network have been extensively analyzed especially in relation to its topological structure , propagation of information , homophily , tie formation and decay , etc @xcite . finding users with thousands or even millions of followers is not exceptional @xcite , so the question is whether the structure of the follower network carries any information on where personal relations ( mentions ) or information transmission events ( retweets ) take place . to answer this question , we first analyze a sample of the follower network with clustering - detection algorithms and identify a set of groups . our dataset is a sample of the network containing @xmath1 users connected with @xmath2 follower relations , as well as the tweets , retweets , mentions , and was gathered through the twitter api during november and december of @xmath3 @xcite ( see the methods section for further detail ) . whether the clusters we identify are traces of underlying social groups ( online or offline ) is a question we can not answer with the available information . we follow an alternative path by checking the correlation between the location of the personal conversations ( mentions ) and information diffusion events ( retweets ) and the structural properties of the link bearing those activities with respect to the detected groups in the network . note that we consider mentions and retweets to happen always on follower links . this allow us to describe user activity in terms of the detected groups . our first step is to identify the groups in the follower network . clustering in large graphs is still a topic of very active research and many algorithms are available @xcite . due to the size , density , and directness of the follower network and in order to capture the possible inclusion of users in multiple groups or in none , we have used oslom @xcite ( see methods ) . the analysis has also been performed with other clustering techniques @xcite , reaching similar conclusions ( see setion @xmath4 in supplementary information [ figs . s6-s14 and table s1 ] for a detailed account on these results ) . we have detected @xmath5 groups , three of which are graphically depicted in figure 1a with each sphere corresponding to a single user . in general , the links can be classified according to their position with respect to the user groups : internal , between groups , intermediary and links involving nodes not assigned to any group as shown in figure 1b . the statistics characterizing the groups and links are displayed in figure 2 . the group size distribution decays slowly for three orders of magnitude and does not show a characteristic group size ( figure 2a ) . for instance , the largest group contains around @xmath6 users . also the number of groups each user belongs to shows high heterogeneity : @xmath7 of the users has not been allocated to any group , while there exists a user belonging to more than @xmath8 groups ( see figure 2b ) . the percentage of links falling in the different types regarding the groups is depicted in figure 2c . although the non - classified users are @xmath9 of the total , the links connected to them are less than @xmath10 and the percentage is even lower for those with mentions or retweets . the most common type of connections is the between - group links . one may wonder if the algorithm for clusters detection is doing a good job when there is such a large proportion of between - group links . the clustering method is trying to find groups of mutually interconnected nodes that would be extremely rare in a randomized instance of the network , rather than optimizing the ratio between number of between - group and internal links . in sections @xmath11 and @xmath12 of the supplementary information ( figs . s1-s5 ) , this argument is further developed and the capacity of oslom to detect planted communities is proved in a benchmark even in situations with a high ratio between the number of between - groups and internal links . another relevant point to highlight is the different potential of each type of links to carry mentions and retweets . as it can be seen in the figure 2c , the red bars for mentions in internal links and intermediary links almost double the abundance of links in the follower network in these categories . the links between groups , on the other hand , attract far less mentions . of internal links as a function of the group size in number of users . the curve for the follower network acts as baseline for mentions and retweets . note that if mentions / retweets were randomly appearing over follower links then the red / green curve should match the black curve . ( b ) distribution of the number of mentions per link . ( c ) fraction of links with mentions as a function of their intensity . the dashed curves are the total for the follower network ( black ) and for the links with mentions ( red ) . while the other curves correspond ( from bottom to top ) to fractions of links with : 1 non - reciprocated mention ( diamonds ) , 3 mentions ( circles ) , 6 mentions ( triangle up ) and more than 6 reciprocated mentions ( triangle down).,width=325 ] besides their location with respect to the groups , the links can be also characterized by their intensity . in twitter mentions are typically used for personal communication , which establishes a parallelism between links with mentions and strength of social ties . the more mentions has been exchanged between two users , even more so if reciprocated , the stronger we consider the tie between them . we define intensity of a link as the number of mentions interchanged on it . different predictors have been considered to estimate social tie strength @xcite including , for instance , time spent together @xcite or the duration of phone calls @xcite . we consider the intensity as an approximation to social strength given that writing a mention involves some effort and addresses only single targeted users . according to granovetter s theory , one could expect the internal connections inside a group to bear closer relations . mechanisms such as homophily @xcite , cognitive balance @xcite or triadic closure @xcite favor this kind of structural configurations . unfortunately , we have no means to measure the closeness of a user - user relation in a sociological sense in our twitter dataset . however we can verify whether the link has been used for mentions , whether the interchange has been reciprocated or whether it has happened more than once . we define the fraction @xmath13 of links with interaction @xmath14 in position @xmath15 with respect to the groups of size @xmath16 as @xmath17 where @xmath18 is the number of links with that type of interaction in position @xmath15 with respect to the groups of size @xmath16 and @xmath19 in the total number of links with interaction @xmath14 . the fractions @xmath20 reveals an interesting pattern as function of the group size as can be seen in figure 3a . note that the fraction of links in the follower network ( black curve ) is taken as the reference for comparison . links with mentions are more abundant as internal links than the baseline follower relations for groups of size up to @xmath21 users . this particular value brings reminiscences of the quantity known as the dunbar number @xcite , the cognitive limit to the number of people with whom each person can have a close relationship and that has recently been discussed in the context of twitter @xcite . although we have identified larger groups , the density of mentions is similar to the density of links in the follower network . in addition , the distribution of the number of times that a link is used ( intensity ) for mentions is wide , which allows for a systematic study of the dependence of intensity and position ( see figure 3b ) . the more intense ( or reciprocated ) a link with mentions is , the more likely it becomes to find this link as internal ( figure 3c ) . this corresponds to granovetter expectation that the stronger the tie is the higher number of mutual contacts of both parties it of links of the different types ( follower , with mentions and with retweets ) as a function of the size of the group at the link origin , and ( c ) at the targeted group . ( d ) frequency of between - group links as a function of the group - group similarity for the different type of links . in the inset , ratio between the frequency of links with retweets and with mentions.,width=325 ] the next question to consider is the characteristics of links between groups . these links occur mainly between groups containing less than @xmath22 users ( figure 4a - c ) . however , their frequency depends on the quality of the links ( if they bear mentions or retweets ) . while links with mentions are less abundant than the baseline , those with retweets are slightly more abundant . according to the strength of weak ties theory @xcite , weak links are typically connections between persons not sharing neighbors , being important to keep the network connected and for information diffusion . we investigate whether the links between groups play a similar role in the online network as information transmitters . the actions more related to information diffusion are retweets @xcite that show a slight preference for occurring on between - group links ( figures 4b and 4c ) . this preference is enhanced when the similarity between connected groups is taken into account . we define the similarity between two groups , a and b , in terms of the jaccard index of their connections : @xmath23 the similarity is the overlap between the groups connections and it estimates network proximity of the groups . the general pattern is that links with mentions more likely occur between close groups and retweets occur between groups with medium similarity ( figure 4d ) . mentions as personal messages are typically exchanged between users with similar environments , what is predicted by the strength of weak ties theory . links with retweets are related to information transfer and the similarity of the groups between which they take place should be small according to the granovetter s theory . the results show that the most likely to attract retweets are the links connecting groups that are neither too close nor too far . this can be explained with aral s theory about the trade - off between diversity and bandwidth : if the two groups are too close there is no enough diversity in the information , while if the groups are too far the communication is poor . these trends are not dependant on the size of the considered groups ( see section 3 [ figs s6-s14 and table s1 ] in the supplementary information ) . the communication between groups can take place in two ways : the information can propagate by means of links between groups or by passing through an intermediary user belonging to more than one group . we have defined as intermediary the links connecting a pair of users sharing a common group and with at least one of the users belonging also to a different group ( see fig . these users and their links have a high potential to pass information from one group to another in an efficient way @xcite . several previous works pointed out to the existence of special users in twitter regarding the communication in the network @xcite . in order to estimate the efficiency of the different types of links as attractors of mentions and retweets , we measure a ratio @xmath24 for links in position @xmath15 and for interaction @xmath14 defined as @xmath25 where , as before , @xmath26 is the number of links with the interaction @xmath14 in position @xmath15 and @xmath27 is the total number of links in that position . the bar plot with the values of @xmath24 is displayed in figure 5a . the efficiency of the different type of links can thus be compared for the attraction of mentions ( red bars ) and retweets ( green bars ) . links internal to the groups attract more mentions and less retweets than links between groups in agreement with the predictions of the strength of weak ties theory . intermediary links attract mentions as likely as internal links : the fraction of intermediary links with mentions is very close to the fraction of internal links with mentions . this is expected because intermediary links are also internal to the groups . however , the aspect that differentiates more intermediary links from other type of links is the way that they attract retweets . intermediary links bear retweets with a higher likelihood than either internal or between - groups connections ( see figure 5a and section @xmath11 [ figs . s1-s4 ] in the supplementary information ) . this fact can be interpreted within the framework of the tradeoff between diversity and bandwidth @xcite : strong ties are expected to be internal to the groups and to have high bandwidth , while ties connecting diverse environments or groups are more likely to propagate new information . high bandwidth links in our case correspond to those with multiple mentions , while links providing large diversity are the ones between groups . intermediary links exhibit these two features : they are internal to the groups and statistically bear more mentions , and introduce diversity through the intermediary user membership in several groups . although some theoretical works @xcite suggest that ties with high bandwidth and high diversity should be scarce , we find that intermediary links are as abundant as internal links ( see fig . moreover , in line with the theories @xcite , higher diversity increases the chances for a link to bear retweets as can be seen in figure 5b , which implies a more efficient information flow . in the inset of the figure it is shown that the number of non - shared groups assigned to the users connected by the link positively correlates with a higher than expected number of retweets . in summary , we have found groups of users analyzing the follower network of twitter with clustering techniques . the activity in the network in terms of the messages called mentions and retweets clearly correlates with the landscape that the presence of the groups introduces in the network . mentions , which are supposed to be more personal messages , tend to concentrate inside the groups or on links connecting close groups . this effect is stronger the larger the number of mentions exchanged and if they are reciprocated . retweets , which are associated to information propagation events , appear with higher probability in links between groups , especially those that connect groups that do not show a high overlap , and more importantly on links connected to users who intermediate between groups . these intermediary users belong to multiple groups and play an important role in the spreading of information . they acquire information in one group and launch retweets targeting the other groups of which they are members . at the same time , the access to new information can transform them into attractive targets to be retweeted by their followers . the relevance of certain users for the spread of information in online social media has been discussed in previous works . our method provides a way to identify these special users as brokers of information between different groups using as only input the follower network . from the sociological point of view , the way that the activity localizes with respect to the groups allow us to establish a parallelism with the organization of offline social networks.in particular , we have shown that the theory of the strength of weak ties proposed by granovetter to characterize offline social network applies also to an online network . furthermore , some of our results can be explained within the framework of burt s brokerage and closure and aral s diversity - bandwidth tradeoff theories . the specific properties of twitter offers an opportunity to study directly the importance of the links for personal communications or for information diffusion . according to these theories , the strong social ties tend to appear at the interior of the groups or between close groups as happens for the links with mentions in twitter . in addition , the socially weak ties are expected to be more common connecting different groups and to be important for the propagation of information in the network . this is similar to what we observe for the links with retweets that concentrate with high probability in links between dissimilar groups or in intermediary links . besides the roles assigned by these two theories to the links , we have found that intermediary users and links are also an important component to take into account for understanding information propagation . these links tend to be characterized by high bandwidth and diversity in the context of aral s study , and exhibit high information diffusion efficiency . based on all these findings , despite the myth of one million friends and the doubts on the social validity of online links , the simplest connections of the online network bear valuable information on where higher quality interactions take place . .overall characteristics of the follower network and of the interactions taking place on it . [ cols="<,^ , > , > , > , > " , ] [ tab_clus ] in this section , we check the reliability of the results of the main text regarding the localization of the activity when the groups are obtained with clustering techniques different from oslom . the reasons to select oslom as the main method are that ( i ) the software is publicly available , ( ii ) the method is able to analyze the full directed follower network in a reasonable amount of time , ( iii ) it detects the overlapping communities ( bridging nodes and bridges ) and nodes not belonging to any group and ( iv ) the clusters obtained are statistically significant according to a clear null model @xcite . we do not ask the same properties to other methods explored here but they should meet the following conditions : the methods should be available online in the form of software tools , they should be able to deal with relatively large samples of dense graphs , and , if possible , they should include a version to analyze directed networks . we have found several methods satisfying fully or partially these conditions and so we will show in the remainder results with groups detected in the follower network ( or in sub - sampling of it ) by infomap @xcite , moses @xcite , a message - passing algorithm proposed by raghavan et al @xcite that we will refer to as real - time community detection and two algorithms to optimize modularity @xcite : the louvain method for community detection @xcite and a slower modularity optimization algorithm adapted to deal with directed networks that was implemented in the software radatools @xcite . cluster detection in graphs is a topic of very active research @xcite . there are plenty of methods in the literature based on different techniques and with different features and capabilities . the methods selected here are representative of some of the most popular approaches used today for searching groups in networks . modularity optimization is based on a comparison between the number of internal links and the average expected number in a random graph . it has been one of the most popular community detection methods in the last few years , although it is not free problems such as resolution limits @xcite or difficulties to find the absolute maximum of the modularity due to a rough landscape of its value in the space of the possible network partitions @xcite . the louvain method is based on a contraction of the network similar to real - space renormalization and attempts to keep the modularity function constant at each contraction step . moses includes an overlapping stochastic blockmodelling approach as the basis to its community detection algorithm . it provides a procedure for the optimization of the log - likelihood introduced by the blockmodelling approach . oslom is based on a structural approach comparing the internal and external links with the best expectations in an equivalent random graph . in this method , the fact that clustering algorithms follow a process of optimization is incorporated to give an assessment of the probability of finding a similar group in a random network . the method then searches for groups in the given network that have low probability of appearance in the equivalent random graphs . the last tested method , infomap , is based on a different approach , trying to optimize the information fluxes in the given graph by the compartmentalization of the network . infomap is based on the optimization of the description of random walkers paths in the network using information theory concepts the clustering methods can be distinguished by other aspects apart from the approach that they use to find groups . to start with , some as infomap , radatools as well as oslom have specific versions to analyze directed graphs . we used them on the original or sub - samples of the directed follower network . in order to use the other three methods ( moses , louvain , real - time ) , we have symmetrized the network . the symmetrization consists in ignoring the directionality of the links , and considering them as undirected . this procedure neglects information that can be important to define the groups and can affect the performance of the methods . a second difference is the ability to find overlapping communities or bridging nodes . only oslom and moses are able to detect users belonging to more than one group . and , finally , the performance of the methods varies . a way to compare clustering methods is to generate benchmark networks in which the groups are a priori known , then to increase the level of disorder in the connections and to test up to which point the methods recover the planted groups . infomap was found to be one of the best performing methods in this sense in a recent comparative work @xcite , while oslom has been thoroughly tested in a later work @xcite getting results that are comparable to infomap s . a final point to consider is that not all the methods are able to cope with the large size and the high density of our empirical follower network . the computational cost of the analysis makes difficult for some of them to deal with large networks . for this reason , and as can be seen in table [ tab_clus ] , we have run the methods against several samplings of the follower network . these samplings include , when possible , the full network and , when not , subnetworks extracted with different procedures . we have implemented a snow - balling technique similar to the one that led to the collection of the original network and obtained subgraphs using shells of @xmath12 and @xmath4 hops starting from different initial seeding nodes . we have also generated subgraphs by choosing @xmath28 nodes at random and extracting all the links between them . another network sample has been extracted by considering the full graph without the hubs ( nodes with total degree higher than @xmath29 ) . this idea comes from the fact that in twitter there can be users such as celebrities with thousands and millions of followers but whose connections bear no information on bona fide social activity in the network . and , finally , as a sanity check , we have also generated a sub - graph formed by some of the groups detected by oslom . in the results of the main text , the links with mentions are more abundant than the baseline inside groups of size up to @xmath21 users . larger groups do not seem to behave in the same way and the abundance of links with mentions fall upon the baseline . we find a similar signal when the groups are extracted with other clustering algorithms , see the summary of table [ tab_clus ] . the results for the full network for @xmath12 out of @xmath4 algorithms tested are in qualitative correspondence with the results of oslom . these include infomap s results for all the network samples ( infomap is supposed to be one of the most trustable methods for community detection @xcite ) . the results , together with results of oslom for the same sample of the network , are depicted in fig . [ fig_panel_sel0 ] . the figure reveals that the fraction of links with mentions inside of groups is higher than the fraction of any links inside groups irrespectively of the algorithm used . in case of all clustering algorithms ( including oslom ) , for both the sample and the full network , the effect is not visible for groups larger than @xmath21-@xmath30 users ( this number varies for different algorithms ) . when we take into account the number of mentions and whether they are reciprocated , the results show a remarkably consistent pattern . the more mentions , especially reciprocated , the link has , the higher the probability that it is inside of a small group independently of the community detection algorithm used or the sample of the network considered ( see figs . [ fig_panel_sel3nei2b ] to [ fig_panel_sel2oslom5k ] ) . in order to check whether the results on the localization patterns of the activity in the links between groups discussed in the figure 4 of the main text and in figs [ maps ] of this supplementary can be reproduced with groups obtained by other clustering methods , we have repeated the analysis of the group - group links using the groups found by infomap . the results are in figure [ fig_infomap ] . even though the shape of some of the curves is different , the main qualitative results confirm the trends observed with oslom regarding the concentrations of mentions in links between more similar groups , and retweets in those connected groups with medium or low similarity . also the between - group links concentrate a higher quantity of retweets and slightly less mentions . the role of bridging users and bridging connections can be investigated with clustering algorithms capable of detecting overlapping communities , and so capable of assigning nodes to more than one group ( moses and oslom ) . the results for oslom has been presented in the main text . here in fig . [ fig_moses]a we show results for the moses algorithm , run for the sample of the network with removed hubs . the shape of the curves , especially for ratios between distributions for different types of links shown in fig . [ fig_moses]b , is consistent for the two clustering algorithms . the probability of having a retweet over a bridging link , which is proportional to the ratio , steadily grows from almost zero with the number of non - shared groups of the pair of interacting users until it reaches a maximum around value 20 , and then it drops down . the pattern for links with mentions remains consistent for both algorithms , with small maximum for smallest values of number of non - shared groups and steady decay for larger values . the theory of strength of weak ties have two formulations . the macroscopic formulation predicts strong ties to be inside of communities , whereas weak ties as the connectors between these communities . the microscopic formulation states that strong ties happen between users having many friends in common . in the paper , we focus on the communities but here we also show that in fact the microscopic predictions can be confirmed in twitter as well . jaccard similarity of two users is equal to number of shared followers by the two users divided by total number of unique followers the two users have . in fig . [ fig_usim ] we plot the distribution of the similarity for pairs of users who either send a mention or a retweet to each other or are simply connected by a follower tie . the distribution of similarity for the interaction which is supposed to be the most personal ( link with mention ) is shifted to the right , showing that indeed mentions happens much more often between users who share friends . this result and the finding , that mentions are more abundant inside of the communities , are consistent with the expectations of the theory of the strength of weak ties . in this section , we discuss in more detail the imbalance between the number of internal and between - group links that is seen in figure 3c and the effect it may have on cluster detection . the objective of a clustering algorithm is to find areas of the network relatively isolated and dense in internal connections . how is then possible that the number of overall internal links is lower than that of links between groups in the clusters found by oslom ? the answer is that oslom ( as many community detection algorithms ) is not attempting to optimize the balance between internal and between - group links in a direct way . the method searches for areas denser in internal connections than a baseline established by the properties of the random graphs obtained by reshuffling the links of the original network while maintaining the nodes degrees constant @xcite . to illustrate this idea , we have generated a benchmark formed by @xmath31 cliques ( fully connected subgraphs ) of size @xmath32 each . the final graph is then obtained by the addition of @xmath33 links between groups connecting nodes of different cliques at random . to quantify the level of similarity between the original cliques and the groups detected by oslom , we use the normalized mutual information between partitions @xmath34 @xcite . this quantity is equal to one when the two divisions of the network in groups the original cliques and the groups detected are strictly equal and tends to zero when there is no relation between them . the results as function of the ratio between the number of links between groups and the number of internal links ( @xmath35 ) is shown in figure [ nmi ] . oslom is able to detect the planted cliques up to high levels of the ratio internal vs between links , higher in any terms than the values seen in the real follower network . at the same time , the performance of the method improves with larger groups and with a larger number of cliques . cliques ( fully connected subgraphs ) of size @xmath36 each . , width=302 ] the reason for this ability is that the connections between groups are introduced at random , without any clear statistical preference for connections between two particular groups . oslom can detect these random links and ignore them to evaluate which nodes belong to each group despite the high ratio of between - groups links over internal links . we revisit the group similarity as defined in the main text . it is important to recall that the group - group similarity is defined as the jaccard index of the connections of two groups ( the ratio between the number of links in common and total links ) . we consider as the connections of a group all links originating from or targeting the nodes of the group . the similarity of the groups is dependent on the size of the groups of origin and destination of the particular link . figure [ ggsim ] shows how the average similarity of groups connected by follower relations depends on the size . the largest similarity concentrates in links connecting groups of similar size . although one should keep in mind the figure 4a of the main text that shows where most of links are located , namely between groups of size from @xmath37 to @xmath22 , making this region the most relevant . taking the average similarity of the groups for the links with mentions and for the links with retweets and plotting the ratio of these quantities divided by the baseline average similarity of the follower network links , we obtain figures [ maps]a and [ maps]b . the signal reproduces the results of the overall histogram in the figure 4d of the main text : ( i ) the links with mentions tend to connect groups with higher similarity compared to those with retweets or the baseline given by the follower network and ( ii ) the retweets normally happen in links between groups with a medium value of the similarity . the same picture is observed almost independently of the size of the group of origin and destination except for some areas that anyway concentrate a lower density of links . finally , we show results that complement the discussion on bridges of figure 5 of the main text . the fraction of follower links that are bridges , bridges with mentions or with retweets as a function of the groups size can be seen in figure 18a . note that although a node can belong to more than one group , the links usually belong to a single group unless they connect two bridging nodes . the size of the group in figure 18a refers to each of the groups a bridging link belongs to . if a link connects two bridging nodes and so belongs to several groups , the link is counted as many times as groups it is in . as mentioned in the main text , bridges are very effective in attracting retweets . this behavior , present also in figure 18a , contrasts with the one observed in figure 3 of the main text for internal links . however , what is similar between internal and bridges is their attraction for mentions . this attraction increases with the number of times a link has been used for mentions ( see figure 18b ) , also in parallel to the results for pure internal links . the authors would like to thank sergio gmez , andrea lancichinetti and ian leung for making network - clustering software available and for their advice in its use . p.a.g . and j.j.r . acknowledge support from the jae program of the csic . funding was provided by the spanish ministry of science through projects modass ( fis2011 - 24785 ) and mosaico ( fis2006 - 01485 ) . d.m . romero and j. kleinberg , _ the directed closure process in hybrid social - information networks , with an analysis of link formation on twitter _ 4th international aaai conference on weblogs and social media 138145 ( 2010 ) .	an increasing fraction of today social interactions occur using online social media as communication channels . recent worldwide events , such as social movements in spain or revolts in the middle east , highlight their capacity to boost people coordination . online networks display in general a rich internal structure where users can choose among different types and intensity of interactions . despite of this , there are still open questions regarding the social value of online interactions . for example , the existence of users with millions of online friends sheds doubts on the relevance of these relations . in this work , we focus on twitter , one of the most popular online social networks , and find that the network formed by the basic type of connections is organized in groups . the activity of the users conforms to the landscape determined by such groups . furthermore , twitter s distinction between different types of interactions allows us to establish a parallelism between online and offline social networks : personal interactions are more likely to occur on internal links to the groups ( the weakness of strong ties ) , events transmitting new information go preferentially through links connecting different groups ( the strength of weak ties ) or even more through links connecting to users belonging to several groups that act as brokers ( the strength of intermediary ties ) .
the phenomenon of superconductivity remains one of the most interesting problems of condensed matter physics . in particular , in recent years the material science development has revealed several interesting systems where high transition temperature superconductivity was found , in particular the families of high-@xmath6 cuprates @xcite with a maximum @xmath6 of about 155 k. soon after their discovery it was realized that the essential physics of cuprates takes place in the cuo@xmath0-planes which is believed to be responsible for the high transition temperature . sr@xmath7cuo@xmath1 and ( b ) layered perovskite structure of the triplet superconductor sr@xmath0ruo@xmath1 . both structures are identical if la ( ba ) is replaced by sr and the cuo@xmath0-plane is replaced by ruo@xmath0-plane.,title="fig:",scaledwidth=40.0% ] sr@xmath7cuo@xmath1 and ( b ) layered perovskite structure of the triplet superconductor sr@xmath0ruo@xmath1 . both structures are identical if la ( ba ) is replaced by sr and the cuo@xmath0-plane is replaced by ruo@xmath0-plane.,title="fig:",scaledwidth=30.0% ] in this connection the discovery of superconductivity in sr@xmath0ruo@xmath1 with @xmath8 k @xcite is of particular interest for several reasons . first , the crystal structure of sr@xmath0ruo@xmath1 is identical to that of the parent compound of the high-@xmath6 superconductor la@xmath0cuo@xmath1 ( see fig . [ fig1 ] for illustration ) . both kinds of materials are highly two - dimensional and as fig . [ fig1 ] shows the structure is almost identical to that of the la@xmath9sr@xmath7cuo@xmath1 superconductors . both materials are oxides with conduction occurring in partially filled @xmath10bands that are strongly hybridized with the oxygen @xmath5-orbitals . therefore , it was generally believed that a comparison of the normal and superconducting properties of the cuprates and sr@xmath0ruo@xmath1 will give a deeper understanding of the nature of the high-@xmath6 in the cuprates . however , it has been found that the differences between sr@xmath0ruo@xmath1 and the cuprates are larger than their general similarities might suggest . in sr@xmath0ruo@xmath1 superconductivity occurs only at low temperatures and the normal state is a well - defined fermi - liquid . this contrasts strongly with the anomalous normal state of the cuprates . furthermore , it was soon found that the superconductivity in sr@xmath0ruo@xmath1 is very interesting on its own . in particular , there are clear indications that the superconducting state is unconventional . for example , the transition temperature is highly sensitive to impurities @xcite and nuclear quadrupole resonance(nqr ) experiments do not show a hebel - slichter peak in the spin - lattice relaxation at @xmath6 @xcite . shortly after the discovery of sr@xmath0ruo@xmath1 it was suggested that superconductivity might arise from odd - parity ( spin - triplet ) cooper - pairs with total spin @xmath11 and a non - zero angular momentum which is reminiscent of the phases of superfluid @xmath12he @xcite . the basis for this suggestion was the presence of ferromagnetism in the related compounds as srruo@xmath13 and thus the expectation of ferromagnetic fluctuations in metallic sr@xmath0ruo@xmath1 . to support this picture the model phase diagram shown in fig . [ phasesr ] has been suggested . ru@xmath14o@xmath15 , taken from ref . the number of layers is the parameter which determines the transition between the two phases : mo = magnetically ordered and sc = superconducting . this phase digram suggests ferromagnetic excitations in the normal state of sr@xmath0ruo@xmath1.,scaledwidth=50.0% ] here , one plots the phase of the ferromagnetic and superconducting members of the so - called rodllesen - popper series ( sr@xmath16ru@xmath14o@xmath15 ) as a function of the numbers of ruo@xmath0-layers per unit cell , @xmath17 . the infinite layer ( srruo@xmath13 ) is a ferromagnet with @xmath18 k. for @xmath17=3 one finds @xmath19 k and for @xmath17=2 the substance orders magnetically at @xmath20 k. this demonstrates the tendency that @xmath21 is reduced with decreasing layer number @xmath17 and suggests that even for @xmath17=1 , when supeconductivity occurs , one expects significant ferromagnetic fluctuations which may play an important role for triplet superconductivity in sr@xmath0ruo@xmath1 . meanwhile , a number of experiments indeed point towards spin - triplet cooper - pairing . the most convincing evidence comes from the @xmath22o nmr knight shift data which shows that the spin susceptibility is not affected by the superconducting state for a magnetic field parallel to the ruo@xmath0-plane @xcite . in fig . [ triplet ] we show the corresponding experimental results . in conventional superconductors the spin part of the knight shift measured by nmr in the superconducting state of sr@xmath0ruo@xmath1 as measured by nmr knight shift @xcite . one clearly sees that in contrast to the singlet cooper - pairing ( illustrated by the dashed curve ) where spin susceptibility decreases upon cooling , the knight shift in sr@xmath0ruo@xmath1 is unchanged by lowering @xmath23 below @xmath6 . the knight shift doe not decrease below @xmath6 , since the polarization induced by the external magnetic field does not change in the superconducting state in case of spin - triplet cooper - pairs.,scaledwidth=50.0% ] decreases rapidly below @xmath6 due to the formation of singlet cooper - pairs . on the other hand , in a triplet superconductor with s=1 the spin part of the knight shift should not change below @xmath6 for some field orientations , since the polarization induced by the weak external magnetic field and probed by nmr does not change . this behavior was observed in sr@xmath0ruo@xmath1 by ishida _ _ @xcite and provides strong evidence for triplet cooper - pairing . however , recently it has became clear that the situation is not that simple . for example , the intensive studies by means of inelastic neutron scattering ( ins ) reveal the presence of strong two - dimensional antiferromagnetic spin fluctuations at * q@xmath24 and no sign of ferromagnetic fluctuations in the normal state of sr@xmath0ruo@xmath1 @xcite . this fact , at first glance , can not be compatible with triplet cooper - pairing and thus the role of the antiferromagnetic fluctuations in the formation of superconductivity has to be clarified . moreover , further analysis of the nmr data indicated the magnetic response to be strongly anisotropic @xcite . in particular , the measured longitudinal component of the spin susceptibility @xmath25 is much larger than the transverse one @xmath26 , while from an isotropic model one expects @xmath27 . this anisotropy increases with decreasing temperature and reaches a maximum close to @xmath6 indicating its influence on the superconducting properties . therefore , in view of the importance of spin fluctuations for the unconventional superconductivity in sr@xmath0ruo@xmath1 their behavior should be understood in more detail . another interesting question concerns the symmetry of the superconducting order parameter . while the analogy to the case of superfluid @xmath12he leads to the suggestion of nodeless @xmath5-wave superconductivity in sr@xmath0ruo@xmath1 , recent analysis of the specific heat data @xcite indicates a more complicated and anisotropic behavior of the superconducting order parameter in the first brillouin zone ( bz ) . therefore , the physics of sr@xmath0ruo@xmath1 is far from being completely understood . note , an extensive review on the experimental situation can be found in ref . @xcite . in contrast to that review , in the present review we mainly analyze the role played by strong electronic correlations and spin fluctuations in sr@xmath0ruo@xmath1 . so far their role in determining the superconducting and normal state properties has not been addressed in detail . probably , this was due to the success of the conventional band theory in explaining the various properties of sr@xmath0ruo@xmath1 . however , its success is mainly due to the correct description of the fermi surface topology in this compound . the fermi surface shape , however , is hardly affected by the electronic correlations , while on the other hand the bandwidth and the energy dispersion are strongly modified . we will show that this is indeed the case for the ruthenates . furthermore , we review the spin - fluctuation mediated cooper - pairing scenario in application to sr@xmath0ruo@xmath1 and address the important role of spin - orbit coupling in this system . in particular , we discuss the symmetry of the relevant superconducting order parameter within this scenario and , finally , investigate the role of impurities in sr@xmath0ruo@xmath1 . in sr@xmath0ruo@xmath1 the formal valence of the ruthenium ion is ru@xmath28 . this leaves four electrons remaining in the @xmath29-shell . furthermore , the ru ion sits at the center of a ruo@xmath30-octahedron and the crystal field of the o@xmath31 ions splits the five @xmath29-states into threefold ( @xmath32 ) and twofold ( @xmath33 ) subshells as illustrated in fig . [ rutenstr0 ] . the negative charge of the @xmath34 ions causes the @xmath32 subshell to lie lower in energy . note , that these orbitals ( @xmath35 , @xmath36 , and @xmath37 ) have lobes that point between the @xmath34 ions lying along the @xmath38 , @xmath39 , and @xmath40-axes . electrons of these orbitals form the fermi surface ( the so - called @xmath41 , @xmath42 , and @xmath43-bands ) . we assume that the most important interaction of the carriers is with spin excitations described by the spin susceptibility @xmath44 and resulting from ru @xmath41- , @xmath42- , and @xmath43-states . the ru@xmath28 ion corresponds to a 4d@xmath45 level . the splitting of the @xmath33 and @xmath32 subshells is due to the ruo@xmath30 crystal field . the orbitals d@xmath46 , d@xmath47 , and d@xmath48 cross the fermi level and form the @xmath41- , @xmath42- , and @xmath43-band . the resulting band structure is quasi - two - dimensional . magnetic activity results from the @xmath32 subshell.,scaledwidth=40.0% ] then , we have 3 bands formed out of the oxygen @xmath49-@xmath50 hybridized orbitals and with a filling factor of 4/3 . since @xmath49-@xmath50 bonding is weaker than the @xmath49-@xmath51 realized in cuprates , the admixture of the oxygen orbitals is smaller in ruthenates than in cuprates . for example , the contribution of the oxygen @xmath5-states close to the fermi level is only 16 percent in sr@xmath0ruo@xmath1 . at the same time the contribution of the ru 4@xmath2-orbitals is about 84 percent @xcite . lda calculations @xcite indeed confirm the existence of the three bands crossing the fermi level . the band structure of sr@xmath0ruo@xmath1 is quasi - two - dimensional and the electronic dispersion along the @xmath52-axis is very small @xcite . ruo@xmath1 using eq . the fermi surface consists of three sheets : two with electron - like topology and one with hole - like topology . note , the quasi - one - dimensional character of the @xmath41(xz)- and @xmath42(yz)-bands . , scaledwidth=50.0% ] this is because the highly planar structure of sr@xmath0ruo@xmath1 prevents substantial energy dispersion along the @xmath40-direction , due to the large interplanar separation of the ruo@xmath30 octahedra . at the same time , in the @xmath53-plane neighboring ruo@xmath30-octahedra share o ions which in turn are @xmath50-bonded to the ru ions . thus , the @xmath37-orbital will form a two - dimensional band , while the @xmath35- and @xmath36-bands have only a restricted one - dimensional character . then , to describe the lda results the following tight - binding energy dispersions can be introduced @xmath54 the lattice constants @xmath55 have been set to unity . this model is a good parameterization of lda calculations @xcite if one chooses for @xmath56- , @xmath57- , and @xmath58-orbitals the values for the parameter set ( @xmath59 ) as ( 0.5 , 0.42 , 0.44 , 0.14 ) , ( 0.23 , 0.31 , 0.045 , 0.01 ) , and ( 0.23 , 0.045 , 0.31 , 0.01 ) ev . the experimental investigation of the fermi surface made by means of de - haas - van - alphen effect completely confirmed the lda results @xcite . the resulting fermi surface is shown in fig . [ fermruten ] . here , the @xmath41-band is hole - like centered around @xmath60 point of the brillouin zone , while @xmath42 and @xmath43-bands are electron - like ones centered around @xmath61-point . furthermore , as one sees the @xmath41 and @xmath42-bands are quasi - one - dimensional and their degeneracy is removed by the introduction of the interband ( between @xmath41 and @xmath42-bands ) hopping @xmath62= 0.025 ev @xcite . despite of the large difference in the magnitudes of the resistivity in the ruo@xmath0-plane and perpendicular to the planes , @xmath63 , the temperature dependences of @xmath64 and @xmath65 for @xmath66 k both follow the fermi - liquid behavior @xmath67 @xcite . above @xmath68 k small deviations from this law occur . at higher temperatures @xmath65 becomes linear with a temperature dependence similar to the cuprates . as already mentioned in the introduction , the spin dynamics of sr@xmath0ruo@xmath1 consists of ferromagnetic and incommensurate antiferromagnetic spin fluctuations at wave vector q@xmath69 . its role in the formation of triplet superconductivity in sr@xmath0ruo@xmath1 requires special consideration . long - range magnetic order is absent in sr@xmath0ruo@xmath1 . on the other hand the uniform magnetic susceptibility is much larger than the pauli susceptibility of the non - interacting electrons . indeed the ab - initio calculations of the stoner factor @xmath70 ( where @xmath71 is the exchange parameter at @xmath72 , @xmath73 is the density of states at the fermi level ) gives the following values @xmath74 @xcite , @xmath75 @xcite . using a simple so - called stoner criterium , this indicates that sr@xmath0ruo@xmath1 is in the vicinity of the instability to ferromagnetism . in srruo@xmath13 the corresponding stoner parameter is @xmath76 @xcite . therefore , this system is a ferromagnetic metal with magnetic moment @xmath77 and curie temperature @xmath78 k. the reason for the enhanced ferromagnetic exchange fluctuations in srruo@xmath13 in comparison to sr@xmath0ruo@xmath1 is the stronger @xmath49 hybridization and thus a larger contribution of the oxygen @xmath5-orbitals in the density of states at the fermi level . however , what is interesting in sr@xmath0ruo@xmath1 is not only the presence of ferromagnetic fluctuations at @xmath72 ( from the stoner instability ) , but also the incommensurate antiferromagnetic fluctuations at the wave vector @xmath79 as observed in inelastic neutron scattering experiments @xcite . what is the origin of these fluctuations ? in fact they originate from the nesting properties of the quasi - one - dimensional @xmath35- and @xmath36-bands . ( @xmath80 refers to the band indices ) in the normal state of sr@xmath0ruo@xmath1 along the route @xmath81 in the first bz for the three different orbitals ( @xmath35,@xmath36 , and @xmath37 ) crossing the fermi level . due to the nesting of the @xmath35 and @xmath36-bands their susceptibilities show an enhancement at the incommensurate antiferromagnetic wave vector @xmath82 . the response of the @xmath37-band is more isotropic , but significantly larger than in the normal metal due to the vicinity of the van - hove singularity . , scaledwidth=60.0% ] in order to see this we show in fig . [ ruthiohne ] the results for the lindhard response functions calculated for the different bands using their tight - binding dispersions @xmath83 where @xmath84 refers to the band index and @xmath85 is the fermi distribution function . as seen in fig . [ ruthiohne ] , due to the pronounced nesting of the @xmath35 and @xmath36-bands their susceptibilities display peaks at q@xmath86 , while the @xmath37-band does not show any significant feature . the response of the @xmath37-band is enhanced due to the presence of the van - hove singularity close to the fermi level . then , it becomes clear that the features observed by ins relate mainly to the magnetic response of the @xmath35- and @xmath36-bands . furthermore , as has been shown in ref . @xcite that the stoner enhancement related to the appearance of the incommensurate antiferromagnetic fluctuations is even stronger than the ferromagnetic one and is of the order of 1 . this means that sr@xmath0ruo@xmath1 is almost unstable with respect to the spin density wave formation ( sdw ) . note , that both ferromagnetic and antiferromagnetic fluctuations are seen experimentally in the normal state of sr@xmath0ruo@xmath1 by means of @xmath22o knight shift @xcite and by inelastic neutron scattering @xcite , respectively . these results suggest that an effective hamiltonian for describing the physics in sr@xmath0ruo@xmath1 is a two - dimensional three - band hubbard hamiltonian @xmath87 where @xmath88 is the fourier transform of the annihilation operator for the @xmath89 orbital electrons ( @xmath90 ) and @xmath91 is an effective on - site coulomb repulsion . the hopping integrals @xmath92 denote the energy dispersions of the tight - binding bands , see eq.([lind ] ) . as a result the spin dynamics in sr@xmath0ruo@xmath1 is determined by the competition and the delicate balance between ferromagnetic and incommensurate antiferromagnetic fluctuations . the natural question arises what is the influence of these magnetic fluctuations on the triplet superconductivity in sr@xmath0ruo@xmath1 ? in the following subsection we will study the formation of triplet of superconductivity in sr@xmath0ruo@xmath1 driven by the spin fluctuation exchange . in the remainder of this section we discuss another aspect of the ruthenates , namely strong electronic correlations . as a matter of fact the success of ab - initio methods in describing magnetism of sr@xmath0ruo@xmath1 appear to invalidate , at first sight , the discussion of the role of strong electronic correlations . on the other hand , the effects of strong electronic correlations are small if the bandwidth of the conduction band @xmath93 is much larger than the on - site coulomb repulsion , @xmath94 . given that the bandwidth is of the order of 1 ev there is no reason to think that @xmath95 in sr@xmath0ruo@xmath1 . to estimate the value of @xmath94 one has to remember that for 4@xmath2-electrons @xmath94 has to be smaller than for @xmath96-electrons , since the average radius of the 4@xmath2-shell is approximately two times larger than for the 3@xmath2-shell . then a simple estimates gives u@xmath97 ev . this means that sr@xmath0ruo@xmath1 is in the regime of intermediate electronic correlations when @xmath98 . as we mentioned earlier , despite of the importance of the electronic correlations in sr@xmath0ruo@xmath1 the relative success of the band structure theory can be easily explained . the correct description of the ferromagnetic and antiferromagnetic fluctuations comes from the correct fermi surface topology which is not affected by the electronic correlations . however , the bandwidth and the energy dispersion are strongly affected by the electronic correlations in sr@xmath0ruo@xmath1 . for example , it has been shown by liebsch and lichtenstein @xcite that in order to explain the bandwidth as measured by angle - resolved photoemission spectroscopy ( arpes ) the electronic correlations have to be taken into account . in particular , using dynamical - mean - field theory ( dmft ) they have shown that , while the fermi surface topology is the same independent on electronic correlations , the bandwidth is different , namely the correlations reduce the bandwidth of the all bands in sr@xmath0ruo@xmath1 . we will further address this issue discussion the phase diagram of ca@xmath9sr@xmath7ruo@xmath1 . even before experimental results have indicated triplet superconductivity in sr@xmath0ruo@xmath1 , its possibility had been theoretically predicted by sigrist and rice @xcite . in particular , they considered the triplet superconductivity on a square lattice as an electronic analogy of the superfluid a - phase in @xmath12he . for the spin - triplet cooper - pairing the wave function can be written as a matrix in spin space @xmath99 where the eigenvalues of the s@xmath100 operator with projection + 1 , 0 , -1 have the form @xmath101 another possibility to express the cooper - pair wave function is to use the basis of the symmetric matrices @xmath102 then the cooper - pair wave function has the form @xmath103 the components of the vector d * can be expressed linearly via the amplitudes of @xmath104 : @xmath105 due to the pauli principle the orbital part of the wave function with total @xmath11 has to be odd , i.e. the orbital quantum number @xmath106 . therefore , one can speak about @xmath107 cooper - pairing in a continuum system . in particular , the amplitudes g@xmath108 and the vector * d(k ) * have to be the odd functions of the momentum * k. in the tetragonal crystal structure such as that of sr@xmath0ruo@xmath1 , the order parameter d(k ) * has to belong to the corresponding irreducible representation of the d@xmath109 group . there are several representations @xcite and the proper choice can be made by comparison to experiment . for example , the spin state of the cooper - pair wave function can be obtained from the measuring the knight shift in nmr and polarized inelastic neutron scattering experiments below @xmath6 . the orbital state can be measured by muon spin relaxation ( @xmath110sr ) indicating time - reversal symmetry breaking . the square symmetry of the vortex lattice can also be interpreted as evidence for @xmath5-wave pairing . as we have shown in the introduction , the nmr knight shift experiments indicate spin - triplet cooper - pairing in sr@xmath0ruo@xmath1 @xcite . the same conclusion has been drawn from the polarized neutron scattering studies @xcite . the time - reversal symmetry breaking detected by means of the @xmath110sr @xcite suggests the superconducting order parameter @xmath111 with @xmath112 and @xmath113 ( so - called chiral @xmath5-wave state ) : @xmath114 the amplitude of this gap is isotropic , @xmath115 , and does not have a node at the fermi surface of sr@xmath0ruo@xmath1 . so far , the above analysis is quite general . next , we assume the superconductivity is driven by the exchange of spin fluctuations . the interesting question in the ruthenates is why does triplet cooper - pairing take place in the presence of strong incommensurate antiferromagnetic spin fluctuations at @xmath82 originating from the @xmath35- and @xmath36-bands and only relatively weak ferromagnetic fluctuations arising from the @xmath37-band . in general , to answer this question the bcs gap equation ( see eq . ( [ gaprut ] ) ) must be solved assuming different possible symmetries of the order parameter and corresponding pairing interaction for singlet , triplet with @xmath116 projection and triplet with @xmath117 projection cooper - pairing . for the analysis of the interaction between quasiparticles and spin fluctuations ( ferromagnetic and incommensurate antiferromagnetic ) we use the generalized eliashberg like theory for the interaction between quasiparticles and spin fluctuations @xcite . we extend this theory to become a three - band theory , as necessary in sr@xmath0ruo@xmath1 @xcite . in the normal state both the self - energy and thermal green s functions become matrices of @xmath118 form , _ i.e. _ @xmath119 and @xmath120 , where @xmath121 refer to the band indexes of the @xmath122 and @xmath35-orbitals . the corresponding dyson equation is given by @xmath123^{-1 } = \left[\hat{g}^0 ( { \bf k},\omega_n)\right]^{-1 } - \hat{\sigma } ( { \bf k},\omega_n ) , \label{dyson3}\ ] ] where @xmath124 is the matrix of the bare green s function determined via the tight - binding energy dispersions for the @xmath37 , @xmath36 , and @xmath35-bands . the self - energy is given by @xmath125 where @xmath126 is an effective interaction [ gamma ] between quasiparticles and spin fluctuations . as shown in fig.7 it consists of an infinite series of diagrams including charge and spin fluctuations . this is similar to the case of cuprates , however , some important differences are present in the random phase approximation . most importantly we must now consider the diagrams shown in fig . [ tripltheory ] with an odd number of bubbles that contribute to the triplet pairing . this is in contrast to singlet pairing in cuprates where an even number of bubbles occur ( see fig . [ gammasf ] ) . furthermore , the transverse ( + - ) and the longitudinal ( zz ) parts of the spin susceptibility are different . the reason for this ( as we show later ) is the presence of spin - orbit coupling . we discuss its role in detail in the next section . then the effective pairing interaction in the @xmath118 form including transverse and longitudinal spin fluctuations and also charge fluctuations is given by @xmath127 where @xmath128 describe coupling to spin density fluctuations and @xmath129 to charge density fluctuations . @xmath130 , @xmath131 are the irreducible parts of the charge and spin susceptibilities respectively . note that in the bethe - salpether equation shown diagrammatically in fig . [ gammasf ] for singlet cooper - pairing it is necessary to have an even number of bubbles and include the ladder diagrams . in the case of triplet cooper - pairing the contribution of the ladder diagrams is zero and an odd number bubbles diagrams occur due to pauli s principle shown in fig . [ tripltheory ] . since the feynman rules require a @xmath132 for each loop an extra minus sign enters the gap equation via @xmath133 . the magnetism in the ruthenates resulting from the quasiparticles in the @xmath32-orbital is itinerant and thus the magnetic response is created by the same electrons that form also the cooper - pairs . then , for example , the irreducible part of the charge susceptibility @xmath130 is defined in terms of the electronic green s functions @xmath134 where @xmath135 is the single - electron green s function ( we omit here the band indices for simplicity ) . the longitudinal and transverse components of the spin susceptibilities are also calculated in terms of the electronic green s functions . in contrast to the cuprates , in the ruthenates @xmath136 and @xmath137 are different due to the magnetic anisotropy @xcite . in the ruthenates the general matrix form of the superconducting gap equation for @xmath138 is similar to the case of cuprates . to determine the superconducting transition temperature @xmath6 one must solve the following set of linearized gap equations to obtain @xmath139 : @xmath140 here , @xmath141 is an eigenvalue and @xmath6 is determined from the condition @xmath142 . note , the interband coupling will provide a single @xmath6 for all three bands . the pairing potential @xmath143 ( taken for singlet or triplet pairing ) controls which state gives the lowest energy and determines singlet or triplet cooper - pairing . three possibilities may occur : + ( a ) singlet pairing , + ( b ) triplet pairing with the total spin s@xmath144 of the cooper - pair + wave function , + ( c ) triplet pairing with the total spin s@xmath145 . + note , @xmath146 @xmath147 one can not judge which pairing state is realized in the ruthenates due to the presence of both antiferromagnetic and ferromagnetic fluctuations . therefore , one has to solve the gap equations for all three possibilities . thus , for singlet pairing we use @xmath148 for triplet pairing with s@xmath145 we take @xmath149 resulting from the set of diagrams with odd number of bubbles ( see fig . [ tripltheory ] ) . finally , for triplet pairing with @xmath150 we get @xmath151 therefore due to magnetic anisotropy we have separated the longitudinal and transverse parts of the spin fluctuations . note , transverse and longitudinal spin fluctuations contribute differently to the singlet and triplet cooper - pairing . in particular , in the spin - triplet case the longitudinal and transverse spin fluctuations do not act additively as in the case of singlet cooper - pairing . as a result , a strong reduction of the superconducting transition temperature @xmath6 is expected . the effect of charge fluctuations is expected to be small in the superconductivity in sr@xmath0ruo@xmath1 . the pairing state symmetry is determined at @xmath6 by the eigenvector of eq.(16 ) corresponding to @xmath152 . as well as in the case of superfluid @xmath12he many other possible states besides the @xmath153-phase can occur . note , that recent experiments indicate the occurrence of the line nodes in the superconducting state of sr@xmath0ruo@xmath1 . these are the power rather an exponential laws of the specific heat , @xmath154 @xcite , spin - lattice relaxation rate @xmath155 @xcite , the heat capacity @xmath156 @xcite , and ultrasound attenuation @xcite . therefore , other symmetries of the superconducting order parameter like @xmath5-wave symmetry with nodes or @xmath157-wave symmetry which are other than simple nodeless @xmath5-wave symmetry of the superconducting gap have to be considered . note , the largest eigenvalue of eq . ( [ gaprut ] ) corresponds to a minimum in the free energy and thus will yield the symmetry of the superconducting order parameter @xmath158 in sr@xmath0ruo@xmath1 . one could expect that the formation of the node should correspond to the influence of the incommensurate antiferromagnetic fluctuations . let us see whether the simple nodeless @xmath5-wave symmetry is possible if the iaf at the wave vector * q=@xmath159 are present . using appropriate symmetry representations @xcite we discuss the solutions of the gap equation assuming @xmath5 , @xmath2 , or @xmath157-wave symmetry of the order parameter in the ruo@xmath0-plane @xcite : @xmath160 here the @xmath161-unit vector refers to the @xmath162 component of the cooper - pairs as observed in the experiment @xcite . we will present a possible explanation for this below . these symmetries of the superconducting order parameter must be substituted into the eq.([gaprut ] ) . for simplicity we consider first the solution of eq . ( [ gaprut ] ) in the ruo@xmath0-planes and then discuss what is happening along the @xmath52-axis . solving eq . ( [ gaprut ] ) in the first bz down to 5 k we have found that @xmath5-wave symmetry yields the largest energy gain for the @xmath163band , while for the @xmath164 and @xmath165bands the situation is more complicated . we find that the gap equations for the @xmath37- and @xmath165 , @xmath35-bands can be separated due to their weak interaction . the result is that for the @xmath37-band the @xmath5-wave is the most stable solution , while for the @xmath164 and @xmath165bands @xmath166-wave symmetry yields the largest eigenvalue due to stronger nesting . = 0 ( solid curves ) and @xmath167 ( dotted curves ) . @xmath41 , @xmath42 , and @xmath43 denote the fs of the corresponding hybridized bands . the wave vectors q@xmath168 and q@xmath168 are the main wave vectors contributing to the susceptibility ( without spin - orbit coupling ) . for @xmath169-wave symmetry the nodes of the real part of the order parameter are shown ( dashed lines ) and the regions + ( - ) where the @xmath166-wave superconducting gap is positive ( negative ) . note , for the real part of a @xmath5-wave order parameter the node would occur only along @xmath170=0 . , scaledwidth=50.0% ] how to understand the competition between @xmath5- and @xmath157-wave pairing ? . for doing this we display in fig . [ fermif ] characterizes the solutions of eq . ( [ gaprut ] ) . note , a good approximation is to linearize eq . ( [ gaprut ] ) with respect to @xmath171 , i.e. @xmath172 . we use also @xmath173 . thus , the main contribution to the pairing comes from the states at the fermi level . we present our results in terms of the fermi surface of the ruo@xmath0 plane . the wave vectors q@xmath168 and q@xmath168 refer to the peaks in @xmath174 . the areas with @xmath175 and @xmath176 are denoted by ( + ) and ( - ) , respectively . note , that the minus sign in eq . ( [ gaprut ] ) is absent for triplet pairing . the summation over @xmath177 in the first bz is dominated by the contributions due to @xmath178 for the @xmath41- and @xmath42-bands and the one due to @xmath179 for the @xmath43-band . as can be seen from fig . [ fermif ] in the case of @xmath166-wave symmetry for the @xmath163band the wave vector @xmath179 bridges the same number of portions of the fs with opposite and equal sign . therefore , for the @xmath37-band no solution with @xmath166-wave symmetry is possible . on the other hand , for the @xmath164 and @xmath36-band the wave vector @xmath178 bridges portions of the fs with _ equal _ signs of the @xmath166-superconducting order parameter . the eigenvalue of this order parameter is also enhanced due the interband nesting for @xmath35 and @xmath36-bands . thus , superconductivity in the @xmath35- and @xmath36-band in the ruo@xmath0 plane is indeed possible with @xmath166-wave order parameter . therefore , solving eq . ( [ gaprut ] ) for the three band picture we found a competition between triplet @xmath5-wave and @xmath166-wave superconductivity in the ruo@xmath0-plane . we can immediately rule out singlet cooper - pairing . in particular , assuming @xmath180-symmetry for sr@xmath0ruo@xmath1 the eq . ( [ gaprut ] ) yields an eigenvalue which is lower than in the case of triplet pairing . as can be seen using fig . [ fermif ] this is plausible . note , we get for d@xmath181-wave symmetry a change of sign of the order parameter upon crossing the diagonals of the bz . according to eq . ( [ gaprut ] ) wave vectors around q@xmath168 connecting areas ( + ) and ( - ) contribute constructively to the pairing . contributions due to q@xmath168 and the background connecting areas with the same sign subtract from the pairing ( see fig . [ fermif ] with nodes at the diagonals ) . we find that the four contributions due to q@xmath168 in the @xmath182 do not allow @xmath183-wave symmetry in the @xmath163band . despite the pair - building contribution due to q@xmath168 one finds that the eigenvalue of the @xmath183-wave symmetry in the @xmath37-band is smaller than for the @xmath166-wave symmetry . this is due to the large contribution from q@xmath168 to the cross - terms for the triplet pairing which are absent for singlet pairing . for @xmath56-symmetry where the nodes are along ( @xmath50,0 ) and ( 0,@xmath50 ) directions we argue similarly . thus , we exclude this symmetry . the @xmath184-wave symmetry solution has vertical lines of nodes along the @xmath185-axis . in principle , this has to be observable in the strong four - fold anisotropy of the thermal conductivity @xmath186 in applied magnetic field along @xmath53-plane . one has to admit , however , the recent measurements have shown only a weak anisotropy @xcite that suggests the line nodes in sr@xmath0ruo@xmath1 are not vertical . a phenomenological mechanism to produce horizontal line nodes has been proposed by zhitomirskiy and rice @xcite . as we have already mentioned , the magnetic fluctuations are different with respect to the various bands . the ferromagnetic fluctuations are originating from the @xmath37-band while the incommensurate antiferromagnetic fluctuations are due to the nesting of @xmath35 and @xmath36-bands . in this case it was proposed that in the active @xmath37-band the superconducting gap has no nodes and can be described by the nodeless @xmath5-wave gap . then , due to interband scattering of the cooper - pairs the superconductivity is induced in the @xmath35 and @xmath36-bands with the symmetry @xmath187 note , this gap has the horizontal line of nodes at @xmath188 . then even a combination of the nodeless @xmath5-wave symmetry and that with nodes will still be present , only the position of the nodes will be slightly shifted along @xmath52-axis . this behavior correctly describes the temperature dependence of the specific heat below @xmath6 @xcite . a similar model was proposed by annett _ _ @xcite . in this model there are two distinct attractive pairing interactions . one , acting between nearest neighbor in - plane @xmath37 orbitals , leads to a nodeless chiral @xmath5-wave state on the @xmath43-band . the second acts between nearest neighbors between planes , and is assumed to act primarily on the d@xmath47 and d@xmath48 orbitals , which have lobes perpendicular to the planes . this second interaction produces a gap with line nodes on the @xmath41 and @xmath42 fermi surface sheets . this gap model was shown to be in quite good agreement with experiments on specific heat @xcite , penetration depth @xcite and thermal conductivity @xcite . similar to the zhitomirskiy rice approach @xcite , this model is also orbital dependent superconductivity , but in this case there are no active and passive bands , since all bands have a pairing interaction and the gap is of similar magnitude on all three fermi surface sheets . alternatively , interlayer coupling models have been proposed @xcite . these also give rise to gap functions of the form of eq . ( [ gapc ] ) , either on one or all of the three fermi surface sheets . the physical mechanisms responsible for the interlayer coupling and hence the gap node are assumed to be either interlayer coulomb interactions , or dipole - dipole interactions . the question remains whether spin fluctuations can also explain the formation of the nodes along the @xmath52-axis in the passive @xmath35 and @xmath36-bands . in order to explain this we have first to take into account the magnetic anisotropy observed by nmr in the normal state @xcite . in the following we will show that spin - orbit coupling is an important interaction in the physics of sr@xmath0ruo@xmath1 and we shall argue that it leads to the formation of the horizontal line nodes in the superconducting state . so far we investigated the three - band hubbard hamiltonian . this model neglects spin - orbit coupling and would fail to explain the magnetic anisotropy at low temperatures observed in nmr experiments @xcite . recently it has been proposed that the spin - orbit coupling may play an important role in the superconducting state of sr@xmath0ruo@xmath1 . in particular , it was shown by ng and sigrist @xcite that the spin - orbit coupling lowers the free - energy of the chiral superconducting order parameter . another indication of the importance of a spin - orbit coupling is the recent observation of the large spin - orbit coupling in the related insulating compound ca@xmath0ruo@xmath1 @xcite . therefore , we extend the theory by adding to the hubbard hamiltonian the spin orbit coupling : @xmath189 here , the effective angular momentum l@xmath168 operates on the three @xmath32-orbitals on the site @xmath84 and @xmath190 are the electron spins . as in an earlier approach of ref . @xcite we restrict ourselves to the three orbitals , ignoring @xmath191-orbitals and choose the coupling constant @xmath192 such that the t@xmath193-states behave like an @xmath194 angular momentum representation . moreover , it is known that the quasi - two - dimensional @xmath37-band is separated from the quasi - one - dimensional @xmath35- and @xmath36-bands . then , one expects that the effect of spin - orbit coupling is relatively small for the @xmath37-band and can be neglected in the first approximation @xcite . therefore , we consider the effect of the spin - orbit coupling on @xmath35- and @xmath36-bands only . then , the kinetic part of the hamiltonian @xmath195 can be diagonalized and the new energy dispersions are obtained @xcite . note , we use @xmath196mev in accordance with earlier estimations @xcite . most importantly , the spin - orbit coupling does not break the time - reversal symmetry and therefore the kramers degeneracy between the spin @xmath197 and @xmath198 is not removed . the resultant fermi surface consists of three sheets as observed in experiment @xcite . then spin - orbit coupling together with the hubbard hamiltonian leads to a new quasiparticle which we label with pseudo - spin and pseudo - orbital indices . note , that despite the spin - orbit coupling causing the spin and orbit quantum numbers not to be good ones we can still identify the cooper - pairing to be triplet or singlet . this refers now then to the pseudo - spin quantum numbers . at the same time , the magnetic behavior of sr@xmath0ruo@xmath1 becomes very anisotropic due to the fact that both one - particle green s functions and land @xmath4-factors will be different if the magnetic field is applied along the @xmath52-axis or in the @xmath53-ruo@xmath0 plane . the anisotropy arises mainly from the calculations of the land @xmath4-factors and , in particular , their orbital parts . the factors @xmath199 and @xmath200 are calculated using the new quasiparticle states . the latter consist , for example , of @xmath35- and @xmath36-orbitals which in turn are the combinations of the initial orbital states @xmath201 and @xmath202 mixed due to the crystal field . then , the matrix elements @xmath203 , @xmath204 are zero for the @xmath35- and @xmath36-orbitals while the @xmath205 matrix element is not . therefore , the longitudinal components of the spin susceptibility of the @xmath35- and @xmath36-band get enhanced in comparison to the transverse one . one of the interesting questions that we will analyze later is the effect of spin - orbit coupling on the antiferromagnetic and ferromagnetic fluctuations . this provides insights into a microscopic explanation of the pairing mechanism and allows us to calculate the spatial structure of the superconducting order parameter . for the calculation of the transverse , @xmath206 , and longitudinal , @xmath207 , components of the spin susceptibility of and pseudo - orbital @xmath80 quantum numbers . the lande s @xmath4-factors are denoted by @xmath208 ( @xmath209 ) and @xmath210.,scaledwidth=50.0% ] each band @xmath80 we use the diagrammatic representation shown in fig . [ chisrdiagram ] . note that because the kramers degeneracy is not removed by spin - orbit coupling @xmath211 . thus , as we have already mentioned before , the anisotropy arises mainly from the calculations of the lande s @xmath4-factors and in particular their orbital parts . the factors @xmath199 and @xmath200 are calculated using the new quasiparticle states . the latter consist , for example , of @xmath35- and @xmath36-orbitals which in turn are the combinations of the initial orbital states @xmath201 and @xmath202 mixed due to the crystal field . then , the same matrix elements as above are zero for the @xmath35- and @xmath36-orbitals while @xmath205 matrix element is again not . therefore , the longitudinal components of the spin susceptibility of the @xmath35- and @xmath36-band get enhanced in comparison to the transverse one . we obtain , for example , for the @xmath212-states for the transverse susceptibility @xmath213 and for the longitudinal susceptibility @xmath214 ^ 2 \nonumber \\ & \times & \frac{f(\epsilon_{{\bf k}xz}^{+})- f(\epsilon_{{\bf k+q}xz}^{+ } ) } { \epsilon_{{\bf k}xz}^{+ } - \epsilon_{{\bf k+q}xz}^{+ } + \omega + io^+ } \quad . \label{lindzz } \end{aligned}\ ] ] here , @xmath215 is again the fermi function and @xmath216 and @xmath217 are the corresponding coherence factors @xcite for all other orbitals the calculations are similar and straightforward . then , one gets within rpa the following expressions for the transverse susceptibility @xmath218 and for the longitudinal susceptibility @xmath219 where @xmath220 . these susceptibilities are used in the corresponding pairing interaction for triplet pairing . in order to compare our results with nmr and ins experiments we take @xmath221 and @xmath222 using the random phase approximation(rpa ) for each particular band we calculate the longitudinal and transverse components of the total susceptibility in the ruo@xmath0-plane . its real parts at @xmath223 are shown in fig . [ ruthimit ] . ( @xmath80 refers to the band indices ) in the normal state of sr@xmath0ruo@xmath1 along the route @xmath81 in the first bz . due to spin - orbit interaction the transverse component ( @xmath224 ) doe not contain incommensurate antiferromagnetic fluctuations ( iaf ) from @xmath35 and @xmath36-orbitals , while in the longitudinal component ( @xmath225 ) they are enhanced . , scaledwidth=90.0% ] as a result of the spin - orbit coupling the magnetic response becomes very anisotropic along the whole brillouin zone . since the spin and orbital degrees of freedom are mixed now , the orbital anisotropy will be reflected in the magnetic susceptibility . as one sees in fig . [ ruthimit ] the longitudinal component has a pronounced peak at q@xmath168 , while the transverse one does not show these features at all and is almost isotropic ( it reproduces mostly the response of the @xmath37-band ) . in order to understand why the longitudinal susceptibility shows mostly the nesting features of the @xmath35- and @xmath36-bands one has to remember that due to spin - orbit coupling the orbital component of the magnetic susceptibility can not be neglected . therefore , the matrix elements such as @xmath226 and @xmath227 have to be taken into account . at the same time the @xmath35- and @xmath36-bands consist of the @xmath201 and @xmath202 orbital states . one sees that while the longitudinal component gets an extra term due to @xmath228 , the transverse component does not , since the transitions between states @xmath201 and @xmath202 are not allowed . thus , the contribution of the nesting of the @xmath35 and @xmath36 orbitals to the longitudinal component of the susceptibility is larger than to the transverse one and incommensurate antiferromagnetic fluctuations are almost absent . and summed over q. note , @xmath229 and @xmath230 refer to the out - of - plane ( solid curve ) and in - plane ( dashed curve ) components of the rpa spin susceptibility . , scaledwidth=90.0% ] of @xmath22o in the ruo@xmath0-plane for an external magnetic field applied along @xmath52-axis ( dashed curve ) and along the @xmath53-plane ( solid curve ) using eq . triangles are experimental results taken from ref . , scaledwidth=85.0% ] in order to demonstrate the temperature dependence of the magnetic anisotropy induced by the spin - orbit coupling we display in fig . [ rutimhi ] the temperature dependence of the quantity @xmath231 for both components . at room temperature both longitudinal and transverse susceptibilities are almost identical , since thermal effects wash out the influence of the spin - orbit interaction . with decreasing temperature the magnetic anisotropy increases and at low temperatures we find the important result that the out - of - plane component @xmath225 is about two times larger than the in - plane one : @xmath232 we also note that our results are in accordance with earlier estimations made by ng and sigrist @xcite however , there is one important difference . in their work it was found that the iaf are slightly enhanced in the longitudinal components of the @xmath35- and @xmath36-bands in comparison to the transverse one . in our case we have found that the longitudinal component of the magnetic susceptibility is strongly enhanced due to orbital contributions . moreover , we have shown by taking into account the correlation effects within rpa that the iaf are further enhanced in the @xmath40-direction . finally , in order to compare our results with experimental data we calculate the nuclear spin - lattice relaxation rate @xmath233 for the @xmath22o ion in the ruo@xmath0-plane for different external magnetic field orientation @xmath234 : @xmath235_i & = & \frac{2k_b \gamma^2_n}{(\gamma_e \hbar)^2 } \sum_{\bf q } \|a_{\bf q}^{p}\|^2 \frac{\chi''_{p } ( { \bf q } , \omega_{sf } ) } { \omega_{sf } } \quad , \label{t1 t } \end{aligned}\ ] ] where @xmath236 is the @xmath237-dependent hyperfine - coupling constant perpendicular to the @xmath84-direction . in fig . [ t1tris ] we show the calculated temperature dependence of the spin - lattice relaxation for an external magnetic field within and perpendicular to the ruo@xmath0-plane together with experimental data . at @xmath238 k the spin - lattice relaxation rate is almost isotropic . due to the anisotropy in the spin susceptibilities arising from spin - orbit coupling the relaxation rates become increasingly different for decreasing temperature . the largest anisotropy occurs close to the superconducting transition temperature in good agreement for experimental data @xcite . note , most recently similar magnetic anisotropy has been observed by means of inelastic neutron scattering @xcite . however , the absolute magnitude of the anisotropy seems to be smaller than in the case of the nmr experiments @xcite . the reason for this might be that in the nmr experiments the magnetic susceptibility is summed over the whole brillouin zone while in the ins only its part around @xmath239 is probed . what is the influence of spin - orbit coupling for determining the symmetry of the superconducting gap in sr@xmath0ruo@xmath1 ? -planes ( in real coordinate representation ) as resulting from the magnetic anisotropy . here , the amplitude of the order parameter along the @xmath40-direction has been drawn in cylindrical coordinates between ruo@xmath0-planes . this horizontal line node in the superconducting gap resulting from antiferromagnetic spin fluctuations along @xmath40-direction seems to be in agreement with thermal conductivity measurements below @xmath6 @xcite . , scaledwidth=70.0% ] first , the spin - orbit coupling affects the spin dynamics and as we have shown induces the anisotropy in the spin subspace . in particular the two - dimensional iaf at q@xmath240 have a polarization along the @xmath40-direction . this simply means that the antiferromagnetic moments associated with these fluctuations align along the @xmath40-direction . at the same time the ferromagnetic fluctuations are in the @xmath53-plane this is illustrated in fig . [ spatial ] . since the interaction in the ruo@xmath0-plane ( @xmath185=0 ) is mainly ferromagnetic nodeless @xmath5-wave pairing is indeed possible and the superconducting order parameter in all three bands will not have nodes . this differs from previous results in which only the @xmath37-band was nodeless . therefore , we safely conclude that triplet @xmath5-wave pairing without line - nodes will be realized in the ruo@xmath0-plane . on the other hand the state between the ruo@xmath0-plane will be determined mainly by the iaf that are polarized along the @xmath52-direction . this implies that the magnetic interaction between the planes has to be rather antiferromagnetic than ferromagnetic . this antiferromagnetic interaction strongly suggests a line of nodes for @xmath241 between the neighboring ruo@xmath0-planes . this is in agreement with the arguments presented in ref . @xcite . then , due to the magnetic anisotropy induced by spin - orbit coupling a nodeless @xmath242wave pairing is possible in the ruo@xmath0-plane as experimentally observed @xcite , while a node would lie between the ruo@xmath0-planes . note , the spin - orbit coupling determines also the orientation of the @xmath243-vector of the cooper - pairs . in a triplet superconductor , the pairing state can be represented by the three - dimensional vector d(k ) * , whose magnitude and direction may vary over the fermi surface in * k-space . like in the case of superfluid @xmath12he , several pairing states related to the different orientation of the total spin moment of the cooper - pair @xmath244 may have the same condensation energy under weak - coupling conditions . this degeneracy is related to the spin rotation symmetry being present in @xmath12he ( which leads to presence of the so - called b - phase ) , but is lifted by the spin - orbit coupling in sr@xmath0ruo@xmath1 @xcite . then , the analogy of the @xmath153-phase - like can be realized in sr@xmath0ruo@xmath1 ( d(k)@xmath245 = ( 0,0,@xmath246 ) ) while the two - dimensional analog of the @xmath247-phase will be suppressed ( d(k)@xmath248 = @xmath249 ( @xmath250 0 ) ) . the different orientation of the spin and orbital angular momenta of the cooper - pair is illustrated in fig . [ triplet11 ] . ruo@xmath1 taken from ref . ( a ) @xmath251 and ( b ) @xmath252 have different spin and total angular momentum . the @xmath253 superconducting state has an angular momentum along @xmath52-axis ( thick arrow ) and the spin is lying in the ruo@xmath0-plane . , scaledwidth=90.0% ] the important consequence of this order parameter is the uncompensated orbital angular momentum of the cooper - pair ( so - called chiral state ) . in analogy to @xmath12he , interesting effects can be observed by applying external magnetic field or external pressure along @xmath53-plane to sr@xmath0ruo@xmath1 . for example , in @xmath12he the transition from the @xmath247 to the @xmath153 phase is observed under pressure while in an external magnetic field the @xmath153 transition splits into @xmath254 , @xmath255 phases . in sr@xmath0ruo@xmath1 the situation is somewhat opposite . superconductivity in sr@xmath0ruo@xmath1 is already in the @xmath153-like phase and therefore , it would be interesting to see whether by applying pressure or magnetic field one could cause a transition into a @xmath247-like phase . note that recently a second phase has been observed in high magnetic field applied along @xmath146-direction ( @xmath256 ) with @xmath257 @xcite . the origin of this feature is not clear at the moment . despite of the similar crystal structures the differences in the electronic states close to the fermi level leads to the different electronic structure of cuprates and ruthenates . as a result the spin fluctuations are very different . in both systems the spin fluctuations are strong but while in cuprates the antiferromagnetic fluctuations at q=@xmath60 in sr@xmath0ruo@xmath1 there is a competition between ferromagnetic and incommensurate antiferromagnetic fluctuations at q@xmath258 . as we have shown earlier the superconductivity in the sr@xmath0ruo@xmath1 seems to be induced by mainly the ferromagnetic fluctuations and the role of iaf is in the formation of the line nodes in the passive @xmath35 and @xmath36-bands between the ruo@xmath0-planes . -planes of cuprates reflecting bonding between a cu@xmath259 and two o@xmath31 ions . ( a ) initial splitting of the bands taking into account crystal field interaction only . we consider only @xmath2-electrons ( holes ) of cu and @xmath260 and @xmath261 orbitals of the oxygen . the numbers indicate the occupations of the different levels in the undoped compound . ( b ) role of the on - site coulomb repulsion and the splitting of 3d@xmath181-orbital into a lower ( lhb ) and upper hubbard band ( uhb ) . u@xmath262 is the coulomb repulsion between electrons at copper . due to @xmath263 the cuprates are mott - hubbard insulators without doping.,title="fig:",scaledwidth=50.0% ] -planes of cuprates reflecting bonding between a cu@xmath259 and two o@xmath31 ions . ( a ) initial splitting of the bands taking into account crystal field interaction only . we consider only @xmath2-electrons ( holes ) of cu and @xmath260 and @xmath261 orbitals of the oxygen . the numbers indicate the occupations of the different levels in the undoped compound . ( b ) role of the on - site coulomb repulsion and the splitting of 3d@xmath181-orbital into a lower ( lhb ) and upper hubbard band ( uhb ) . u@xmath262 is the coulomb repulsion between electrons at copper . due to @xmath263 the cuprates are mott - hubbard insulators without doping.,title="fig:",scaledwidth=50.0% ] the comparison of the superconductivity in ruthenates and cuprates was done in ref . @xcite on the basis of the @xmath264 model . for all cuprate compounds the degeneracy between the copper 3@xmath2-orbitals is removed by the lattice structure . after some straightforward calculations it can be shown that the hybridized copper and oxygen orbitals separate . as shown in fig . [ cuintro](a ) , the state with highest ( lowest ) energy has mainly d@xmath181-wave character in the undoped cuprates and the @xmath265 copper has one missing electron ( _ i.e. _ a hole ) and gives the cu - ion a spin @xmath266 . thus , in the absence of doping , the cuprate material is well described by a model of mostly localized spin-@xmath266 states . the other orbitals are occupied and therefore can be neglected . however , the coulomb repulsion between holes in the same orbital is strong and must be included . due to the large coulomb repulsion the d@xmath181-orbitals split into two so - called lower and upper hubbard bands ( lhb and uhb ) , respectively as shown in fig . [ cuintro](b ) . then , at half - filling the system becomes an insulator . furthermore , due to a very large @xmath267 the splitting of d@xmath181-band is so large that the oxygen @xmath5-band lies between the uhb and the lhb . the charge transfer gap in cuprates @xmath268 ( @xmath269 ) is smaller than @xmath267 , and so underdoped cuprates are charge - transfer insulators . when the cuprates are doped the @xmath270 model appears as an effective low - energy hamiltonian describing the moving of a hole on the antiferromagnetic background @xcite . in contrast to the cuprates , in ruthenates the fermi level lies in the center of the @xmath271 , and @xmath58 manifold of states . in the ruthenates considering the active @xmath37-band one could also consider the moving of the electrons on the ferromagnetic background . the instability of the system can be analyzed with respect to the triplet @xmath5-wave and singlet @xmath2-wave cooper - pairing . in a weak - coupling bcs theory it was found @xcite that for @xmath5-wave pairing the van hove singularity does not contribute to the @xmath6 since in the effective density of states the corresponding electron velocity at the fermi level is cancelled by the angular dependence of the @xmath5-wave superconducting gap @xmath272 . at the same time this cancellation does not occur for @xmath183-wave cooper pairing and thus the van hove singularity enhances @xmath6 . we can also contrast the spin fluctuation pairing mechanism in the cuprates and ruthenates . in the strong - coupling eliashberg - like theory of superconductivity in the cuprates @xcite the enhancement of @xmath6 occurs due to pronounced antiferromagnetic spin fluctuations and due to a large spectral weight of the pairing interaction at high frequencies . therefore , by comparison of superconductivity in sr@xmath0ruo@xmath1 and high-@xmath6 cuprates one may conclude that the higher @xmath6 in cuprates can be explained by the stronger antiferromagnetic fluctuations and the density of states effects ( van hove singularity ) . this , however , are only quantitative arguments since a complete theory of superconductivity in cuprates is still lacking . originally , the investigations of ruthenates were concentrated on sr@xmath0ruo@xmath1 and srruo@xmath13 , since it was believed that in contrast to cuprates in ruthenates only ferromagnetic fluctuations take place . however , it was soon realized that the ruthenates reveal much more interesting behavior . in particular , the richness of the electronic , magnetic and structural phase transitions in ca@xmath9sr@xmath7ruo@xmath1 has received a lot of attention recently . in particular , in ca@xmath9sr@xmath7ruo@xmath1 a transition from triplet superconductivity in sr@xmath0ruo@xmath1 to an antiferromagnetic mott - hubbard insulator in ca@xmath0ruo@xmath1 has been found @xcite . for various values of @xmath38 the different phase transitions occur as shown in fig . [ rutendoping ] . for example for @xmath273 this material is an insulator and for the interval @xmath274 ca@xmath9sr@xmath7ruo@xmath1 is a metal with short - range antiferromagnetic order . at @xmath275 there is a crossover which is accompanied by the sharp enhancement of the ferromagnetic fluctuations in the uniform spin susceptibility . for @xmath276 the system becomes superconducting with triplet cooper - pairing . p refers to paramagnetic , caf to canted antiferromagnetic , m to magnetic , sc to the superconducting phase , -m to the metallic phase , and -i to the insulating phase @xcite . , scaledwidth=50.0% ] a neutron diffraction study on ca@xmath9sr@xmath7ruo@xmath1 has been made recently @xcite , and results in the structural phase diagram of this system , shown in fig . [ friedtfig](a ) . here , one can see also the experimental data of the structural ( @xmath277 ) , magnetic ( @xmath278 ) phase transition and the dashed curve refers to the temperature @xmath279 where the magnetic susceptibility shows a maximum . as one sees for @xmath273 the metal - insulator transition is accompanied by the structural transition as well where both low and high temperature phases have the sr@xmath7ruo@xmath1 as measured by means of neutron diffraction @xcite including the different structural and magnetic phases and the occurrence of the maxima in the magnetic spin susceptibility . in the lower part the tilting and rotation distortion of the octahedra are shown together with elongation of the basal planes ( ru@xmath280small points , o(1)@xmath280larger points ) . all phases are metallic except s@xmath280pbca . , scaledwidth=50.0% ] symmetry @xmath281 with only difference of the parameter @xmath52 ( s - short @xmath52 , l - long @xmath52 ) . in the metallic phase there are two structural phase transitions which are influencing the temperature and concentration dependencies of the magnetic susceptibility . in fig . [ friedtfig](b ) we illustrate the evolution of the structural transitions with increasing x , starting from sr@xmath0ruo@xmath1 . the rotation of the ruo@xmath30 octahedra around @xmath282axis leads to a phase @xmath283 . for @xmath284 we have @xmath285 . below @xmath286 there are addition rotations around the @xmath53-plane axis which increase the @xmath287 angle and suppression of the octahedra along the @xmath52-axis . an attempt to explain the magnetic phase diagram of sr@xmath9ca@xmath7ruo@xmath1 has been made on a basis of ab - initio band structure calculations @xcite , as shown in fig . [ terakura ] . it was shown that the rotations of the ruo@xmath30 octahedra around the @xmath282axis stabilize the ferromagnetic state , since they are sufficient to reduce the @xmath288 hybridization between the @xmath37-orbital and the oxygen @xmath289-states . then the @xmath37-band becomes narrower and the van hove singularity shifts towards the fermi level . at the same time the @xmath35- and @xmath36-bands are only slightly affected . on the contrary , the rotation around the @xmath53-plane changes all three bands completely and that increases the nesting of @xmath35- and @xmath36-bands and may stabilize the antiferromagnetic phase @xcite . ruo@xmath1 including structural distortions , taken from ref . the solid lines are calculated phase boundaries , while the triangles linked by dashed lines correspond to the experimental data . , scaledwidth=40.0% ] in this picture it is easy to understand why the ferromagnetic fluctuation are strongly enhanced around the so - called critical doping @xmath275 . using the lda parameters for the tight - binding energy dispersion we calculated the lindhard response function of the @xmath37-band for ca@xmath290sr@xmath291ruo@xmath1 and sr@xmath0ruo@xmath1 as shown in fig . [ chica ] . one could clearly see that due to vicinity to the van hove singularity the ferromagnetic response of the along the symmetry points in the bz in comparison for the @xmath43-band for ca@xmath290sr@xmath291ruo@xmath1 and sr@xmath0ruo@xmath1 . due to vicinity to the van hove singularity the response of the @xmath43-band is strongly enhanced and ferromagnetic ( around @xmath72 ) . this is in good agreement with experiment @xcite . , scaledwidth=50.0% ] @xmath292-band is strongly enhanced , at @xmath72 . furthermore , since the @xmath41 and @xmath42-bands are almost unchanged the ferromagnetic response gets much stronger than the antiferromagnetic ones indicating the dominance of the ferromagnetic fluctuations around this critical point @xmath275 . this is in good agreement with experiment @xcite . however , this model can not explain the mott - hubbard transition in ca@xmath0ruo@xmath1 at finite temperatures , since ca@xmath0ruo@xmath1 remains an insulator even above @xmath278 . therefore , strong electronic correlations has to be taken into account , as was proposed recently @xcite . here it was suggested that while the @xmath35- and @xmath36-bands are split into the lower and upper hubbard bands , the @xmath37-band is not split even though it is close to the mott - hubbard transition . the splitting between @xmath35(@xmath41)- and @xmath36(@xmath42)-bands is increased due to interband coulomb repulsion . indeed , in a mean - field approximation @xmath293 the energy dispersions are renormalized @xmath294 and the splitting of the @xmath41- and @xmath42-bands is given by @xmath295 here we use the data for the filling of the @xmath41 and @xmath42-bands from ref . @xcite . at @xmath296 ev the splitting is @xmath297 ev which is comparable with the band width of these bands . note in this case one will have still three fermi surfaces ( as in lda calculations ) due to the @xmath37-band , upper hubbard band of @xmath42-band and one hole - like fermi surface from the @xmath41-band with filling factor @xmath298 . due to the rotation of ruo@xmath30-octahedra around the @xmath52-axis the width of the @xmath299-band is reduced and the system approaches the critical value @xmath300 . as we mentioned for the rotation of the ruo@xmath30-octahedra around @xmath52-axis , the @xmath41 and @xmath42-bands are not changing in this doping range ( @xmath301 ) . further decreasing @xmath38 and the rotation of the ruo@xmath30-octahedra with respect to the @xmath53-plane reduces the bandwidth of all three bands so for @xmath302 the @xmath43-band is split into the filled lower and empty upper hubbard bands . a similar situation occurs for the @xmath42-band while the @xmath41-band is now just completely filled . this allows us to explain the formation of the mott - hubbard insulator for ca@xmath0ruo@xmath1 in a way consistent with experimental data of x - ray absorption spectroscopy @xcite . note that while lda+dmft results @xcite also predict correctly the formation of mott - hubbard antiferromagnetic insulator in ca@xmath0ruo@xmath1 the resulting filling of the bands is inconsistent with the observed data @xcite . in addition , a new phase transition in ca@xmath0ruo@xmath1 has been found recently by applying an external hydrostatic pressure @xcite . above a very small pressure ( @xmath303 gpa ) ca@xmath0ruo@xmath1 shows metallic properties , and most unusually it becomes a ferromagnetic metal with a curie temperature @xmath6 strongly dependent on pressure . for example , just above the critical pressure @xmath304 k and it reaches its maximum value of @xmath305 k at 5 gpa . in general , in the hubbard model one would expect a transition from the antiferromagnetic insulator to the paramagnetic metal . however , if one keeps in mind the multi - band nature of the metallic state , a ferromagnetic ordering is possible . of course , a study of the mechanism of magnetic order in the multiband metallic state in ca@xmath0ruo@xmath1 should be done in the future . it is known that unconventional superconductors like heavy - fermion compounds or high-@xmath6 cuprates reveal peculiar behavior if one add magnetic or non - magnetic impurities . in contrast to conventional @xmath306-wave superconductors the non - magnetic as well as the magnetic impurities act as strong pair breakers and suppress the transition temperature @xmath6 of unconventional superconductors . the suppression of @xmath6 reflects the sensitivity to translational symmetry breaking and is characteristic of anisotropic ru@xmath307ti@xmath7o@xmath1 @xcite . the superconductivity is completely suppressed for @xmath308 and it becomes magnetically ordered with q@xmath309 indicating an importance of incommensurate antiferromagnetic fluctuations of the @xmath35- and @xmath36-bands for the magnetically - ordered state . , scaledwidth=65.0% ] cooper - pairing . in cuprates the effects induced by magnetic and non - magnetic impurities are not well understood . for example by doping the non - magnetic zn the local magnetic moments in cuo@xmath0-plane are induced around these impurities @xcite which demonstrate kondo - like behavior . on the contrary , the magnetic ni impurities also show quite puzzling features which so far have not been understood @xcite . the complication of the cuprates are the highly correlated electrons ( holes ) of the cuo@xmath0-plane . in this respect it is of high interest to analyze the effect of impurities in sr@xmath0ruo@xmath1 , since it is a fermi - liquid system in the normal state . early studies of the impurity effects in sr@xmath0ruo@xmath1 revealed some features reflecting the unconventional nature of superconductivity . for example , the rapid suppression of @xmath6 by native impurities and defects occurs @xcite as well as the large enhancement of the residual density of states in the superconducting state seen in the specific heat @xcite and nmr measurements @xcite . quite recently the first accurate study of the effects induced by the magnetic ( ir@xmath28 ) and non - magnetic ( ti@xmath28 ) impurity substitution in the ruo@xmath0-plane has been performed @xcite . here , the observed effects are also quite peculiar . in particular , the substitution of the _ non - magnetic _ impurity ti@xmath28 ( 3@xmath310 ) in sr@xmath0ruo@xmath1 induces a local magnetic moment with an effective moment @xmath311ti @xcite . the induced moment has ising anisotropy with an easy axis along the @xmath52-direction . the corresponding phase diagram is shown in fig . [ tiphase ] . furthermore , magnetic ordering with glassy behavior appears for @xmath38(ti)@xmath312 in sr@xmath0ru@xmath307ti@xmath7o@xmath1 while retaining the metallic conduction along the in - plane direction . when @xmath313 is further increased to @xmath314 the elastic neutron scattering measurements detect an incommensurate bragg peak whose wave vector q@xmath315 is close to the position of the inelastic neutron scattering peak in pure sr@xmath0ruo@xmath1 @xcite . most interestingly , in the vicinity of a magnetic ordering a deviation from a pure fermi - liquid behavior seen in sr@xmath0ruo@xmath1 is observed by means of resistivity and transport measurements showing linear and logarithmic temperature dependencies , respectively @xcite . these results indicate that the two - dimensional incommensurate antiferromagnetic spin fluctuations arising from the nesting of @xmath35- and @xmath36-bands become a static spin density wave state ( sdw ) by introducing ti substitution . as a function of the impurity concentration @xmath38 ( ti , ir ) @xcite . the superconductivity is completely suppressed for @xmath316 . the dashed curve shows the fit to the abrikosov - gorkov pair - breaking function generalized to the case of magnetic and non - magnetic impurities in unconventional superconductors @xcite . , scaledwidth=70.0% ] on the other hand , the system sr@xmath0ru@xmath307ir@xmath7o@xmath1 in which the substitution is magnetic ir@xmath28 ( 5@xmath317 in the low spin configuration ) shows a weak ferromagnetism for @xmath318 @xcite . thus , substitution of magnetic and non - magnetic impurities in sr@xmath0ruo@xmath1 leads to a different ground states . despite of these differences both magnetic and non - magnetic impurities act similarly suppressing superconductivity in sr@xmath0ruo@xmath1 for @xmath316 @xcite as shown in fig . [ kiki1 ] . from this behavior it is clear that magnetic and non - magnetic impurities act mainly as a potential scatters for superconductivity in sr@xmath0ruo@xmath1 and that the magnetic scattering does not play any role . although this behavior could be explained in other ways , this observation is also consistent with the existence of a spin triplet state . magnetic impurities break singlet cooper - pairs mainly because of exchange splitting of the single particle state . the equal spin paired state would not be subject to such an effect . more detailed theoretical investigation is still needed in order to fully understand the role played by the impurities in sr@xmath0ruo@xmath1 and illuminate the role played by magnetic fluctuations in the ruthenates . in this review we have presented some aspects of superconductivity in sr@xmath0ruo@xmath1 which have not been reviewed previously in a unified way so far . most importantly , we analyzed the role of electronic correlations and the competition between ferromagnetic and incommensurate antiferromagnetic spin fluctuations for the formation of unconventional triplet superconductivity in sr@xmath0ruo@xmath1 . we have found that the magnetic response is strongly anisotropic and is due to spin - orbit coupling . as we have shown it plays a crucial role in the formation of triplet @xmath5-wave superconductivity with line nodes between the ruo@xmath0-layers . we further argue that spin - fluctuation - mediated cooper - pairing scenario in a frame of the three - band hubbard hamiltonian can explain well the symmetry of the superconducting order parameter in sr@xmath0ruo@xmath1 . furthermore , we point out that strong electronic correlations are responsible for the rich phase diagram of ca@xmath9sr@xmath7ruo@xmath1 compound upon changing the ca content . we have also discussed the open issues in sr@xmath0ruo@xmath1 , like the influence of magnetic and non - magnetic impurities on the superconducting and normal state of sr@xmath0ruo@xmath1 . it is clear that the physics of triplet superconductivity in sr@xmath0ruo@xmath1 is still far from being completed and remains to be analyzed more in detail . 10 j. g. bednorz and k. a. mller , z. phys . b 64 * , 189 ( 1986 ) . y. maeno _ _ , nature * 372 * , 532 ( 1994 ) . a. p. mackenzie _ et al . , _ phys . lett . * 80 * , 161 ( 1998 ) . k. ishida _ et al . , _ phys . rev . b * 56 * , 505 ( 1997 ) . t. m. rice and m. sigrist , j. phys . matter * 7 * , l643 ( 1995 ) . m. sigrist , d. agterberg , a. furusaki , c. honerkamp , k. k. ng , t. m. rice , and m. e. zhitomirsky , physica c * 317 - 318 * , 134 ( 1999 ) . y. sidis _ lett . * 83 * , 3320 ( 1999 ) . k. ishida , h. mukuda , y. kitaoka , z. q. mao , h. fukuzawa , and y. maeno , phys . b * 63 * , 060507 ( 2001 ) . t. dahm , h. won , and k. maki , cond - mat/0006151 ( unpublished ) . a. p. mackenzie and y. maeno , rev . mod . phys.75 , 657 ( 2003 ) . c. noce and m. cuoco , phys . b * 59 * , 2659 ( 1999 ) . t. oguchi , phys . b * 51 * , 1385 ( 1995 ) . d. singh , phys . b52 , 1358 ( 1995 ) . a. liebsch and a. lichtenstein , phys . lett . * 84 * , 1591 ( 2000 ) . a. p. mackenzie _ et al . , _ phys . * 76 * , 3786 ( 1996 ) ; y. yoshida _ et al . _ , j. phys . soc . jpn . * 68 * , 3041 ( 1999 ) . i. i. mazin and d. singh , phys . b * 56 * , 2556 ( 1997 ) . i. i. mazin and d. singh , phys . lett . * 82 * , 4324 ( 1999 ) . et al . _ , lett . * 81 * , 3006 ( 1998 ) . j. f. annett , adv . phys . * 39 * , 83 ( 1990 ) . j. a. duffy _ et al . _ , lett . * 85 * , 5412 ( 2000 ) . g. m. luke _ et al . _ , nature * 394 * , 558 ( 1998 ) . d. j. scalapino , phys . rep . * 250 * , 329 ( 1995 ) ; n. f. berk and j. r. schrieffer , phys . rev . lett . * 17 * , 433 ( 1966 ) ; a. j. layzer and d. fay , _ proc . . low temp . andrews ( lt-11 ) , academic press , london , ( 1968 ) ; p. w. anderson and w. f. brinkmann , in _ `` the physics of liquid and solid helium '' _ , eds . k. h. bennemann and j. b. ketterson ( v. 2 , wiley - interscience , 1978 ) . h. kontani and k. ueda , phys . lett . * 80 * , 5619 ( 1998 ) . k. kuroki , m. ogata , r. arita , and h. aoki , phys . rev . b * 63 * , 060506(r ) ( 2001 ) . s. nishizaki , y. maeno , and z. mao , j. low temp . phys . * 117 * , 1581 ( 1999 ) ; j. phys . . jpn . * 69 * , 572 ( 2000 ) . k. ishida _ et al . _ , rev . lett . * 84 * , 5387 ( 2000 ) . m. tanatar _ et al . _ , b * 63 * , 064505 ( 2001 ) . i. bonalde _ rev . lett . * 85 * , 4775 ( 2000 ) . i. eremin , d. manske , c. joas , and k. h. bennemann , europhys . lett . * 57 * , 447 ( 2002 ) . m. e. zhitomirsky and t. m. rice , phys . lett . * 87 * , 057001 ( 2001 ) . annett , g. litak , b. l. gyorffy , and k. i. wysokinski , phys . b * 66 * , 134514 ( 2002 ) . k. izawa , h. takahashi , h. yamaguchi , y. matsuda , m. suzuki , t. sasaki , t. fukase , y. yoshida , r. settai , and y. onuki , phys . 86 * , 2653 ( 2001 ) . y. hasegawa and m. yakiyama , j. phys . * 72 * , 1318 ( 2003 ) . y. yanase and m. ogata , j. phys . * 72 * , 673 ( 2003 ) . s. koikegami , y. yoshida , and t. yanagisawa , phys . b * 67 * , 134517 ( 2003 ) . k. k. ng and m. sigrist , europhys . lett . * 49 * , 473 ( 2000 ) . t. mizokawa , l. h. tjeng , g. a. sawatzky , g. ghiringhelli , o. tjernberg , n. b. brookes , h. fukazawa , s. nakatsuji , and y. maeno , phys . lett . * 87 * , 077202 ( 2001 ) . i. eremin , d. manske , and k. h. bennemann , phys . b * 65 * , 220502(r ) ( 2002 ) ; i. eremin , d. manske , j .- tarento , and k. h. bennemann , j. supercond . * 15 * , 447 ( 2002 ) . k. k. ng and m. sigrist , j. phys . * 69 * , 3764 ( 2000 ) . m. braden , p. steffens , y. sidis , j. kulda , p. bourges , s. hayden , n. kikugawa , and y. maeno , cond - mat/0307662 ( unpublished ) . k. deguchi , m. tanatar , z. mao , t. ishiguro , and y. maeno , j. phys . . jpn . * 71 * , 2839 ( 2002 ) . y. maeno , t. m. rice , and m. sigrist , phys . today , no . 1 , 42 ( 2001 ) . e. v. kuzmin , s. g. ovchinnikov , and i. o. baklanov , phys b * 61 * , 15392 ( 2000 ) . f. c. zhang and t. m. rice , phys . b * 37 * , ( 1987 ) . k. h. bennemann and j. ketterson ( eds . ) , physics of superconductors ( springer - verlag , berlin , 2003 ) . s. nakatsuji and y. maeno , phys . * 84 * , 2666 ( 2000 ) ; phys . b * 62 * , 6458 ( 2000 ) . o. friedt _ et al . _ , b * 63 * , 174432 ( 2001 ) . z. fang and k. terakura , phys . b * 64 * , 020509 ( 2001 ) . s. g. ovchinnikov , in : ruthenate and rutheno - cuprate materials , lecture notes in physics ( springer , berlin , 2002 ) . t. mizokawa _ et al . _ , * 87 * , 077201 ( 2001 ) . f. nakamura , t. goko , m. ito , t. fujita , s. nakatsuji , h. fukazawa , y. maeno , p. alireza , d. forsythe , and s. r. julian , phys . b * 65 * , 220402(r ) ( 2002 ) . v. i. anisimov , i. a. nekrasov , d. e. kondakov , t. m. rice , and m. sigrist , eur . journal b * 25 * , 191 ( 2002 ) . a. v. mahajan et al . , phys . lett . * 72 * , 3100 ( 1994 ) . a. polkovnikov , s. sachdev , and m. vojta , phys . * 86 * , 296 ( 2001 ) . s. nishizaki , y. maeno , and z. q. mao , j. low temp . phys . * 117 * , 1581 ( 1999 ) . k. ishida , h. mukuda , y. kitaoka , z. q. mao , h. fukazawa , and y. maeno , phys . 84 * , 5387 ( 2000 ) . n. kikigawa , a. p. mackenzie , and y. maeno , j. phys . . jpn . * 72 * , 237 ( 2003 ) . m. minakata and y. maeno , phys . b * 63 * , 180504(r ) ( 2001 ) . m. braden _ et al . , _ phys . lett . * 88 * , 197002 ( 2002 ) . n. kikigawa and y. maeno , phys . rev . lett . * 89 * , 117001 ( 2002 ) . r. j. cava , b. batlogg , k. kiyono , and h. takagi , phys . b * 49 * 11890 ( 1994 ) . r.j . radtke , k. levin , h .- b . schttler , and m. norman , phys . b * 48 * , 653 ( 1993 ) .	we review the normal and superconducting state properties of the unconventional triplet superconductor sr@xmath0ruo@xmath1 with an emphasis on the analysis of the magnetic susceptibility and the role played by strong electronic correlations . in particular , we show that the magnetic activity arises from the itinerant electrons in the ru @xmath2-orbitals and a strong magnetic anisotropy occurs ( @xmath3 ) due to spin - orbit coupling . the latter results mainly from different values of the @xmath4-factor for the transverse and longitudinal components of the spin susceptibility ( i.e. the matrix elements differ ) . most importantly , this anisotropy and the presence of incommensurate antiferromagnetic and ferromagnetic fluctuations have strong consequences for the symmetry of the superconducting order parameter . in particular , reviewing spin fluctuation - induced cooper - pairing scenario in application to sr@xmath0ruo@xmath1 we show how @xmath5-wave cooper - pairing with line nodes between neighboring ruo@xmath0-planes may occur . we also discuss the open issues in sr@xmath0ruo@xmath1 like the influence of magnetic and non - magnetic impurities on the superconducting and normal state of sr@xmath0ruo@xmath1 . it is clear that the physics of triplet superconductivity in sr@xmath0ruo@xmath1 is still far from being understood completely and remains to be analyzed more in more detail . it is of interest to apply the theory also to superconductivity in heavy - fermion systems exhibiting spin fluctuations .
the search for supersymmetry ( susy ) is one of the main goals at present and future colliders @xcite as susy is generally accepted as one of the most promising concepts for physics beyond the standard model ( sm ) @xcite . a special feature of susy theories is the existence of the neutralinos , the spin1/2 majorana superpartners of the neutral gauge bosons and higgs bosons . in the mssm , the neutralinos are expected to be among the light supersymmetric particles that can be produced copiously at future high energy colliders @xcite . once several neutralino candidates are observed at such high energy colliders , it will be crucial to establish the majorana nature and cp properties of the neutralinos . in this light , many extensive studies of the general characteristics of the neutralinos in their production and decays @xcite as well as in the selectron pair production @xcite at @xmath4 and/or @xmath5 linear colliders have been performed . + in the present work , we analyze two body tree level decays of neutralinos into a neutralino plus a @xmath0 boson or a lightest neutral higgs boson @xmath1 in order to probe the majorana nature of the neutralinos and cp violation in the neutralino system . a comprehensive analysis of the two body decays of neutralinos as well as charginos was given previously in ref . we however note that a rather light higgs boson mass was assumed and no @xmath0 boson polarization was considered in the previous work . one powerful diagnostic tool in the present analysis is @xmath0 polarization , which can be reconstructed with great precision through @xmath0boson leptonic decays , @xmath6 , in particular , with @xmath7 . + it is possible that due to the masses of the relevant particles , no two body tree level decays are allowed , in which case the dominant decays would consist of three body tree level @xcite or two body one loop decays @xcite . however , a sufficiently heavy neutralino can decay via tree level two body channels containing a @xmath0 or @xmath1 with its mass less than 135 gev in the context of the mssm @xcite . if some sfermions are sufficiently light , two body tree level decays of neutralinos into a fermion and a sfermion may be also be important . however , neutralinos heavier than the squarks will be extremely difficult to isolate at hadron colliders , because the squarks and gluinos are strongly produced and they decay subsequently into lighter neutralinos and charginos . on the other hand , at @xmath4 colliders , squarks and sleptons , if they are kinematically accessible , are fairly easy to produce and study directly . with these phenomenological aspects in mind , we assume in the present work that all the sfermions are heavier than ( at least ) the second lightest neutralino @xmath8 . then , we investigate the mssm parameter space for the two body tree level decays of the neutralino @xmath8 and show how the majorana nature and cp properties of the neutralinos can be probed through the two body decays @xmath9 , once such two body decays are kinematically allowed . + the paper is organized as follows . section [ sec : mixing ] is devoted to a brief description of the mixing for the neutral gauginos and higgsinos in cp noninvariant theories with non vanishing phases . in sec . [ sec : two - body ] , after explaining the reconstruction of @xmath0boson polarization through the @xmath0 decays into two lepton pairs , we present the formal description of the ( polarized ) decay widths of the two body neutralino decays into a lightest neutralino @xmath10 plus a @xmath0 boson or a lightest higgs boson @xmath1 with special emphasis on the polarization of the @xmath0 boson . in sec . [ sec : analysis ] , we first investigate the region of the mssm parameter space where the two body neutralino decays are allowed and discuss the dependence of the branching ratios and decay widths on the relevant susy parameters . then , we give a simple numerical demonstration of how the majorana nature and cp properties of the neutralinos can be probed through the two body decays @xmath11 . finally , we conclude in sec . [ sec : conclusion ] . in the mssm , the mass matrix of the spin-1/2 partners of the neutral gauge bosons , @xmath12 and @xmath13 , and of the neutral higgs bosons , @xmath14 and @xmath15 , takes the form @xmath16 0 & m_2 & m_z c_\beta c_w & -m_z s_\beta c_w\\[1 mm ] -m_z c_\beta s_w & m_z c_\beta c_w & 0 & -\mu \\[1 mm ] m_z s_\beta s_w & -m_z s_\beta c_w & -\mu & 0 \end{array}\right)\ , , \label{eq : massmatrix}\end{aligned}\ ] ] in the @xmath17 basis . here @xmath18 and @xmath19 are the fundamental susy breaking u(1 ) and su(2 ) gaugino mass parameters , and @xmath20 is the higgsino mass parameter . as a result of electroweak symmetry breaking by the vacuum expectation values of the two neutral higgs fields @xmath21 and @xmath22 ( @xmath23 , @xmath24 where @xmath25 ) , non diagonal terms proportional to the @xmath0boson mass @xmath26 appear and the gauginos and higgsinos mix to form the four neutralino mass eigenstates @xmath27 ( @xmath28@xmath29 ) , ordered according to increasing mass . in general the mass parameters @xmath18 , @xmath19 and @xmath20 in the neutralino mass matrix ( [ eq : massmatrix ] ) can be complex . parameterization of the fields , @xmath19 can be taken real and positive , while the u(1 ) mass parameter @xmath18 is assigned the phase @xmath30 and the higgsino mass parameter @xmath20 the phase @xmath31 . for the sake of our latter discussion , it is worthwhile to note that in the limit of large @xmath32 the gaugino higgsino mixing becomes almost independent of @xmath32 and the neutralino sector itself becomes independent of the phase @xmath31 in this limit . + the neutralino mass eigenvalues @xmath33 ( @xmath28-@xmath29 ) can be chosen positive by a suitable definition of the mixing matrix @xmath34 , rotating the gauge eigenstate basis @xmath17 to the mass eigenstate basis of the majorana fields : @xmath35 . in general the mixing matrix @xmath34 involves 6 non trivial angles and 9 non trivial phases , which can be classified into three majorana phases and six dirac phases @xcite . the neutralino sector is cp conserving if @xmath20 and @xmath18 are real , which is equivalent to vanishing dirac phases ( mod @xmath36 ) and majorana phases ( mod @xmath37 ) . majorana phases of @xmath38 do not signal cp violation but merely indicate different intrinsic cp parities of the neutralino states in cp invariant theories @xcite . before describing the two body decays @xmath39 in detail , we explain how to reconstruct the @xmath0 polarization through the lepton angular distributions of the @xmath0boson leptonic decays , @xmath40 , particularly with @xmath7 . in the rest frame of the decaying @xmath0 boson , which can be reconstructed with great precision by measuring the lepton momenta , the lepton angular distributions are given by @xmath41}\ , \frac{d\gamma[z(\pm)\rightarrow l^+l^-]}{d\cos\theta_l } = \frac{3}{8}\left[1+\cos^2\theta_l\pm 2\,\xi_l \cos\theta_l\right]\ , , \nonumber\\ & & \frac{1}{\gamma[z\rightarrow l^+l^-]}\ , \frac{d\gamma[z(0)\rightarrow l^+l^-]}{d\cos\theta_l } = \frac{3}{4}\,\sin^2\theta_l\ , , \label{eq : polar_angle}\end{aligned}\ ] ] for the @xmath0boson helicities , @xmath42 and @xmath43 , respectively , where @xmath44 with @xmath45 and @xmath46 , and @xmath47 is the polar angle of the @xmath48 momentum with respect to the @xmath0 boson polarization direction . here , the decay width @xmath49 $ ] is the average of three polarized decay widths , @xmath50 = \frac{1}{3 } \left\{\ , \gamma[\ , z(+)\rightarrow l^+l^-]+\gamma[\ , z(0)\rightarrow l^+l^- ] + \gamma[\ , z(-)\rightarrow l^+l^-]\,\right\}.\end{aligned}\ ] ] we emphasize that the three polar angle distributions ( [ eq : polar_angle ] ) can be determined without knowing the full kinematics of the decay @xmath51 . in contrast , the distributions involving the interference of the amplitudes with different @xmath0 helicities are always accompanied with azimuthal angle dependent terms . as the lightest neutralino @xmath10 assumed to be the lightest susy particle ( lsp ) always escapes detection , the kinematics of the two body decay @xmath51 can not be fully reconstructed so that the azimuthal angle dependent distributions are not fully available . + the decay width of the decay @xmath52 producing a @xmath0 boson with its helicity , @xmath42 or @xmath43 , reads @xmath53 = \frac{g^2_z\,\lambda^{1/2}_z}{16\pi m^3_i}\,(\|v\|^2+\|a\|^2 ) \left[\,m^2_i+m^2_j - m^2_z- 2m_i m_j\ , { \cal a}_n \pm \frac{\lambda^{1/2}_z}{2 } { \cal a}_t\right]\ , , \nonumber\\ & & \gamma\left[\tilde{\chi}^0_i\rightarrow \tilde{\chi}^0_j\ , z(0)\right ] = \frac{g^2_z\,\lambda^{1/2}_z}{16\pi m^3_i}\,(\|v\|^2+\|a\|^2 ) \left[\frac{\lambda_z}{m^2_z}+m^2_i+m^2_j - m^2_z - 2\,m_i m_j\,{\cal a}_n\,\right]\ , , \label{eq : polarized_decay_width}\end{aligned}\ ] ] respectively , where the asymmetries @xmath54 and @xmath55 are defined in terms of the vector and axial vector couplings @xmath56 and @xmath57 of the @xmath0 boson to the neutralino current as @xmath58 with @xmath59 and the kinematical factor @xmath60[(m_i - m_j)^2-m^2_z]$ ] . combining the leptonic @xmath0boson decay distributions ( [ eq : polar_angle ] ) with the polarized decay widths ( [ eq : polarized_decay_width ] ) , we obtain the correlated polar angle distribution : @xmath61\,\bigg\ { \left(\gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j\ , z(+ ) ] + \gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j\ , z(- ) ] \right)\ , ( 1+\cos^2\theta_l)\nonumber\\ & & { } \hskip 2.cm + 2\left(\gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j\ , z(+ ) ] -\gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j\ , z(- ) ] \right)\ , \xi_l\,\cos\theta_l \nonumber\\ & & { } \hskip 2.cm + 2\,\gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j\ , z(0 ) ] \,\sin^2\theta_l\bigg\}\ , . \label{eq : correlated_polar_angle}\end{aligned}\ ] ] consequently , each polarized decay width can be extracted from the correlated polar angle distribution by projecting out the distribution with a proper lepton polar angle distribution . + the explicit forms of the vector and axial vector couplings @xmath56 and @xmath57 in eq . ( [ eq : asymmetry ] ) are given in terms of the @xmath62 neutralino diagonalization matrix @xmath34 in the mssm by @xmath63 note that _ the vector coupling @xmath56 is pure imaginary and the axial vector coupling @xmath57 is pure real . _ this characteristic property of the @xmath0-@xmath64-@xmath65 coupling due to the majorana nature of neutralinos leads to one important relation between the polarized decay widths with the @xmath0boson helicities , @xmath42 : @xmath66 \ , = \ , \gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j z(-)]\ , , \label{eq : majorana}\end{aligned}\ ] ] which is valid even in the cp non invariant theory . this relation can be checked by measuring the forward angle asymmetry of the correlated polar angle distribution ( [ eq : correlated_polar_angle ] ) . however , because of the small analyzing power @xmath67 , it will be necessary to have sufficient large number of decay events to measure the asymmetry with good precision . in addition to the relation ( [ eq : majorana ] ) , the relative intrinsic cp parity of two neutralinos in the cp invariant theory can be determined by measuring the ratio of the longitudinal decay width to the transverse decay width , which satisfies @xmath68 } { \gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j z(+ ) ] + \gamma[\,\tilde{\chi}^0_i\rightarrow\tilde{\chi}^0_j z(- ) ] } = \frac{(m_i\mp m_j)^2}{m^2_z}\ , , \label{eq : ratio r}\end{aligned}\ ] ] for the even / odd relative intrinsic cp parity with @xmath69/@xmath70 , _ i.e. _ @xmath71 , respectively . in the cp non invariant theory , both the vector and axial couplings are in general non vanishing , leading to the value of the asymmetry @xmath54 different from @xmath42 . therefore , any precise measurements of the asymmetry @xmath54 will provide us with an important probe of cp violation in the neutralino system under the assumption that _ the neutralino masses are measured with good precision , independently of the decay modes . _ + next , we give the decay formulas into final states containing a lightest neutral higgs boson @xmath1 . the explicit form of the decay width of the decay @xmath72 is written as @xmath73 = \frac{g^2\,\lambda^{1/2}_h}{16\pi m^3_i } \left[\|s\|^2\,((m_i+m_j)^2-m^2_h ) + \|p\|^2\ , ( ( m_i - m_j)^2-m^2_h)\right]\,,\end{aligned}\ ] ] with the kinematical factor @xmath74[(m_i - m_j)^2-m^2_h]$ ] . the scalar and pseudoscalar couplings , @xmath75 and @xmath76 , of the higgs boson @xmath1 to the neutralino current are defined in terms of the mixing matrix @xmath34 as @xmath77\,,\nonumber\\ & & p = \frac{i}{2}{\im { \rm m}}\left[(n_{j2}-t_w n_{j1})(s_\alpha n_{i3}+c_\alpha n_{i4 } ) + ( i\leftrightarrow j)\right]\,,\end{aligned}\ ] ] where @xmath78 , @xmath79 and @xmath80 for the neutral higgs mixing angle @xmath81 . if the charged higgs boson mass in the mssm is very large , @xmath82 this decoupling approximation of the cosine and sine of the mixing angle @xmath81 is very good if the charged higgs mass is larger than twice the @xmath0 boson mass @xcite . for the sake of discussion , we take the decoupling limit in the present work . + in some susy scenarios , the lightest neutralino @xmath10 is the lsp and the second lightest neutralino @xmath8 among the other three neutralino states are expected to be lighter than sfermions and gluino @xcite . then , the two body decays @xmath83 or @xmath1 as well as the two body decays of the heavier neutralinos @xmath84 @xcite will constitute the major decay modes of the neutralinos , respectively , once the two body tree level decay modes are kinematically allowed . in the following numerical analysis we will ignore all other modes except for the two body tree level decays of the neutralinos . for the branching ratio calculations for the two body decays @xmath85 , we assume that all the susy parameters are real , @xmath18 is related to @xmath19 by the gaugino mass unification condition @xmath86 and the higgs boson mass @xmath87 is 115 gev . in addition , we assume that the mssm higgs system is in the decoupling regime so that the characteristics of the lightest higgs boson @xmath1 is similar to the sm higgs boson to a good approximation @xcite . + + -0.5 cm figure [ fig : bratio ] shows the regions of the decays of @xmath8 on the @xmath88 plane . in the region denoted by three body decays " , no two body modes are kinematically allowed . in the @xmath0 region " ( red colored ) , only the two body decay into a @xmath0 is allowed and in the @xmath89 region " , both the two body decays @xmath90 are allowed . we divide the @xmath89 region " into three parts , according to @xmath91\leq 10\%$ ] , @xmath92\leq 20\%$ ] and @xmath91\geq 20\%$ ] . ( here , @xmath91\equiv { \rm br}[\tilde{\chi}^0_2\rightarrow\tilde{\chi}^0_1\,z]$ ] . ) in addition , as a reference , the region excluded by the experimental bound @xcite on the lighter chargino mass @xmath93 gev is displayed by the blue hatched region . + we first note that , if @xmath94 $ \sim$}}~}2m_z$ ] , the mass difference @xmath95 is less than @xmath26 for all @xmath20 and the mass difference is very small for @xmath96 . so , as clearly shown in fig . [ fig : bratio ] the two body decay @xmath97 is allowed only when @xmath98 $ \sim$}}~}m_2 { \raisebox{-0.13cm}{~\shortstack{$<$\\[-0.07 cm ] $ \sim$}}~}2 \|\mu\|$ ] under the assumption of the gaugino mass unification condition . in addition , we find from the figure that for the two body decays the magnitude of @xmath20 is required to be larger than about 270 gev and that , once the two body higgs mode @xmath99 is open kinematically , this two body decay mode dominates in most of the @xmath89 region . the region where the decay @xmath100 is appreciable is not symmetric between positive and negative @xmath20 in the @xmath89 region . the branching ratio @xmath91 $ ] is significant only in a small area of the positive @xmath20 region , but in a large area of the negative @xmath20 region . + on the other hand , we find numerically that , for the heavier neutralinos @xmath84 , the @xmath101 region for the two body decays @xmath102 expands drastically . a large region with small @xmath103 but large @xmath19 as well as with small @xmath19 but large @xmath103 also allows for the two body decays of the heavier neutralinos , @xmath102 . only in the wedge shaped band region of the width of about 100 gev around the line satisfying the relation @xmath104 no two body decays for the heavier neutralino @xmath105 are allowed , while the heaviest neutralino can still decay into @xmath10 and @xmath89 in the ( almost ) entire parameter space , possibly except for the region excluded by the experimental lighter chargino mass bound . + consequently , for most of the parameter space of the mssm the decays of the two heavier neutralinos are dominated by two body tree level processes of which the final state consists of a @xmath0 boson or @xmath1 boson together with one of the lighter neutralinos , or a @xmath106 boson and one of the charginos . furthermore , the two body decays of the second lightest neutralino @xmath8 can be significant in a large region of the parameter space of the mssm . + -0.3 cm in addition to the branching ratios , it is also crucial to analyze the absolute size of the decay width @xmath107 $ ] . depending on the values of the relevant couplings , the two body decay widths could be smaller than the three body decay widths involving virtual sfermion exchanges , unless the sfermions are too heavy . we exhibit in fig . [ fig : width ] the dependence of the decay width @xmath108 $ ] on the higgsino mass parameter @xmath20 , assuming again that @xmath20 is real and taking @xmath109 , @xmath110 gev and @xmath111 . the decay width decreases rapidly with increasing @xmath103 . this is because the couplings of the @xmath0 boson to the neutralino current are governed by the higgsino components of the neutralinos ( see eq . ( [ eq : va ] ) ) so that the @xmath0-@xmath64-@xmath65 couplings are strongly suppressed for large @xmath103 . therefore , for large @xmath103 , some three body decays could be more dominant than the two body decays . + -0.3 cm in the previous subsection , we restrict ourselves to the cp invariant case with real parameters , as the qualitative results obtained from the cp even quantities are not expected to change so significantly even if the parameters are complex . but , the parameters @xmath18 and @xmath20 are in general complex so that it is important to check whether they indeed have complex phases or not . the existence of the complex phases in the neutralino system , which in general cause cp violation , can be established by the measurements of the ratio @xmath112 with @xmath113 defined in eq . ( [ eq : ratio r ] ) as well as the neutralino masses @xmath114 and @xmath115 . the ratio @xmath116 is @xmath117 ( @xmath118 ) in the cp invariant theory for the positive ( negative ) relative intrinsic cp parity of the neutralinos , @xmath64 and @xmath65 , taking part in the decay @xmath119 , respectively . + to explicitly show the dependence of the ratio @xmath120 on the cp phases @xmath30 and @xmath31 , we chose a specific set of real parameters @xmath121 as a simple numerical example with @xmath122 , while varying the phases @xmath30 and @xmath31 . numerically , we find that the ratio @xmath116 is insensitive to the phase @xmath31 . this is because the phase dependence is always accompanied with @xmath123 for @xmath111 , which is already small , and the higgsino components of the neutralinos are small for large @xmath103 . so , we show in fig . [ fig : etacp ] the dependence of the ratio @xmath124 ( red solid ) as well as the asymmetry @xmath125 ( blue dot dashed ) only on the phase @xmath30 for the real parameter set for one fixed value of @xmath126 . clearly , in the cp invariant case with @xmath127 or @xmath128 , the absolute magnitude of the ratio @xmath116 as well as the asymmetry @xmath125 is 1 , but it is different from 1 in the cp non invariant case . in the given real parameter set , we find that the ratio @xmath116 is quite sensitive to the phase @xmath30 near @xmath129 , while it is not so sensitive to the phase near @xmath130 and @xmath128 . + in the present work the analysis for probing the majorana nature and cp violation in the neutralino system has been carried out at the tree level . however , it will be important to include loop corrections to the two body tree level decays because , if they are small , the tree level cp violation effects might be diluted by loop induced cp violation effects originating from other sectors of the mssm . + for a large portion of the mssm parameter space , the decay of the second lightest neutralino @xmath8 as well as the heavier neutralinos @xmath84 could be dominated by two body processes in which the final state consists of a @xmath0 or a lightest higgs boson @xmath1 together with a lightest neutralino @xmath10 , assumed to be the lightest supersymmetric particle . the main conclusion of the present work is that , unless the two body decay @xmath119 is strongly suppressed , the @xmath0 polarization , which can be reconstructed through great precision via the leptonic @xmath0boson decays @xmath131 , provides us with a powerful probe of the majorana nature of the neutralinos and cp violation in the neutralino system . the work of syc was supported in part by the korea research foundation grant ( krf2002070c00022 ) and in part by kosef through chep at kyungpook national university and the work of ygk was supported by the korean federation of science and technology societies through the brain pool program . tesla technical design report , part : iii physics at an @xmath4 linear collider , _ eds . _ heuer , d. miller , f. richard and p. zerwas , desy 2001 - 011 [ hep - ph/0106315 ] ; t. abe _ [ american linear collider working group collaboration ] , `` linear collider physics resource book for snowmass 2001 . 2 : higgs and supersymmetry studies '' , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ , ed . n. graf , hep - ex/0106056 ; k. abe _ et al . _ , jlc roadmap report , presented at the acfa lc symposium , tsukuba , japan 2003 , http://lcdev.kek.jp/rmdraft/ ; g. guignard ( ed . ) , _ a 3tev @xmath4 linear collider based on clic technology _ , cern2000 - 008 . choi , j. kalinowski , g. moortgat - pick and p.m. zerwas , eur . phys . j. c * 22 * 563 ( 2001 ) [ hep - ph/0108117 ] , _ ibid . _ c * 23 * 769 ( 2002 ) [ hep - ph/0202039 ] ; j. kalinowski , acta phys . polon . * b34 * ( 2003 ) 3441 [ hep - ph/0306272 ] ; s.y . choi , hep ph/0308060 . petcov , phys . b139 * ( 1984 ) 421 ; s.m . bilenky , e.kh . khristova and n.p . * 13 * ( 1986 ) 283 ; g. moortgat - pick and h. fraas , phys . d * 59 * ( 1999 ) 015016 [ hep - ph/9708481 ] ; g. moortgat - pick , h. fraas , a. bartl and w. majerotto , eur . j. c * 9 * ( 1999 ) 521 [ erratum - ibid . c * 9 * ( 1999 ) 549 ] [ hep - ph/9903220 ] ; g. moortgat - pick , a. bartl , h. fraas and w. majerotto , eur . j. c * 18 * ( 2000 ) 379 [ hep - ph/0007222 ] ; m.m . nojiri , d. toya and t. kobayashi , phys . d * 62 * ( 2000 ) 075009 [ hep - ph/0001267 ] . bilenky , n.d . nedecheva and s.t . petcov , nucl . phys . * b247 * ( 1984 ) 61 ; s.t . petcov , phys . lett . * b178 * ( 1986 ) 57 ; n. oshimo , z. phys . * c21 * ( 1988 ) 129 ; y. kizukuri and n. oshimo , phys . lett . * b249 * ( 1990 ) 449 ; s.y . choi , h.s . song and w.y . song , phys . d * 61 * ( 2000 ) 075004 [ hep - ph/9907474 ] ; v.d . barger , t. falk , t. han , j. jiang , t. li and t. plehn , phys . d * 64 * ( 2001 ) 056007 [ hep ph/0101106 ] ; a. bartl , h. fraas , o. kittel and w. majerotto , hep ph/0308141 ; a. bartl , t. kernreiter and o. kittel , hep ph/0309340 ; s.y . choi , m. drees , b. gaissmaier and j. song , hep ph/0310284 . j. ellis , j.m . frre , j.s . hagelin , g.l . kane and s.t . petcov , phys . b * 132 * ( 1983 ) 436 ; a. bartl , h. fraas and w. majerotto , nucl . * b278 * ( 1986)1 ; g. moortgat - pick and h. fraas , eur . j. c * 25 * ( 2002 ) 189 [ hep - ph/0204333 ] . m. peskin , in _ proc . of the third workshop on physics and experiments with linear colliders _ , ed . a. miyamoto _ et al . _ , ( world scientific , singapore , 1996 ) p.248 ; int * c13 * ( 1998 ) 2299 [ hep - ph/9803279 ] ; s. thomas , int * c13 * ( 1998 ) 2307 [ hep - ph/9803420 ] ; j.l . feng and m.e . peskin , phys . d * 64 * ( 2001 ) 115002 [ hep ph/0105100 ] ; a. freitas , d.j . miller and p.m. zerwas , eur . j. * c21 * ( 2001 ) 361 [ hep ph/0106198 ] ; a. datta , a. djouadi and m. muhlleitner , eur . j. * c25 * ( 2002 ) 539 [ hep ph/0204354 ] ; c. blochinger , h. fraas , g. moortgat pick and w. porod , eur . j. * c24 * ( 2002 ) 297 [ hep ph/0201282 ] ; j.a . saavedra and a.m. teixeira , hep ph/0307001 ; a. freitas , a. von manteuffel and p.m. zerwas , hep ph/0310182 . m.m . nojiri and y. yamada , phys . d * 60 * ( 1999 ) 015006 [ hep - ph/9902201 ] ; a. bartl , w. majerotto and w. porod , phys . * b465 * ( 1999 ) 187 [ hep - ph/9907377 ] ; a. djouadi , y. mambrini and m. muhlleitner , eur . j. * c20 * ( 2001 ) 563 [ hep ph/0104115 ] . h. komatsu and j. kubo , phys . b157 * ( 1985 ) 90 ; h.e . haber and d. wyler , b323 * ( 1989 ) 267 ; s. ambrosanio and b. mele , phys . d * 53 * ( 1996 ) 2541 [ hep - ph/9508237 ] ; h. baer and t. krupovnickas , jhep * 0209 * ( 2002 ) 38 [ hep ph/0208277 ] . berger , phys . d * 41 * ( 1990 ) 225 ; y. okada , m. yamaguchi and t. yanagida , prog . * 85 * ( 1991 ) 1 ; phys . lett . * b262 * ( 1991 ) 54 ; j. ellis , g. ridolfi and f. zwirner , phys . b257 * ( 1991 ) 83 ; h.e . haber and r. hempfling , * 66 * ( 1991 ) 1815 . h.e . haber and y. nir , phys b306 * ( 1993 ) 327 ; h.e . haber , in _ physics from the planck scale to the electroweak scale _ , proc . of the us polish workshop , warsaw , poland , september 2124 , 1994 , edited by p. nath , t. taylor and s. pokorski ( world scientific , singapore , 1995 ) pp 4963 ; j.f . gunion and h.e . haber , phys . d * 67 * ( 2003 ) 075019 [ hep ph/0207010 ] .	in the minimal supersymmetric standard model ( mssm ) , the neutralinos , the spin1/2 majorana superpartners of the neutral gauge and higgs bosons , are expected to be among the light supersymmetric particles that can be produced copiously at future high energy colliders . we analyze two body neutralino decays into a neutralino plus a @xmath0 boson or a lightest neutral higgs boson @xmath1 , allowing the relevant parameters to have complex phases . we show that the two body tree level decays of neutralinos are kinematically allowed in a large region of the mssm parameter space and they can provide us with a powerful probe of the majorana nature and cp properties of the neutralinos through the @xmath0boson polarization measured from @xmath0boson leptonic decays . = 100000 kias p03079 + kupt0304 + hep - ph/0311037 + * analysis of the neutralino system in two body decays of neutralinos * + s. y. choi@xmath2 and y. g. kim@xmath3 0.5 cm 1 . @xmath2 _ department of physics , chonbuk national university , chonju 561 - 756 , korea _ + 2 . @xmath3 _ department of physics , korea university , seoul 136701 , korea _
strange stars are astronomical compact objects which are entirely made of deconfined quarks ( for a review , see glendenning 1996 ; weber 1999 ) . the possible existence of strange stars is a direct consequence of the conjecture that strange quark matter ( sqm ) may be the absolute ground state of the strong interaction , i.e. , absolutely stable with respect to @xmath1fe ( bodmer 1971 ; witten 1984 ) . sqm with a density of @xmath2 g @xmath3 might exist up to the surface of a strange star . recently , the thermal emission from bare sqm surfaces of strange stars was considered ( usov 1998 , 2001a ) . it was shown that the surface emissivity of sqm in both equilibrium photons and @xmath0 pairs created by the coulomb barrier at the sqm surface is @xmath4% of the black body surface emissivity at the surface temperature @xmath5 k. below this temperature , @xmath6 k , the sqm surface emissivity decreases rapidly with decrease of @xmath7 . at the moment of formation of a strange star the surface temperature may be as high as @xmath8 k ( e.g. , haensel , paczyski , & amsterdamski 1991 ) . since sqm at the surface of a bare strange star is bound via strong interaction rather than gravity , such a star can radiate at the luminosity greatly exceeding the eddington limit , up to @xmath9 ergs s@xmath10 at @xmath11 k. ( alcock , farhi , & olinto 1986 ; chmaj , haensel , & slomiski 1991 ; usov 1998 , 2001a ) . a young strange star cools rapidly , and within about a month after its formation the surface temperature is less than @xmath12 k ( e.g. , pizzochero 1991 ) . in this case , the thermal luminosity from the stellar surface in both equilibrium photons and @xmath0 pairs is negligibly small , @xmath13 ergs s@xmath10 . a bare strange star with such a low surface temperature may be a strong source of radiation only if its surface is reheated . recently , response of a bare strange star to accretion of a massive comet - like object with the mass @xmath14 g onto the stellar surface was considered ( usov 2001b ) . it was shown that the light curves of the two giant bursts observed from the soft @xmath15-ray repeaters sgr 0526 - 66 and sgr 1900 + 14 may be easily explained in this model . in this paper we report on our numerical simulations of the response of a bare strange star to energy input onto its surface . we consider a wide range of the rate of the energy input . both normal and colour superconducting sqm are discussed . the model to be studied is the following . the energy input onto the surface of a bare strange star starts at the moment @xmath16 , and it is spherical and steady at @xmath17 . since in our simulations the surface temperature is not higher than @xmath18 k , @xmath0 pairs created by the coulomb barrier are the main component of the thermal emission from the stellar surface ( usov 2001a ) . we assume that the process of the energy input has no effect on the outflow of both created pairs and photons which form due to annihilation of some of these pairs ( cf . usov 2001b ) . the equation of heat transfer that describes the temperature distribution at the surface layers of a strange star is @xmath19 where @xmath20 is the specific heat for sqm per unit volume , @xmath21 is the thermal conductivity , and @xmath22 is the neutrino emissivity . the heat flux due to thermal conductivity is @xmath23 at the stellar surface , @xmath24 , the heat flux directed into the strange star is equal to ( usov 2001b ) @xmath25 where @xmath26 is the rate of the energy input onto the stellar surface , @xmath27 cm is the radius of the star , @xmath28 is the energy flux in @xmath0 pairs emitted from the sqm surface , @xmath29 is the mean energy of created pairs , @xmath30 is the flux of pairs from the unit sqm surface , @xmath31 @xmath32 , and @xmath33 is the surface temperature in units of @xmath34 k. eqs . @xmath35 give a boundary condition on @xmath36 at the stellar surface . we assume that at the initial moment , @xmath16 , the temperature in the surface layers is constant , @xmath37 k. it has been suggested ( bailin & love 1979 , 1984 ) that the quarks may eventually form cooper pairs . recently , superconductivity of sqm was considered in detail ( for a review , see rajagopal , & wilczek 2000 ; alford , bowers , & rajagopal 2001 ) , and it was shown that sqm is plausibly a colour superconductor if its temperature not too high . below , we consider both normal and superconducting sqm . for non - superconducting sqm , the contribution of the quarks to both the specific heat and the thermal conductivity prevails over the contributions of the electrons , photons and gluons . in this case we have ( iwamoto 1982 ; heiselberg & pethick 1993 ; benvenuto & althaus 1996 ) @xmath38 @xmath39 @xmath40 where @xmath41 @xmath3 is normal nuclear matter density , @xmath42 is the baryon number density of sqm , @xmath43 is the qcd fine structure constant , @xmath44 is the quark - gluon coupling constant , @xmath45 is the number of electrons per baryon , and @xmath46 is the temperature in units of @xmath34 k. sqm may be a colour superconductor if its temperature is below some critical value . in the classic model of bardeen , cooper , and schrieffer the critical temperature is @xmath47 , where @xmath48 is the energy gap at zero temperature and @xmath49 is the boltzmann constant ( e.g. , carter & reddy 2000 ) . the value of @xmath50 is very uncertain and lies in the range from @xmath51 mev ( bailin & love 1984 ) to @xmath52 mev ( alford , rajagopal , & wilczek 1998 ; alford , berges , & rajagopal 1999 ; pisarski & rischke 2000 ) . we use the following interpolation formula for the temperature dependence of the energy gap(e.g . , carter & reddy 2000 ) : @xmath53^{1/2}.\ ] ] superconductivity modifies the properties of sqm significantly . the specific heat of superconducting sqm increases discontinuously as the temperature falls below the critical temperature , and then decreases exponentially at lower temperatures . for the quark specific heat at @xmath54 we adopt the theoretical results of m " uhlschlegel ( 1959 ) for superconducting nucleons , i.e. , ( horvath , benvenuto , & vucetich 1991 ; blaschke , kl " ahn , & voskresensky 2000 ) @xmath55 @xmath56\,,\ ] ] here and below tilde signifies that this value relates to superconducting sqm . even at @xmath57 the suppression of the specific heat of sqm is never complete because the electrons remain unpaired . the specific heat of superconducting sqm which is used in our simulations is @xmath58 , where @xmath59 is the specific heat of the electron subsystem of sqm ( blaschke , grigorian , & voskresensky 2001 ) . at @xmath60 both the thermal conductivity of sqm and its neutrino emissivity are suppressed by a factor of @xmath61 ) . in our simulations , we adopt @xmath62+k_e\,,\ ] ] @xmath63\,,\ ] ] where @xmath64 and @xmath22 are given by equations ( 7 ) and ( 8) , respectively , and @xmath65 is the thermal conductivity of the electrons ( e.g. , blaschke et al . the set of equations @xmath66 was solved numerically for both normal @xmath67 and superconducting @xmath68 states of sqm . we assumed the typical values of @xmath69 , @xmath70 , and @xmath71 . figure 1 shows a typical temporal behaviour of the strange star luminosity , @xmath72 , in @xmath0 pairs . from this figure , we can see that @xmath73 increases eventually to its maximum value @xmath74 . the rate of this increase may be characterized by the rise time @xmath75 that is determined as a time interval from the initial moment , @xmath16 , to the moment when @xmath73 is equal to @xmath76 . the results of our simulations are presented in tables 1 and 2 . there is a critical value , @xmath77 , of the input luminosity at which the dependence of @xmath74 on @xmath26 changes qualitatively ( see table 1 ) . @xmath77 is @xmath78 ergs s@xmath10 for normal sqm and @xmath79 ergs s@xmath10 for superconducting sqm with @xmath80 mev . at @xmath81 , @xmath74 is about @xmath26 while when @xmath26 is a few times below than @xmath77 the thermal emission of @xmath0 pairs from the stellar surface is negligible . in our simulations , the rise time @xmath75 varies in a very wide range from @xmath82 s at @xmath83 ergs s@xmath10 and @xmath84 mev to @xmath85 s at @xmath86 ergs s@xmath10 and @xmath87 mev ( see table 2 ) . at @xmath87 mev the thermal emission of @xmath0 pairs does not depend on @xmath48 because in this case both the specific heat of the quarks and their thermal conductivity are strongly suppressed , and the heat transport is mostly determined by the electron subsystem of sqm . the rise time of the luminosity in neutrinos is many orders of magnitude larger that @xmath75 , especially when @xmath75 is small . in our model the neutrino luminosity may increase up to @xmath88 when @xmath89 goes to infinity . this is because all energy which is delivered onto the stellar surface is radiated either from the surface by @xmath0 pairs or from the stellar interior by neutrinos . since bare strange stars can radiate at the luminosities greatly exceeding the eddington limit ( alcock et al . 1986 ; chmaj et al . 1991 ; usov 1998 , 2001a ) , these stars are reasonable candidates for soft @xmath15-ray repeaters ( sgrs ) that are the sources of brief ( @xmath90 s ) bursts with super - eddington luminosities , up to @xmath91 . the bursting activity of a sgr may be explained by fast heating of the bare , rather cold ( @xmath92 k ) surface of a strange star up to the temperature of @xmath93 k and its subsequent thermal emission ( usov 2001a ) . the heating mechanism may be either fast decay of superstrong ( @xmath94 g ) magnetic fields ( usov 1984 ; duncan & thompson 1992 ; paczy ' nski 1992 ; thompson & duncan 1995 ; cheng & dai 1998 ; heyl & kulkarni 1998 ) or impacts of comet - like objects onto the stellar surface ( harwitt & salpeter 1973 ; newman & cox 1980 ; zhang , xu , & qiao 2000 ; usov 2001b ) . the magnetar model of sgrs which is based on the first mechanism is most popular now . in this model , the magnetic energy of a strongly magnetized strange star ( magnetar ) may be released from time to time due to mhd instabilities . a violent release of energy inside a magnetar excites surface oscillations ( e.g. , thompson & duncan 1995 ) . in turn , this shaking may generate strong electric fields in the magnetar magnetosphere which accelerate particles to high energies . these high energy particles bombard the surface of the strange star and heat it . at the input luminosities of @xmath95 ergs s@xmath10 , which are typical for sgrs , the efficiency of reradiation of the partical energy by the stellar surface to @xmath0 pairs is very high , @xmath96 , especially for superconducting sqm ( see table 1 ) . since for sgrs the burst luminosities are at least a few orders of magnitude higher than @xmath97 , the outflowing @xmath0 pairs mostly annihilate in the vicinity of the strange star ( e.g. , beloborodov 1999 ) . therefore , at @xmath98 ergs s@xmath10 far from the star the luminosity in x - ray and @xmath15-ray photons practically coincides with @xmath99 , @xmath100 . two giant bursts were observed on 5 march 1979 and 27 august 1998 from sgr @xmath101 and sgr @xmath102 , respectively . the peak luminosities of these bursts were @xmath103 ergs s@xmath10 ( fenimore , klebesadel , & laros 1996 ; hurley et al . 1999 ) . in this case , from table 2 the rise time expected in our model is @xmath104 s that is consistent with available data on the two giant bursts ( mazets et al . 1999 ) . this is valid irrespective of that sqm is a colour superconductor or not . for typical bursts of sgrs the luminosities are @xmath105 ergs s@xmath10 ( e.g. , kouveliotou 1995 ) , and the observed rise times ( @xmath106 s ) may be explained in our model only if sqm is a superconductor with the energy gap @xmath107 mev ( see table 2 ) . alcock , c. , farhi , e. , & olinto , a. 1986 , apj , 310 , 261 alford , m. , berges , j. , & rajagopal , k. 1999 , nucl . phys . b , 558 , 219 alford , m. , bowers , j.a . , & rajagopal , k. 2001 , j. phys . g , 27 , 541 alford , rajagopal , k , & wilczek , f. 1998 , pjys . lett . b , 422 , 247 bailin , d. , & love , a. 1979 , j. phys . a , 12 , l283 bailin , d. , & love , a. 1984 , phys . , 107 , 325 beloborodov , a.m. 1999 , mnras , 305 , 181 benvenuto , o.g . , & althaus , l.g . 1996 , apj , 462 , 364 blaschke , d. , grigorian , h. , & voskresensky , d.n . 2001 , a&a , 368 , 561 blaschke , d. , kl " ahn , t. , & voskresensky , d.n . 2000 , apj , 533 , 406 bodmer , d. 1971 , phys . d. , 4 , 1601 carter , g.w . , & reddy , s. 2000 , phys . d , 62 , 103002 cheng , k.s . , & dai , z.g . 1998 , phys . , 80 , 18 chmaj , t. , haensel , p. , & slomiski , w. 1991 , nucl . phys . b , 24 , 40 duncan , r.c . , & thompson , c. 1992 , apj , 392 , l9 fenimore , e.e . , klebesadel , r.w . , & laros , j.g . 1996 , apj , 460 , 964 glendenning , n.k . 1996 , compact stars : nuclear physics , particle physics , and general relativity ( verlag new york : springer ) haensel , p. , paczyski , b. , & amsterdamski , p. 1991 , apj , 375 , 209 harwitt , m. , & salpeter , e.e . 1973 , apj , 186 , l37 heiselberg , h. , & pethick , c.j . 1993 , phys . d , 48 , 2916 heyl , j.s . & kulkarni , s.r . 1998 , apj , 506 , l61 horvath , j.e . , benvenuto , o.g . , & vucetich , h. 1991 , phys . rev . d , 44 , 3797 hurley , k. , et al . 1999 , nature , 397 , 41 iwamoto , n. 1982 , ann . , 141 , 1 kouveliotou , c. 1995 , ap&ss , 231 , 49 mazets , e.p . , et al . 1999 , lett . , 25 , 635 m " uhlschlegel , b. 1959 , z. phys . , 155 , 313 newman , m.j . , & cox , a.n . 1980 , apj , 242 , 319 paczy ' nski , b. 1992 , acta astron . , 42 , 145 pisarski , r.d . , & rischke , d.h . 2000 , phys . d , 61 , 051501 pizzochero , p.m. 1991 , phys . rev . lett . , 66 , 2425 rajagopal , k. & wilczek , f. 2000 , preprint ( hep - ph/0011333 ) thompson , c. & duncan , r.c . 1995 , mnras , 275 , 255 usov , v.v . 1984 , ap&ss , 107 , 191 usov , v.v . 1998 , phys . , 80 , 230 usov , v.v . 2001a , apj , 550 , l179 usov , v.v . 2001b , phys . lett . , 87 , 021101 weber , f. 1999 , j. phys . g : nucl . part . phys . , 25 , 195 witten , e. 1984 , phys . d , 30 , 272 zhang , b. , xu , r.x . , & qiao , g.j . 2000 , apj , 545 , l127	we study numerically the thermal emission of @xmath0 pairs from a bare strange star heated by energy input onto its surface ; heating starts at some moment , and is steady afterwards . the thermal luminosity in @xmath0 pairs increases to some constant value . the rise time and the steady thermal luminosity are evaluated . both normal and colour superconducting states of strange quark matter are considered . the results are used to test the magnetar model of soft gamma - ray repeaters where the bursting activity is explained by fast decay of superstrong magnetic fields and heating of the strange star surface . it is shown that the rise times observed in typical bursts may be explained in this model only if strange quark matter is a superconductor with an energy gap of more that 1 mev .
spintronics , which exploits electron spin rather than charge to develop a new generation of electronic devices , has emerged as an active field in condensed matters because of both the underlying fundamental physics and its potential impact on the information industry.prinz98science,wolf01science,awschalom02 one key issue in spintronics is the generation and efficient control of spin current . spin - orbit interaction of electrons exists extensively in metals and semiconductors and mix spin states . it provides an efficient way to control the coherent motion of electron spins . recently it is proposed theoretically that an electric field may generate a spin current in hole - doped semiconductors and in two - dimensional electron gases ( 2deg ) in heterostructures with spin - orbit coupling due to the spin helicity and the noncollinearity of the velocity of the single particle wave function.@xcite studies of this intrinsic spin hall effect has evolved into a subject of intense research . shen03xxx , sinitsyn03xxx , culcer03xxx , schliemann04prb , hu03xxx , rashba03prb , hulb04xxx the spin hall effect in a paramagnetic metal with magnetic impurities has also been discussed , in which a transverse spin imbalance will be generated when a charge current circulates.dyakonov71,hirsch99,zhang00prl,hu03prb we also note that the spin chirality in systems with strong spin - orbit interaction may induce a pure spin current@xcite over the past two decades , remarkable phenomena have been observed in the 2deg , most notably , the discovery of integer and fractional quantum hall effect.@xcite research in spin transports provides a good opportunity to explore spin physics in the 2deg with spin - orbit couplings . the spin - orbit coupling leads to a zero - field spin splitting , and it competes with the zeeman spin splitting when a perpendicular magnetic field is applied . the result can be detected as beating in shubnikov - de haas oscillations.@xcite very recently we have studied the spin hall effect in the 2deg with spin - orbit coupling in a strong perpendicular magnetic field , and predicted a resonant spin hall effect caused by the landau level crossing near the fermi energy.@xcite in this paper we present detailed calculations of the problem . we analyze symmetries in systems with the rashba and/or dresselhaus couplings . by using linear response theory , we calculate the spin hall conductance @xmath2 , including its magnetic field and temperature dependences for realistic parameters of ingags / ingaalgs . the non - linearity in the electric field of the spin hall current near resonance is also studied beyond the linear response theory . the resonance is a low temperature property , which shows up at a characteristic temperature of the order of the zeeman energy @xmath3 . the peak of the resonance diverges as @xmath4 ( @xmath5 : the magnetic length ) , and its weight diverges as @xmath6 at low @xmath7 and at @xmath8 . near the resonant magnetic field @xmath9 , @xmath10 . the resonance arises from the fermi level degeneracy of the zeeman - split landau levels in the presence of the spin - orbit coupling . among the two types of the spin couplings we consider , the rashba coupling reduces the zeeman splitting and is the interaction responsible for the resonance . the dresselhaus coupling further separates the zeeman splitting and suppresses the resonance . the resonant condition in the presence of both rashba and dresselhaus couplings is derived within a perturbation theory , which is accurate for small ratio of the zeeman energy to the cyclotron frequency . the paper is organized as follows . in section ii we introduce the hamiltonian of the system under consideration and analyze its symmetries . in section iii , we study the spin hall current for systems with only rashba or only dresselhaus coupling . in section iv , we consider systems with both rashba and dresselhaus couplings . by treating the couplings as small parameters , we develop a perturbation method to derive the resonance condition . the paper is concluded with a summary and discussions in section v. as an introduction , we start with the three - dimensional ( 3d ) spin - orbit interaction known for iii - v compounds such as gaas and inas , which is of the form@xcite @xmath11where @xmath12 ( @xmath13 ) are the pauli matrices for spin of electrons , @xmath14 is the momentum of the charge carrier , and @xmath15 in eq.(1 ) , the first term is the dresselhaus coupling which originates from the lack of bulk inversion symmetry,@xcite while the second term is the rashba coupling which arises from the lack of structure inversion symmetry.@xcite the effective field @xmath16 is induced by the asymmetry of the external voltage to the system . in quantum wells , by neglecting the weak interband mixing and retaining the linear contribution of @xmath14 parallel to the @xmath17 plane , the spin - orbit interaction in 3d is reduced to an effective one in 2d , @xmath18 where @xmath19 and @xmath20 , with the average taken over the lowest energy band of the quasi-2d quantum well . the rashba coupling can be modulated up to fifty percent by a gate voltage perpendicular to the plane@xcite . in some quantum wells such as gaas the two terms are usually of the same order of magnitude , while in narrow gap compounds like inas the rashba coupling dominates.@xcite experimentally the relative strength of the rashba and dresselhaus couplings can be extracted from photocurrent measurements . @xcite and @xmath21 are the charge and spin hall currents.,width=321 ] in this paper we consider a spin-1/2 particle of charge @xmath22 and effective mass @xmath23 confined by a semiconductor quantum well to a 2d @xmath17 plane of length @xmath24 and width @xmath25 . @xcite the particle is subjected to a spin - orbit interaction @xmath26 . a perpendicular magnetic field @xmath27 and an electric field @xmath28 along the y - axis are applied as shown in fig . 1 . both electron - electron interaction and impurities will be neglected in our study . the hamiltonian reads @xmath29 where @xmath30 is the lande g - factor , and @xmath31 is the bohr magneton . in @xmath32 the momentum @xmath14 is replaced by the canonical momentum , @xmath33 . we choose the landau gauge @xmath34 and consider periodic boundary condition in the @xmath35 direction , hence @xmath36 is a good quantum number . below we rewrite the hamiltonian in terms of lowering and raising operators . for each @xmath37 , we introduce the lowering operator @xmath38\]]and the corresponding raising operator @xmath39 , with the magnetic length @xmath40 . @xmath41 and @xmath42 satisfy the commutations @xmath43 = \delta _ { kk^{\prime } } $ ] , and @xmath44 = 0 $ ] . in terms of @xmath45 and @xmath46 we have @xmath47where @xmath48 is the cyclotron frequency , @xmath49 , and @xmath50 is twice of the ratio of the zeeman energy to the cyclotron frequency ( @xmath51 : free electron mass ) . @xmath52 and @xmath53 , both inversely proportional to @xmath54 are the dimensionless rashba and dresselhaus coupling respectively . the velocity operator plays an important role in study of transport properties including the spin hall conductance . the velocity operator of a single particle is @xmath55/i\hbar $ ] ( @xmath56 ) , from which we obtain @xmath57 , \\ v_{y } & = & \frac{i\hbar } { \sqrt{2}ml_{b}}\left [ a_{k}^{+}-a_{k}+i\sqrt{2}\eta _ { d}\sigma _ { y}+i\sqrt{2}\eta _ { r}\sigma _ { x}\right ] .\end{aligned}\ ] ] comparing this with the standard expression of velocity for a charged particle in a magnetic field , @xmath58 the spin - orbit coupling effectively induces a spin - dependent vector potential . we analyze three symmetries of the hamiltonian in this subsection , which we will use in our calculations . _ interchange symmetry of the two couplings . _ under the unitary transformation , @xmath59 @xmath60 @xmath61 the rashba and dresselhaus couplings are interchanged@xcite , @xmath62 therefore a system with rashba coupling @xmath63 , dresselhaus coupling @xmath64 , and lande g - factor @xmath30 is mapped on to a system with rashba coupling @xmath63 , dresselhaus coupling @xmath65 , and lande g - factor @xmath66 . in particular , a system with only dresselhaus coupling can be mapped on to a system with only rashba coupling and an opposite sign in @xmath67 . this symmetry will be used in section iii . at the symmetric point @xmath68 , @xmath26 is invariant under the transformation . @xmath69 is another symmetric point under the transformation , @xmath70 @xmath71 @xmath72 for physical parameters , we will always consider @xmath73 . _ signs of the couplings . _ under the transformation , @xmath74 @xmath75 @xmath76we have @xmath77 and @xmath78 . the eigenenergy spectrum is invariant under the simultaneous sign changes of the two couplings . the eigenenergy spectrum is even in @xmath79 if @xmath80 and is even in @xmath81 if @xmath82 . _ charge conjugation . _ under the charge conjugation transformation , @xmath83 , the magnetic moment of the carrier also changes its sign , or effectively @xmath84 in eq . this transformation is equivalent to the flip of the external magnetic field @xmath85 . therefore , a system of hole carriers has the same physical properties as the corresponding electron system except for possible directional changes in the observables . @xmath86 can be solved analytically in the systems with only rashba or only dresselhaus coupling . an analytical solution is currently not available for @xmath87 with both couplings . @xcite in the next section , we shall discuss the charge and spin hall conductance of the electron system with a pure rashba coupling . the results can be mapped easily onto the system with a pure dresselhaus coupling and to the hole system in semiconductors by using the symmetries discussed above . in this section we focus on systems with the rashba coupling only . after a brief review of the single particle solution in the absence of an electric field , we will discuss the spin hall conductance by using linear response theory in iiib , and its non - linear effect and scaling behavior near the resonance in iiic . we will use the interchange symmetry to comment on the dresshaus coupling system . the single particle problem of @xmath86 with @xmath80 can be solved rashba60 . the rashba coupling hybridizes a spin down state in the @xmath88 landau level with a spin - up state in the @xmath89 landau level , and the eigenenergies are given by @xmath90 with @xmath91 for positive integer @xmath92 , and @xmath93 @xmath94 . there is a large degeneracy @xmath95 to each eigenergy . the corresponding eigenstates are given by @xmath96 where @xmath97 is the eigenstate of the n@xmath98 landau level with @xmath99 in the absence of the spin - orbit coupling . @xmath100 and @xmath101 for @xmath102 , with @xmath103 . the eigenenergies for the system with dresselhaus coupling only can be obtained by replacing @xmath79 by @xmath81 and @xmath104 by @xmath105 , @xmath106 the energy spectra versus @xmath79 or @xmath81 are plotted in fig . 2 . in the absence of the spin - orbit coupling , the zeeman energy splits the two degenerate @xmath107 landau levels of spin - up and spin - down electron states into two nearby ones with the lower level for spin - up and the higher level for spin - down . as @xmath79 increases from zero , the energy of the @xmath88 landau level state of spin - down is lowered because of its hybridization with the spin - up state at the @xmath89 landau level due to the rashba coupling . the rashba interaction competes with the zeeman energy and there is an energy crossing at certain values of @xmath79 or the magnetic fields as we can see in fig . the spin hall resonance we examine is closely related to this level crossing . the energy level diagram in fig . 2(b ) for the dresselhaus coupling has different features . in that case , a spin - up state , which is at the lower level due to the zeeman splitting , mixes with a spin - down state at a higher landau level , which separate further the zeeman splitting , thus there is no resonance in the spin hall current . as a function of the dimensionless rashba coupling @xmath108 . the parameters are @xmath109 evm , @xmath110m@xmath111 , @xmath112 and @xmath113 , taken from reference @xcite for the inversion heterostructures in @xmath114ga@xmath115as / in@xmath116al@xmath117as . ( b ) same as in ( a ) , but for the dresselhaus coupling @xmath118 . , width=321 ] we consider the charge and spin hall currents along the @xmath119axis induced by an electric field along the @xmath120axis as in fig . 1 . in terms of the velocity operator , the charge and spin-@xmath121 component current operators are defined by @xmath122respectively . we refer readers to ref . @xcite for the discussions on the other spin components . the symmeterized form of the spin current operator guarantees that it is hermitian . each single particle state @xmath123 carries a current @xmath124 . the average current density carried by the 2deg is then given by @xmath125where @xmath126 is the fermi - dirac distribution function . note that since spin is not a conserved quantity in the presence of spin - orbit couplings , the spin current defined above and the spin density do not satisfy a continuity equation . nevertheless , the expectation values of the spin density and the spin current are well defined . the charge and spin hall conductance are then given by @xmath127unlike a free electron in an uniform magnetic field , the single particle problem with the spin - orbit coupling in the presence of an electric field @xmath128 is not analytically solvable , since the landau levels mixing no longer truncates . after a replacement of @xmath129 in the operator @xmath130 by @xmath131 the hamiltonian of the system in the presence of the electric field reads , apart from a constant , @xmath132where @xmath133 is the one in eq . ( [ hh ] ) by replacing @xmath130 by @xmath134 and @xmath135 . we now consider @xmath136 as a perturbative hamiltonian to study the charge and spin hall currents . up to the first order in @xmath137 , we obtain @xmath138where the superscript refers to the 0@xmath98 order and 1@xmath139 order in the perturbation in @xmath136 , and @xmath140 in the above equations , the summation is over the complete and orthogonal set of eigenstates @xmath141 corresponding to the spectra @xmath142 in eq . ( 8) , and @xmath143 since the matrix elements vanish for other values of @xmath144 . the charge hall conductance is found to be independent of the spin - orbit coupling , @xmath145 , with @xmath146 being the filling factor . within the perturbation theory , the spin hall conductance @xmath2 can be divided into two parts . the part arising from the @xmath147 order in @xmath148 is found to be the product of the spin polarization @xmath149 per electron and the hall conductance @xmath150 , divided by the electron charge ( @xmath22 ) , @xmath151the expectation value of the spin polarization per electron is , @xmath152@xmath153 at @xmath154 is plotted in fig . the oscillation is due to the alternate filling by electrons of the energy levels with mainly spin - up and spin - down . a jump is visible at @xmath155 ( correspondi ng to inverse magnetic field @xmath156t@xmath157 ) because of the energy crossing . the second part in g@xmath158 arises from the first order in @xmath136 , @xmath159at @xmath154 , if the two degenerate energy levels ( crossing point in fig . 2(a ) ) are partially occupied , @xmath160 may become divergent . mathematically , the resonance is given by the condition @xmath161 for the electron filling factor @xmath162 , with @xmath92 an integer satisfying the equation @xmath163from the above condition , for a system with any @xmath164 , and @xmath73 , there is a unique resonant magnetic field @xmath9 such that the resonant condition is satisfied . by symmetry , we obtain the resonance condition for the system with a pure dresselhaus coupling , which is given by the solution for @xmath92 of the equation , @xmath165unlike the pure rashba coupling case , there is no solution for any @xmath73 in the pure dresselhaus coupling system . this is because the energy levels @xmath166 and @xmath167 with @xmath168 do not cross over , so the pairs of the crossing levels in fig . 2(b ) correspond to @xmath169 and do not contribute to the spin hall conductance . we have calculated the spin hall conductance numerically . @xmath160 at @xmath170 is shown in fig . in addition to the oscillation in @xmath171 similar to that of @xmath172 , there is a pronounced resonant peak at the filling @xmath155 . in fig . 4 , we show @xmath160 at several temperatures . the height of the resonance peak increases drastically as the temperature decreases below a few kelvin . in fig . 5 , we show the @xmath7-dependence of the height of the resonant peak and the two nearby side peaks . the characteristic temperature for the occurrence of the peak can be estimated to be the zeeman energy @xmath3 , which is about @xmath173 at the resonant field for the parameters in fig . more explicit derivation of this will be given in the next subsection . and spin hall conductance as a function of 1/b at @xmath154 . the parameters are the same as in fig . 2 . ( a ) and ( b ) for the rashba coupling , and ( c ) and ( d ) for the dresselhaus coupling.,width=302 ] at several temeratures for rashba coupling systems . the parameters are the same as those in fig . 2.,width=302 ] in this section we study the non - linear effect of the electric field to the resonant spin hall current and the scaling behavior . since the resonance originates from the interference of two degenerate levels near the fermi energy , we will focus on those two levels to examine the problem . as an example , we shall consider in@xmath114ga@xmath174as / in @xmath116ga@xmath117 as with the parameters given in fig . 2 , in which case the resonance occurs at the filling factor @xmath155 ( see fig . 3b ) and the relevant two levels are @xmath175 and @xmath176 . the energy levels below the two levels are assumed to be fully filled , and all levels above the two to be empty . this is valid if @xmath177 @xmath178 . the hamiltonian is then , up to a constant , reduced to a @xmath179 matrix , @xmath180 where @xmath181 as we can see from the reduced hamiltonian and from fig . 6 , the electric field breaks the level degeneracy and opens an energy gap @xmath182 . denoting the two eigenstates of the reduced hamiltonian by @xmath183 , the spin hall current density is given by @xmath184where the fermi - dirac distribution @xmath185/k_{b}t\right ) + 1\}^{-1}$ ] , with @xmath186 , @xmath187 the chemical potential , and @xmath188 . the electric field and temperature dependences of the spin current @xmath21 near the resonance point is plotted in fig.7 . at low temperatures the resonant spin current approaches to a constant in a weak electric field . the filling factor at the resonance is @xmath189,width=321 ] now we analyze the scaling behavior of the spin conductance near the resonance point . for simplicity we limit our discussion to the case of @xmath190 and @xmath191 . near the resonant point , @xmath192 where @xmath193 is the zeeman energy and @xmath194 is the reduced dimensionless magnetic field . using the identity @xmath195 , \]]we obtain the singular part of the spin hall conductance to be @xmath196 .\]]where the factor @xmath197 is a slowly varying function of @xmath7 ranging from 1 at low temperatures to @xmath198 at high temperatures . at low temperatures @xmath2 is given by , @xmath199it is only a function of the reduced magnetic field and the excess part of the filling factor from @xmath200 . at the resonant magnetic field , _ i.e. _ , @xmath201 the spin hall current approaches with lowering temperature to a constant , @xmath202 as can be seen in fig . 7 . using the resonance condition in eq.([condition ] ) , @xmath203 ( @xmath204 ) and using the fact that for large @xmath92 , @xmath92 is proportional to @xmath205 the resonant magnetic field @xmath206 approximately . thus the resonant spin current is proportional to @xmath207therefore for a given filling factor , the larger the spin - orbit coupling @xmath208 is , the stronger the spin hall resonance . the resulted spin hall conductance diverges at @xmath154 as @xmath209at temperatures @xmath210 , @xmath211and the integral @xmath212this integral reflects the weight of the resonant peak of the spin hall conductance . since the method used in this subsection is beyond perturbation theory , we conclude that the resonance spin hall conductance we predict is not an artifact of the perturbation method . instead , the resonance is caused by the interference between the two degenerate energy levels at the fermi energy . in this section we briefly discuss the resonance in the spin hall conductance in systems with both rashba and dresselhaus couplings . the hamiltonian including the electric potential reads , @xmath132 with @xmath213 . in this case @xmath86 is not solvable analytically . a state @xmath214 ( in the basis of the landau levels with @xmath215 ) is coupled to @xmath216 via the rashba coupling , which is further coupled to @xmath217 due to the dresselhaus coupling . in this way , a landau level is coupled to infinite number of other landau levels , and the analytic solution is not available . the problem , however , may be approximately solved by using perturbation theory to treat @xmath218 and @xmath81 as small parameters . this is equivalent to the limit @xmath219 , since @xmath220 . for the sample we consider in the present paper given in fig . 2 , @xmath221 at the resonant field @xmath222 tesla . in the absence of the electric field , the single particle energy , up to the second order in @xmath79 and @xmath81 , is given by [ two - e ] @xmath223 note that the mixed term of @xmath224 does not appear in the perturbation to the second order . the two levels become degenerate if the following equation is satisfied , @xmath225 it follows that a necessary condition for the resonant spin hall current is @xmath226 , for @xmath191 . at @xmath227 and in the limit @xmath228 , eq . ( [ perturbation ] ) is consistent with eq . ( [ condition ] ) for the resonant condition we derived for the pure rashba system . alternatively the resonant magnetic field is @xmath229 the large number @xmath230 increases with @xmath231 for a specific density of particles . thus for a certain rashba coupling the increasing of dresselhaus coupling will decrease the resonant magnetic field @xmath232 the singular part of the spin hall conductance can be studied by examining the two level system in the presence of an electric field as we described in section iiic . at the resonant point and at low temperature , @xmath233 as the dresselhaus coupling increases from zero , the resonance is shifted to lower magnetic fields , the resonance occurs at higher landau levels with a weaker resonant strength . in summary , we have studied the spin hall effect in a two - dimensional electron system with spin - orbit couplings in a strong perpendicular magnetic field . in systems with the rashba coupling dominating over the dresselhaus coupling , there is a resonant magnetic field at which the spin hall conductance diverges at low temperature and low electric field . the physics for this resonance is the energy level crossing of the two landau levels due to the competition of the zeeman splitting and the rashba coupling . for a given system , there is a unique resonant magnetic field , at which the two landau levels become degenerate at the fermi energy . in this case , some physical properties may show singularity . as studied earlier , the spin polarization will change its sign as the magnetic field is varied passing through the resonant field . namely the magnetic susceptibility is divergent . the spin hall conductance is another singular response due to this level crossing . when an infinitesimally weak d.c . electric field is applied in the plane , the two degenerate landau levels are split accordingly and a finite spin hall current is induced . the resonance is macroscopic in the sense that a huge number of the states in the same landau level are involved in the process . we have calculated the temperature and electric field dependences of the resonance . the characteristic temperature for the resonant spin hall current is of order of the zeeman energy . as the temperature decreases , the height of the resonance peak diverge like @xmath234 and the weight diverges like @xmath235 . while the spin orbit coupling has a dramatic effect on the spin hall conductance , the charge hall conductance is not affected and remains quantized . the spin hall current is non - linear with the electric field at the resonant field . at low temperatures , the spin hall current rapidly rises linearly with the electric field and saturates at higher electric fields . at @xmath154 , the spin hall conductance diverges as @xmath236 at resonance . near the resonant magnetic field @xmath9 , it is @xmath237 . contrary to the rashba coupling , the dresselhaus coupling further increases the zeeman energy splitting to suppress the effect of the rashba coupling . the strength of the rashba coupling necessary to surpass the dresselhaus coupling by in order to have the resonant spin hall current was estimated by using a perturbation method treating the couplings as small parameters . this is accurate as long as the zeeman energy is much smaller than then cyclotron frequency . we have assumed no potential disorder in our theory . the effects of disorder in 2deg with rashba coupling , especially in a strong magnetic field , is not well understood at this point.@xcite nevertheless , it seems reasonable to assume that the spin - orbit coupling does not change the effects of disorder qualitatively . this is likely to be the case in the presence of a strong magnetic field , which ensures extended states in the landau levels when the disorder is not sufficiently strong as evidenced by the experimentally observed quantization of the hall conductance . we then assume that the disorder gives rise to broadening of the landau level and localization so that the extended states in a landau levels are separate in energy from those in the next one by localized states . inspection of the spin - orbit coupling shows that laughlin s gauge argument still holdslaughlin , halperin , and each landau level with its extended states completely filled contribute @xmath238 to the charge hall conductance . thus we conclude that the quantum hall conductance remains intact with the spin - orbit interaction , except at the special degeneracy point . as the fermi energy varies across this degenerate extended state , the charge hall conductance @xmath239 is expected to change by @xmath240 , instead of @xmath238 for the other extended levels . this fact can be used experimentally to determine the rashba interaction induced degeneracy discussed in this paper . w. knap , c. skierbiszewski , a. zduniak , e. litwin staszewska , d. bertho , f. kobbi , j. l. robert , g. e. pikus , f. g. pikus , s. v. iordanskii , v. mosser , k. zekentes , yu . b. lyanda geller , phys . rev . b * 53 * , 3912 ( 1996 ) . y. xiong and x. c. xie , cond - mat/0403083 ; j. inoue , g. e. w. bauer , and l. w. molenkamp , phys . b 70 , 041303 ( 2004 ) ; s. murakami , phys . b 69 , 241202 ( 2004 ) ; e. g. mishchenko , a. v. shytov , and b. i. halperin , cond - mat/0406730 ; k. nomura , j. sinova , t. jungwirth , q. niu , and a. h. macdonald , cond - mat/0407279 ; l. sheng , d. n. sheng , and c. s. ting , cond - mat/0409038 .	following a previous work [ shen , ma , xie and zhang , phys . rev . lett . 92 , 256603 ( 2004 ) ] on the resonant spin hall effect , we present detailed calculations of the spin hall conductance in two - dimensional quantum wells in a strong perpendicular magnetic field . the rashba coupling , generated by spin - orbit interaction in wells lacking bulk inversion symmetry , introduces a degeneracy of zeeman - split landau levels at certain magnetic fields . this degeneracy , if occuring at the fermi energy , will induce a resonance in the spin hall conductance below a characteristic temperature of order of the zeeman energy . at very low temperatures , the spin hall current is highly non - ohmic . the dresselhaus coupling due to the lack of structure inversion symmetry partially or completely suppresses the spin hall resonance . the condition for the resonant spin hall conductance in the presence of both rashba and dresselhaus couplings is derived using a perturbation method . in the presence of disorder , we argue that the resonant spin hall conductance occurs when the two zeeman split extended states near the fermi level becomes degenerate due to the rashba coupling and that the the quantized charge hall conductance changes by @xmath0 instead of @xmath1 as the magnetic field changes through the resonant field .
the detection of compact signals ( sources ) embedded in a background is a recurrent problem in many fields of astronomy . some common examples are the separation of individual stars in a crowded optical image , the identification of local features ( lines ) in noisy one - dimensional spectra or the detection of faint extragalactic objects in microwave frequencies . the detection , identification and removal of the extragalactic point sources ( eps ) is fundamental for the study of the cosmic microwave background radiation ( cmb ) data ( franceschini et al . 1989 , toffolatti et al . 1998 , de zotti et al . 1999 ) . in particular , the contribution of eps is expected to be very relevant at the lowest and highest frequency channels of the future esa planck mission ( mandolesi et al . 1998 , puget et al . 1998 ) . the heterogeneous nature of the eps that appear in cmb maps as well as their unknown spatial distribution make difficult to separate them from the other physical components ( cmb , galactic dust , synchrotron , etc ) by means of statistical component separation methods . techniques based on the use of linear filters , however , are well - suited for the task of detecting compact spikes on a background . several techniques based on different linear filters have been proposed in the literature : the mexican hat wavelet ( mhw , cayn et al . 2000 , vielva et al . 2001a , b , 2003 ) , the classic _ matched _ filter ( mf , tegmark and oliveira - costa 1998 ) , the adaptive top hat filter ( chiang et al 2002 ) and the scale - adaptive filter ( saf , sanz et al . 2001 ) . a certain deal of controversy has appeared about which one , if any , of the previous filters is _ optimal _ for the detection of point sources in cmb data . in order to answer that question it is necessary to consider first a more fundamental issue , the concept of _ detection _ itself . the detection process can be posed as follows : given an observation , the problem is to _ decide _ whether or not a certain signal was present at the input of the receiver . the decision is not obvious since the observation is corrupted by a random process that we call ` noise ' or ` background ' . formally , the _ decision _ is performed by choosing between two complementary hypotheses : that the observed data is originated by the background alone ( _ null hypothesis _ ) , and the hypothesis that the observation corresponds to a combination of the background and the signal . to decide , the detector should use all the available information in terms of the probabilities of both hypotheses given the data . the _ decision device _ separates the space @xmath1 of all possible observations in two disjoint subspaces , @xmath2 and @xmath3 , so that if an observation @xmath4 the null hypothesis is accepted , and if @xmath5 the null hypothesis is rejected , that is , a source is ` detected ' ( @xmath2 is called the region of acceptance ) . hence , we will call any generic decision device of this type a _ detector_. the simplest example of detector , and one that has been extensively used in astronomy , is _ thresholding _ : if the intensity of the field is above a given value ( e.g. 5@xmath6 ) , a detection of the signal is accepted , on the contrary one assumes that only background is present . thresholding has a number of advantages , among them the facts that it is straightforward and that it has a precise meaning in the case of gaussian backgrounds in the sense of controlling the probability of spurious detections . however , it does not use all the available information contained in the data to perform decisions . for example , the inclusion of spatial information ( such as the curvature ) could help to distinguish the sources from fluctuations in the background with similar scale but a different shape . a general detector that can use more information than simple thresholding is given by the neyman - pearson ( np ) decision rule : @xmath7 where @xmath8 is called the likelihood ratio , @xmath9 is the probability density function ( _ pdf _ ) associated to the null hypothesis ( i.e. there is no source ) whereas @xmath10 is the _ pdf _ corresponding to the alternative hypothesis ( i.e. there is a source ) . @xmath11 are a set of variables which are measured from the data . @xmath12 is an arbitrary constant , which defines the region of acceptance @xmath2 , and must be fixed using some criterion . for instance , one can adopt a scheme for object detection based on maxima . the procedure would consist on considering the intensity maxima of the image as candidates for compact sources and apply to each of them the np rule to decide whether they are true or spurious . for a 1d image , the ratio of probabilities would then correspond to the probability of having a maximum with a given intensity and curvature ( which are the variables @xmath11 in this case ) in the presence of background plus signal over the probability of having a maximum when only background is present . if this ratio is larger than a given value @xmath12 , the candidate is accepted as a detection , if not , it is rejected . unfortunately , in many cases the sources are very faint and this makes very difficult to detect them . in order to improve the performance of the detector , a prior processing of the image could be useful . here is where _ filtering _ enters in scene . the role of filtering is to transform the data in such a way that a detector can perform better than before filtering . once the detector is fixed , it is interesting to compare the performance of different filters , which has been rarely considered in the literature . in a recent work , barreiro et al . ( 2003 ) introduce a novel technique for the detection of sources based on the study of the number density of maxima for the case of a gaussian background in the presence or absence of a source . in order to define the region of acceptance the neyman - pearson decision rule is used with _ pdf _ s associated to the previous number densities and using the information of both the intensities @xmath13 and the curvatures @xmath14 of the peaks in a data set . in addition , @xmath12 is fixed by maximising the _ significance _ , which is the weighted difference between the probabilities of having and not having a source . in that work the performances of several filters ( saf , mf and mhw ) is compared in terms of their _ reliability _ , defined as the ratio between the number density of true detections over the number density of spurious detections . they find that , on the basis of this quantity , the choice of the optimal filter depends on the statistical properties of the background . however , the criterion chosen to fix @xmath12 based on the significance does not necessarily leads to the optimal reliability . therefore , if we are considering the reliability as the main criterion to compare filters , a different criterion for @xmath12 , based on number densities must be used . in a posterior article , vio et al . ( 2004 ) , following the previous work , adopt the same neyman - pearson decision rule , based on the _ pdf _ s of maxima of the background and background plus source , to define the region of acceptance . however , they propose to find @xmath12 by fixing the number density of spurious detections and compare the performance of the filters based on the number density of true detections . in this case , the mf outperforms the other two filters . note that in these last two works different criteria have been used to fix @xmath12 , thus leading to different results . in the present work , our goal will also be to find an optimal filter that gives a maximum number density of detections fixing a certain number density of spurious sources . in order to define the detector , we will use a decision rule based on the neyman - pearson test . we will consider some standard filters ( mf , saf and mh ) introduced in the literature as well as a new filter that we call the biparametric scale adaptive filter ( bsaf ) . in all the filters appears in a natural way the scale of the source . we will modify such a scale introducing an extra parameter . in fact , it has been shown by lpez - caniego et al . ( 2004 ) that the standard matched filter can be improved under certain conditions by filtering at a different scale from that of the source . the performance of the bsaf will be compared with the other filters . the overview of this paper is as follows . in section 2 , we introduce two useful quantities : number of maxima in a gaussian background in the absence and presence of a local source . in section 3 , we introduce the detection problem and define the region of acceptance . in section 4 , we introduce an estimator of the amplitude of the source that is proven to be unbiased and maximum efficient . in section 5 and 6 , we obtain different analytical and numerical results regarding weak point sources and scale - free background spectra and compare the performance of the new filter with others used in the literature . in section 7 , we describe the simulations performed to test some theoretical aspects and give the main results and finally , in section 8 , we summarize the conclusions and applications of this paper . appendix a is a sketch to obtain a sufficient linear detector whereas we obtain the linear unbiased and maximum efficient estimator in appendix b. let us assume a 1d background ( e. g. one - dimensional scan on the celestial sphere or time ordered data set ) represented by a gaussian random field @xmath15 with average value @xmath16 and power spectrum @xmath17 : @xmath18 , where @xmath19 is the fourier transform of @xmath15 and @xmath20 is the 1d dirac distribution . the distribution of maxima was studied by rice ( 1954 ) in a pioneering article . the expected number density of maxima per intervals @xmath21 , @xmath22 and @xmath23 is given by @xmath24 being @xmath25 the expected total number density of maxima ( i.e. number of maxima per unit interval @xmath26 ) @xmath27 @xmath28 where @xmath29 and @xmath30 represent the normalized field and curvature , respectively . @xmath31 is the moment of order @xmath32 associated to the field . @xmath33 are the coherence scale of the field and maxima , respectively . as an example , figure [ fig : fig1 ] shows the values of the ratio @xmath34 for the case @xmath35 ( a typical value for the backgrounds we are considering ) . in this case , the expected density of maxima has a peak around @xmath36 and @xmath37 , that is , most of the peaks appear at a relatively low threshold and curvature , and the density of peaks decreases quickly for extreme values of @xmath38 and @xmath14 . if the original field is linear - filtered with a circularly - symmetric filter @xmath39 , dependent on @xmath40 parameters ( @xmath41 defines a scaling whereas @xmath42 defines a translation ) @xmath43 we define the filtered field as @xmath44 then , the moment of order @xmath32 of the linearly - filtered field is @xmath45 being @xmath46 the power spectrum of the unfiltered field and @xmath47 the fourier transform of the circularly - symmetric linear filter . now , let us consider a position @xmath48 in the image where a gaussian source ( i.e. profile given by @xmath49 , where @xmath41 is the beam width ) is superimposed to the previous background . then , the expected number density of maxima per intervals @xmath21 , @xmath22 and @xmath50 , given a source of amplitude @xmath51 in such spatial interval , is given by ( barreiro et al . 2003 ) @xmath52 @xmath53 where @xmath29 and @xmath30 , @xmath54 is the normalized amplitude of the source and @xmath55 is the normalized curvature of the filtered source . the last expression can be obtained as @xmath56 note that due to the statistical homogeneity and isotropy of the background , the previous equations are independent of the position of the source . we consider that the filter is normalized such that the amplitude of the source is the same after linear filtering : @xmath57 . we want to choose between different filters based on _ detection_. to make such a decision , we will focus on the following two fundamental quantities : a ) the number of spurious sources which emerge after the filtering and detection processes and b ) the number of real sources detected . as we will see in this section , these quantities are properties of the gaussian field and source that can be calculated from equations ( [ nbackground ] ) and ( [ nsource ] ) . as we will see , the previous properties are not only related to the snr gained in the filtering process but depend on the filtered momenta up to 4th - order ( in the 1d case ) , i.e. the amplification and the normalized curvature . let us consider a local peak in the 1d data set characterised by the normalized amplitude and curvature @xmath58 . let @xmath59 : n.d.f . @xmath60 represents the _ null _ hypothesis , i.e. the local number density of background maxima , and @xmath61 : n.d.f . @xmath62 represents the _ alternative _ hypothesis , i.e. the local number density of maxima when there is a compact source : @xmath63 in the previous equation , we have introduced a priori information about the probability distribution of the sources : we get the number density of source detections weighting with the a priori probability @xmath64 . to construct our detector , we will assume a neyman - pearson ( np ) decision rule using number densities instead of probabilities : @xmath65 where @xmath12 is a constant . the previous equation defines a region in @xmath66 , the so - called region of acceptance @xmath2 . therefore , the decision rule is expressed such that if the values of @xmath66 of the candidate maximum is inside @xmath2 ( i.e. @xmath67 ) we decide that the signal is present . on the contrary , if @xmath68 we decide that the signal is absent . ) is equivalent to the one defined by the usual neyman - pearson test in terms of probabilities @xmath69 where @xmath70 , @xmath71 are the _ pdf _ s associated to the number densities given by equations ( [ nbackground ] ) and ( [ eq : number_b+s ] ) and , in order to compare different filters , the constant @xmath72 must be found by fixing the number density of spurious sources in the region of acceptance instead of the _ false alarm _ probability . ] it can be proved that the previous region of acceptance @xmath2 is equivalent to the sufficient linear detector ( see appendix a ) @xmath73 where @xmath74 is a constant and @xmath75 is given by @xmath76 we remark that the assumed criterion for detection leads to a _ linear _ detector @xmath75 ( i.e. linear dependence on the threshold @xmath38 and curvature @xmath14 ) . moreover , this detector is independent of the _ pdf _ of the source amplitudes . using this detector , the expected number density of spurious sources and of true detections are given by @xmath77 @xmath78 we remark that in order to get the true number of real source detections such a number must be multiplied by the probability to have a source in a pixel in the data set . note that for a fixed number density of spurious sources @xmath79 , the np detector leads to the maximum number density of true detections @xmath80 . taking into account equations ( [ eq : r _ * ] ) to ( [ eq : ndet ] ) , one can find @xmath79 and @xmath80 for a gaussian background . after a straightforward calculation , the number density of spurious sources found using the np rule is given by : @xmath81 , \label{eq : nb}\end{aligned}\ ] ] @xmath82 similarly , the number density of detections is obtained as : @xmath83e^{- \frac{(1 - \rho^2)\varphi^2}{2{(1 - \rho y_s)}^2 } } , \label{eq : nb}\end{aligned}\ ] ] where @xmath84 @xmath85 @xmath86 the signal has an unknown parameter , the amplitude @xmath51 , that has to be estimated from the data @xmath87 . we shall assume that the most probable value of the distribution @xmath88 gives an estimation of the amplitude of the source ( criterion for amplitude estimation ) . the result @xmath89 is given by the equation @xmath90 where the function @xmath75 is given by equation ( [ eq : phi ] ) . one can prove that the previous expression corresponds to a linear estimator that is unbiased and maximum efficient ( minimum variance ) , i.e. @xmath91 where @xmath92 denotes average value over realizations ( see appendix b ) . we will consider as an application the detection of compact sources characterised by a gaussian profile @xmath93 , and fourier transform @xmath94 , though the extension to other profiles will be considered in the future . such a profile is physically and astronomically interesting because it represents the convolution of a point source ( dirac @xmath95 distribution ) with a gaussian beam . the source profile above includes a `` natural scale '' @xmath41 that characterises the source . this is a fundamental scale that will appear in all the filters we will consider here . by construction , the standard mf and saf operate on this scale , as well as the canonical mhw at the scale of the source . however , it has been shown that changing the scale at which the mhw and the mf filter the image can improve its performance in terms of detection ( vielva et al . 2001a , lpez - caniego et al . following this idea , we will introduce another degree of freedom in all the filters that allows us to change their scale in a continuous way ( similarly to the scaling of a continuous wavelet ) . this degree of freedom is obtained by multiplying the scale @xmath41 by a new parameter @xmath96 . we will show that with this new parameter the improvement in the results is significant . the idea of a scale - adaptive filter ( or optimal pseudo - filter ) has been recently introduced by the authors ( sanz et al . 2001 ) . by introducing a circularly - symmetric filter , @xmath97 , we are going to express the conditions in order to obtain a scale - adaptive filter for the detection of the source @xmath98 at the origin taking into account the fact that the source is characterised by a single scale @xmath99 . the following conditions are assumed : @xmath100 , i.e. @xmath101 is an _ unbiased _ estimator of the amplitude of the source ; @xmath102 the variance of @xmath103 has a minimum at the scale @xmath99 , i.e. it is an _ efficient _ estimator ; @xmath104 @xmath105 has a maximum with respect to the scale at @xmath106 . then , the filter satisfying these conditions is given by ( sanz et al . 2001 ) @xmath107 , \nonumber\end{aligned}\ ] ] @xmath108 assuming a scale - free power spectrum , @xmath109 , and a gaussian profile for the source , the previous set of equations lead to the filter @xmath110 , \ \ q\alpha r , \nonumber\end{aligned}\ ] ] @xmath111 @xmath112 where we have modified the scale as @xmath113 . in this case the filter parameters @xmath114 and the curvature of the source @xmath115 are given by @xmath116 @xmath117 @xmath118 @xmath119 figure [ fig : fig3 ] shows the saf for two values of the spectral index . if one removes condition ( 3 ) defining the saf in the previous subsection , it is not difficult to find another type of filter after minimization of the variance ( condition ( 2 ) ) with the constraint ( 1 ) @xmath120 this will be called _ matched _ filter as is usual in the literature . note that in general the matched and adaptive filters are different . for the case of a gaussian profile for the source and a scale - free power spectrum given by @xmath121 , the previous formula leads to the following modified matched filter @xmath122 @xmath123 where @xmath124 and @xmath125 is given by equation ( [ eq : saf_m ] ) and we have included the scale parameter @xmath126 . figure [ fig : fig3 ] shows the mf for the case @xmath127 ( standard mf ) and values of the spectral index @xmath128 . we remark that for @xmath129 the scale - adaptive filter and the matched filter coincide , and for @xmath130 ( not shown in the figure ) , the matched filter and the mexican hat wavelet are equal . for the mf the parameters @xmath131 and the curvature of the source @xmath115 are given by @xmath132 we remark that the linear detector @xmath133 is reduced to @xmath134 for the standard matched filter ( @xmath135 ) . i.e. curvature does not affect the region of acceptance for such a filter . the mh is defined to be proportional to the laplacian of the gaussian function : @xmath136 thus , in fourier space @xmath137 in this case the filter parameters @xmath138 and the curvature of the source @xmath115 are given by @xmath139 the generalization of this type of wavelet for two dimensions has been extensively used for point source detection in 2d images ( cayn et al . 2000 , vielva et al . 2001 , 2003 ) . as for the previous filter , the mh is modified by including the scale parameter @xmath126 in the form @xmath140 @xmath141 for the mh the parameters @xmath142 and the curvature of the source @xmath115 are given by @xmath143 figure [ fig : fig3 ] shows the mh for different values of the spectral index . if one removes condition ( 3 ) defining the saf in subsection 5.1.1 and introduces the condition that @xmath144 has a spatial maximum in the filtered image at @xmath145 , i.e. @xmath146 @xmath147 , it is not difficult to find another type of filter @xmath148 where @xmath149 is an arbitrary constant that can be related to the curvature of the maximum . we remark that the constraint @xmath150 is automatically satisfied for any circularly - symmetric filter if the source profile has a maximum at the origin . for the case of a scale - free power spectrum , the filter is given by the parametrized equation @xmath151 where we have modified the scale as @xmath113 . hereinafter , we will call this new filter containing two arbitrary parameters , @xmath152 and @xmath149 , the biparametric scale - adaptive filter ( bsaf ) . a calculation of the different moments leads to @xmath153 where m and t are defined in equation ( [ eq : saf_m ] ) and @xmath154 and @xmath155 are given by @xmath156\gamma ( m),\ ] ] @xmath157 \delta^m \gamma ( m).\ ] ] note that the bsaf contains all the other considered filters as particular cases : the mf is recovered for @xmath158 , when @xmath159 the bsaf defaults to the saf and , finally , the mh wavelet is obtained in the two cases : @xmath160 , @xmath161 and @xmath162 , @xmath163 . we will test two different _ pdf _ @xmath164 : a uniform distribution in the interval @xmath165 and a scale - free distribution with a lower and upper cut - off @xmath166 . in particular , we will especially focus on values for the cut - off s that lead to distributions dominated by weak sources . it is in this regime where sophisticated detection methods are needed , since bright sources can be easily detected with simple techniques . in this case , @xmath167.\ ] ] this allows us to obtain @xmath168,\ ] ] @xmath169 in general , we will consider a cut - off in the amplitude of the sources such that @xmath170 after filtering with he standard mf . note that this correspond to different thresholds for the rest of the filters . in this case , @xmath171,\ \ \beta \neq 1,\ ] ] where the normalization constant n and @xmath172 are @xmath173 @xmath174}.\\\ ] ] in general , we will consider @xmath175 and @xmath176 , @xmath177 after filtering with the standard mf and the corresponding thresholds for the other filters . for a fixed number density of spurious sources @xmath178 , we want to find the optimal filter that produces the maximum number density of true detections @xmath80 for different spectral indices ( @xmath179 ) , values of @xmath41 and point source distributions . in order to do this , we first obtain implicitly the value of @xmath74 from equation ( [ eq : nb ] ) ( for a fixed value of @xmath178 ) and then substitute it in equation ( [ eq : nb ] ) to calculate @xmath80 . we consider two different distributions of sources to test the robustness of the method : a uniform distribution and a scale - free distribution . given that bright point sources are relatively easy to detect , we mainly concentrate on the more interesting case of weak sources . in any case , we also mention some results for distributions containing bright sources . we remark that the bsaf has an additional degree of freedom , the parameter @xmath149 , as it appears in equation ( [ eq : eqnndf ] ) . note that the bsaf and the saf are not the same filter . the parameter @xmath149 in the bsaf can take any positive or negative value , while the coefficient @xmath180 , for the saf , is a known function of @xmath179 . by construction , the bsaf always outperforms the mf and saf or , in the worst case , defaults to the best of them . as a first case , we consider a uniform distribution of sources with amplitudes in the interval @xmath181\sigma_0 $ ] , where @xmath182 is the zero - order moment of the linearly - filtered map with the standard mf . therefore , the threshold @xmath38 in the image filtered with this filter is in the interval @xmath183 $ ] . thus , the corresponding upper limit for @xmath38 in the original ( unfiltered ) map is below 2 , what means that we are considering the detection of weak sources . as a reference example , in figure [ fig : numdet_alfa_unif_g0 ] , we plot @xmath80 , the number density of detections , as a function of @xmath126 for the case @xmath184 , @xmath185 and @xmath186 , where @xmath41 is given in pixel units . for completeness , the theoretical values of @xmath80 are given , in this figure , for values of @xmath126 down to zero ( note that @xmath80 @xmath187 @xmath178 when @xmath188 ) . however , from a practical point of view , we do not expect the theoretical results to reproduce the values obtained for a pixelized image when filtering at small scales ( since the effect of the pixel is not taken into account ) . therefore , hereinafter , we will only consider those results obtained when filtering at scales larger ( or of the order ) of the pixel size , which corresponds to @xmath189 . taking into account this constraint , the best results are obtained for @xmath190 for the bsaf , that clearly outperforms the standard mf ( i.e. , @xmath127 ) with an improvement of the @xmath0 in @xmath80 . if we compare with the mf at @xmath191 , the improvement is of @xmath192 . in figure [ fig : numdet_alfa_unif_g05 ] , we give the same results for the case @xmath193 . in this case , the bsaf at @xmath191 improves again significantly the standard mf , with an increase in the number density of detections of @xmath194 . as @xmath179 increases , the improvement of the bsaf with respect to the standard mf decreases . in fact , for values of @xmath195 they produce very similar results . as an example , we give the number of detections achieved for each filter for the case @xmath196 , @xmath197 and @xmath186 in fig . [ fig : unif_g1p5 ] . it can be seen that the maximum number of detections is approximately found for the standard mf . however , we would like to point out that the saf and mh wavelet at the optimal scale give approximately the same number of detections as the standard mf . these results show the importance of filtering at scales @xmath113 instead of the usual scale of the source @xmath41 . this can also be seen in fig . [ fig : unif_gamma ] , that summarizes how the relative performance of the considered filters with respect to the standard mf changes with the spectral index @xmath179 ( again for @xmath197 and @xmath186 ) . for each filter , the results are given for the optimal scale ( and parameter @xmath149 in the case of bsaf ) . the improvement of the bsaf with respect to the mf ranges from @xmath198 ( for white noise ) to zero ( for the largest values of @xmath199 . we would also like to point out that the mh at the optimal scale performs similarly to the standard mf . in addition , the mh has an analytical expression which makes it very robust and easy to implement . therefore , it can be a useful filter in some practical cases . we have also explored how the previous results change when varying @xmath178 and @xmath41 . in particular , we have considered vaules of @xmath178 in the interval 0.01 - 0.05 , @xmath200 and @xmath186 and values of @xmath201 . the results are summarized in table [ tab : tabla1_n ] for the bsaf and the standard mf ( we present only those cases where the bsaf improves at least a few per cent the standard mf ) . the values of @xmath126 and @xmath149 for the bsaf are found as the ones that maximise @xmath80 in each case . . number density of detections @xmath80 for the standard mf ( @xmath127 ) and the bsaf with optimal values of c and @xmath126 . rd means relative difference of number densities in percentage : @xmath202.[tab : tabla1_n ] [ cols="^,^,^,^,^,^,^,^",options="header " , ] we have also explored how the results depend on the values of @xmath41 and @xmath79 . in table [ tab : tabla2_n ] , we show the number density of detections for the bsaf and for the standard mf ( @xmath127 ) for @xmath200 and @xmath186 , with @xmath178 ranging from 0.01 to 0.05 , and for values of @xmath203 ( we only include the results for those cases where the relative difference between the bsaf and standard mf is at least a few per cent ) . we also give the optimal values of @xmath149 and @xmath126 where the bsaf performs better ( taking into account the constraint @xmath204 ) . as for the previous case of the uniform distribution , the relative performance of the bsaf improves when increasing @xmath41 and @xmath79 . it is also interesting to consider other values of the parameter @xmath205 . for instance @xmath206 $ ] has an intrinsic interest for astronomy , because they describe the distribution of compact sources in the sky at microwave wavelengths . we have also explored the performance of the filters for these values ( for the reference case @xmath184 , @xmath197 , @xmath186 ) and the improvement of the bsaf ( with optimal values of @xmath191 and @xmath207 versus the standard mf is still significant and of the order @xmath208 . to test the effect of the presence of bright sources on our results , we have also considered a scale - free distribution with @xmath209 \sigma_0 $ ] ( i.e. , a mixture of weak and bright sources ) for @xmath175 . we find , for the reference case ( @xmath184 , @xmath197 , @xmath186 ) , that the bsaf improves the standard mf around a @xmath210 , with optimal parameters @xmath191 and @xmath211 . we would like to point out that for for a given set of @xmath179 , @xmath41 and @xmath178 , this distribution of weak and bright sources leads to very similar optimal parameters for the bsaf as the scale - free distribution of weak sources . in addition , we have also tested the performance of the filters for a scale - free distribution of bright sources with @xmath212 \sigma_0 $ ] , for the same case as before ( @xmath184 , @xmath197 , @xmath186 ) . we explore the parameter space of @xmath213 , looking for the best filter regarding detection . we find that , for this distribution , the optimal parameters for the bsaf are @xmath158 and @xmath127 , that is , the bsaf defaults to the standard mf . the filters considered here depend on a number of parameters ( @xmath126 in the case of saf , mf and mh and @xmath126 and @xmath149 in the case of bsaf ) that must be determined in order to get the maximum number of detections for a fixed number of spurious detections . while for a given filter the region of acceptance is explicitly independent of the source distribution , the methodology presented here for the estimation of the optimal filter parameters depends on some assumed parameters of the source distribution ( namely @xmath205 , @xmath214 and @xmath215 ) and the noise power spectrum ( @xmath179 ) . a full study of the robustness of the method for all the filters is out of the scope of this work . however , we have considered some interesting cases as tests of the robustness of the method . in order to ascertain to what extent the uncertainties in the @xmath205 parameter of the source distribution affects the determination of the optimal filter parameters , we repeated our calculations using wrong assumptions on its value . an interesting case corresponds to assume that the source distribution is uniform when , in reality , it is scale - free and vice versa . in order to do this , we first construct the bsaf using the optimal @xmath216 values that were obtained for the uniform distribution . we calculate then the number density of sources @xmath80 obtained from a map that contains sources that follow a scale - free distribution with @xmath175 . we find that the differences in the number of detections when using the wrong filter with respect to using the optimal one are very small ( lower than @xmath217 ) . the same happens if the filter is constructed assuming an underlying scale - free distribution and applied to a map with sources uniformly distributed . this is not surprising , since tables [ tab : tabla1_n ] and [ tab : tabla2_n ] show that both uniform and scale - free distributions lead to similar values of the optimal @xmath126 and @xmath149 parameters . another source of uncertainty that appears in any real case is the value of the limiting cut - offs of the source distribution , @xmath214 and @xmath215 . we have seen in the previous subsections that for a given set of values @xmath79 , @xmath179 and @xmath41 , different cut - offs for the same distribution lead to similar optimal @xmath126 and @xmath149 parameters . for instance , in our reference example ( @xmath184 , @xmath197 , @xmath186 ) and a uniform distribution of sources with @xmath218 and @xmath219 the optimal filter parameters are @xmath191 and @xmath220 , whereas if the upper cut - off value is @xmath221 , the parameters take the values @xmath191 and @xmath222 . then , the shape of the optimal filter is only weakly dependent on the value of the cut - offs . this suggests that the methodology presented here is robust against uncertainties in the prior knowledge of the cut - offs of the distribution . in order to test this idea we proceeded in an analogous way to the case of the @xmath205 parameter explained before : we apply wrong filters ( that is , filters whose parameters have been determined assuming wrong values of the cutoffs ) to test cases with real distributions of sources . we tested several interesting cases : for uniform distributions , we studied the case when the sources are assumed to be weak but in reality some of them are bright ( in our example , @xmath215 is assumed to be 2 but in reality its true value is @xmath221 ) and the opposite situation . for the scale - free distribution , we studied the effect of mistaking the lower cut - off value ( assuming @xmath223 instead of its true value @xmath224 and vice versa ) . for all the cases , we plotted the curves @xmath225 versus @xmath126 . we observe that using a wrong filter changes the number of detections of all the filters , but the qualitative behaviour of the @xmath225 @xmath126 curves does not change . the relative behaviour of the filters is basically the same , and therefore the conclusions we obtained in the previous sections are still valid . thus , we conclude that the uncertainty in the knowledge of the source distribution is , in general , not a critical issue in the cases we have considered . a more delicate issue is the one related to the assumption of the @xmath179 parameter . if one assumes a value for @xmath179 that is very different from the true one , the shape of the filters changes dramatically ( except in the case of the mh whose shape is independent of @xmath179 ) with respect to the optimal ones and this may lead to wrong results . however , note that there are very well established techniques to estimate the power spectrum . albeit in this academic case we consider power law - type backgrounds , it is straightforward to apply the method to any kind of power spectrum that can be present in the data . as an example , we consider a case where the background corresponds to a true value @xmath226 whereas the filters have been constructed with a wrong @xmath227 , that is , a @xmath228 error in the determination of @xmath179 . the resulting @xmath229 curves are given in figure [ fig : gamma_falsa ] . the behaviour of the bsaf , the mf and the saf is qualitatively similar to the case where the noise power spectrum is perfectly known , but the performance of the three filters is poorer . the mh curve is identical to the ideal case since the shape of the filter does not depend on @xmath179 . the bsaf still outperforms all the other filters , although the improvement in the number of detections with respect to the mf slightly decreases . in order to see how our theoretical framework works in a practical example , we run a large set of simulations and study the performance of the np detector after filtering them with the different filters considered in the previous sections . we choose as a practical example the interesting case of a gaussian background characterised by a white noise power spectrum ( @xmath179=0 ) and sources whose intensity distribution is uniform . we will focus on the detection of weak sources . for the sake of simplicity , we give the results only for the bsaf and the mf since the other two filters ( saf and mh ) perform worse in the considered case . the different simulations are performed as follows . the images contain a number @xmath230 pixels , which is sufficiently large so that the addition of a single source does not modify significantly the dispersion of the images . the background is generated as a random field with dispersion ( before filtering ) @xmath231 ( in arbitrary units ) . the sources that we have considered for this example have a characteristic scale @xmath186 pixels . since we are interested in the detection of weak sources , we add point sources with a uniform amplitude distribution in the interval @xmath232 $ ] in the same arbitrary units of the background . the images filtered with the standard mf ( @xmath127 ) for this scale ( @xmath186 ) have dispersion @xmath233 . thus , the sources are distributed in the interval @xmath234 $ ] , where @xmath235 is the normalized amplitude of the sources with respect to the dispersion of the field filtered with the standard mf . for every maximum in a given image , it is possible to apply an empirical np criterion to decide whether the maximum corresponds or not to a source . the quantities in equations ( [ eq : r _ * ] ) and ( [ eq : phi ] ) can be obtained from simulations in the following way : the momenta @xmath182 , @xmath236 and @xmath237 ( and , therefore , the quantities @xmath238 , @xmath115 needed to know the value of the linear detector @xmath75 ) can be straightforwardly calculated from the image . for every maximum in the image , it is possible as well to measure directly its amplitude @xmath51 and curvature @xmath14 . the normalized curvature is easily obtained by fourier transforming the image , multiplying by @xmath239 and going back to real space . this gives the value of @xmath240 at each point and @xmath14 is obtained dividing by @xmath237 . the critical value @xmath74 that defines the acceptance region using equation ( [ eq : r _ * ] ) can be obtained as well directly from the simulations . for each considered filter , it is in principle possible to calculate @xmath74 semi - empirically , inverting equation ( [ eq : nb * ] ) ( with the empirically obtained values of @xmath238 and @xmath115 ) just as we did in the previous sections , and hence to proceed with the np decision rule . instead , since we are dealing with simulations , we will follow a fully empirical approach . the argument goes as follows . we fix the number density of spurious detections , i.e. , the number of maxima of the background that are misidentified as sources " by our detection criterion . then we simulate a set of images containing only background and filter them with the filter under study . we focus on the background maxima and try to determine the value of @xmath74 that makes the np rule to produce the specified number of spurious detections . for example let us consider that we perform @xmath241 noise realisations and focus on what happens in a certain pixel ( we choose the central pixel of the simulation in order to avoid border effects ) . for every realisation , we check if there is a maximum at this position or not . if a maximum is present , the value of @xmath75 is calculated . all the values of @xmath75 obtained in that way are sorted into descending numbers ( large to small ) . the value of @xmath74 is then given by the @xmath75 corresponding to the r - th element @xmath242 of the sorted list ( that is , @xmath74 is the value of @xmath75 so that there are @xmath243 background maxima with @xmath244 ) . for this example , we considered @xmath185 and therefore @xmath245 . once the value of @xmath74 has been empirically determined we add a source with a gaussian profile of dispersion @xmath246 pixels at the central position ( pixel @xmath247 ) of the unfiltered background image and then we filter it with the considered filter . we proceed to apply the detector to any maxima located at the pixel we are considering . finally , @xmath80 , for each of the filters , will be the number of sources with an estimated @xmath244 divided by the total number of realizations @xmath248 . we have performed numerical simulations for our reference case ( @xmath185 , @xmath186 and @xmath184 ) , assuming a uniform distribution with @xmath170 . we have compared the performance of the filters and the empirical np detector in the simulated images with the theoretical predictions as a function of the parameter @xmath126 . for each @xmath126 value and each filter , we have done five sets of simulations of the background , each one with a number of realizations large enough to have 5000 of them containing a maximum in the central pixel . ] . we used these simulations to obtain @xmath74 as explained before . in the figure [ fig : numdet_sims_todo ] we present the results from the simulations for this case and the comparison with the theoretical calculations . the lines in this plot show the theoretical results for each filter , the triangles the result from the simulations for the bsaf and the squares the results for the mf . we concentrate on the bsaf , which corresponds to the dot - dash line . as we mentioned in previous sections , the bsaf significantly improves the standard mf for @xmath184 . the simulations follow the theoretical results well . in the region where @xmath190 , there is a small deviation from theory which we believe is related to the fact that we are close to the scale of the pixel , but still , significantly close to the expected theoretical value . we can estimate the amplitude of a source using the unbiased and maximum efficient estimator from equation ( [ eq : estim_amplitude ] ) and then compare it with the amplitude that we have randomly generated . in figure [ fig : estim_ampli1 ] , we plot the real amplitude versus the estimated one for the bsaf ( top panels ) and the mf ( bottom panels ) . two amplitude regimes have been explored . in the left panels , we have simulated a uniform distribution of weak sources , @xmath181\sigma_0 $ ] . the parameters used for these simulations are @xmath185 , @xmath184 and @xmath186 . the optimal filter parameters have been chosen at each case . the points and the error bars are calculated as the average and the dispersion of the detected sources that fall in each of the amplitude bins from a total of @xmath249 10000 detected sources . we find a similar positive bias in the determination of the amplitude for the bsaf ( @xmath250 , @xmath220 ) and mf ( @xmath250 ) . however , the error bars corresponding to the bsaf are slightly smaller than those of the mf . in the right panels , we give the results for a uniform distribution of sources with @xmath251\sigma_{0}$ ] . the simulation parameters are the same as before ( @xmath185 , @xmath184 and @xmath186 ) . as before , the points and the error bars are calculated as the average and the dispersion of the detected sources that fall in each of the amplitude bins from a total of @xmath249 20000 detected sources . we find that the bsaf ( @xmath191 , @xmath222 ) is unbiased for bright sources whereas the estimation of the amplitude in the case of the standard mf ( @xmath127 ) shows some bias even for bright sources . therefore , the bsaf with @xmath191 outperforms the standard mf in both the detection and estimation for this distribution . the fact that sources with small amplitudes are significantly affected by a positive bias can be explained taking into account that these sources are more easily detected if they lie over a positive contribution of the background . this contributes systematically to the overestimation of the amplitude . we would like to point out that this estimator produces appreciably better results than a naive estimation using directly the measured values at the maxima . nowadays , the detection of compact sources on a background is a relevant problem in many fields of science . a number of detection techniques use linear filters and thresholding - based detectors . our approach to the problem of detector design is different . we use a neyman - pearson rule that takes into account _ a priori _ information of the distribution of sources and the number density of maxima to define the region of acceptance . in our case , we take advantage not only of the amplification but also of the spatial information : the curvature of the background is different from that of the sources , and we use this to improve our detection rule . the background is modelled by a homogeneous and isotropic gaussian random field , characterized by a scale - free power spectrum @xmath252 , @xmath253 . we design a new filter that we call bsaf in such a way that the use of our improved detection rule based on amplification and curvature on the filtered field will increase the number of detections for a fixed number of spurious sources . we generalize the functional form of this filter , as well as other standard filters , and introduce another degree of freedom , @xmath126 , that allows us to filter at any scale , including that of the source @xmath41 . we have shown the benefits of filtering at scales smaller than @xmath41 , which significantly improves the number of detections . as an example , we have considered two different distributions of sources . a uniform distribution in the interval @xmath234 $ ] and a scale - free power law distribution in the interval @xmath254 $ ] ( where the threshold @xmath38 corresponds to the field filtered with the standard mf ) , i.e. we are considering weak sources . the bsaf has proven to be significantly better than the standard mf , the saf and mh wavelet in certain cases . in particular , we have considered a reference case with parameters @xmath184 , @xmath197 and @xmath186 , where the improvement in the number of detections of the bsaf at @xmath191 with respect to the standard mf is @xmath255 . we have also tested the performance of the filters for a mixture of weak , intermediate and bright sources . for a uniform distribution with @xmath256 $ ] and for a scale - free distribution with @xmath257 $ ] , the bsaf also improves the mf . however , for a scale - free distribution with @xmath258 $ ] , i.e. , dominated by bright sources , we find that the optimal bsaf defaults to the standard mf , which gives the maximum number of detections in this case . we find that the bsaf gives in any case the best performance among the considered filters . indeed , the saf and the mf are particular cases of the bsaf and the strategy we follow , i.e. maximization of the detections , guarantees that the parameters of the bsaf will default to the best possible of these filters in each case . in addition , we also find that the bsaf performs at least as well as the mh in all the considered cases . therefore , the number density of detections obtained with the bsaf will be at least equal to the best of the other three filters , and in certain cases superior . however , in some other cases , the gain is small and it is justified to use an analitically simpler filter . our results suggest that for power law spectra , from the practical point of view , one could use the bsaf when @xmath259 since , in this range , clearly improves the number of detections with respect to the other filters . however , for @xmath260 the usage of the mh is justified due to its robustness ( since it has an analytical form ) and it gives approximately the same number of detections obtained either with the bsaf or mf . for all the studied cases of source distributions ( except for the one dominated by bright sources ) and fixing the values of @xmath179 , @xmath79 and @xmath41 , we find that the optimal parameters of the bsaf are only weakly dependent on the distribution of the sources . we have done some simple tests in order to study the robustness of the method when the knowledge about the source _ pdf _ or the background spectral index is not perfect . we find that the values of the optimal filter parameters vary slightly when we assume that the source distribution is uniform when , in reality , it is scale - free and vice versa . the uncertainties in the cut - off values of the source _ pdf _ affect the number of detections , but in a similar way for all the filters , and therefore the relative behaviour of the filters do not change . errors in the estimation of the spectral index @xmath179 reduces the efectiveness of the bsaf , but it still outperforms the other filters . all of this indicates that our detection scheme is robust against uncertainties in the knowledge of the distribution of the sources and spectral index . to test the validity of our results in a practical example , we have tested our ideas with simulations for the uniform distribution ( using our reference case @xmath197 , @xmath186 , @xmath184 ) and find that the results follow approximately the expected theoretical values regarding source estimation , we propose a linear estimator which is unbiased and of maximum efficiency , that we have also tested with simulations . the ideas presented in this paper can be generalized : application to other profiles ( e.g. multiquadrics , exponential ) and non - gaussian backgrounds is physically and astronomically interesting . the extension to include several images ( multi - frequency ) is relevant . the generalization to two - dimensional data sets ( flat maps and the sphere ) and nd images is also very interesting . finally the application of our method to other fields is without any doubt . we are currently doing research in some of these topics . the authors thank enrique martnez - gonzlez and patricio vielva for useful discussions . mlc thanks the ministerio de ciencia y tecnologa ( mcyt ) for a predoctoral fpi fellowship . rbb thanks the mcyt and the universidad de cantabria for a ramn y cajal contract . dh acknowledges support from the european community s human potential programme under contract hprn - ct-2000 - 00124 , cmbnet . we acknowledge partial support from the spanish mcyt project esp2002 - 04141-c03 - 01 and from the eu research training network ` cosmic microwave background in europe for theory and data analysis ' . de zotti , g. , toffolatti , l. , argeso , f. , davies , r.d . , mazzotta , p. , partridge , r.b . , smoot g.f . & vittorio , n. , 1999 , 3k cosmology , proceedings of the ec - tmr conference held in rome , italy , october , 1998 . woodbury , n.y . : american institute of physics , vol . 476 , 204 . the ratio @xmath261 can be explicitly written as @xmath262 and taking into account the np criterion for detection , we find @xmath263 where @xmath12 is a constant . by differentiating the previous equation with respect to @xmath75 @xmath264 therefore , @xmath265 is equivalent to @xmath244 , where @xmath74 is a constant , i.e. @xmath133 given by equation ( [ eq : phi ] ) is a sufficient linear detector . let us assume a linear estimator combination of the normalized amplitude @xmath38 and normalized curvature @xmath14 with the constraint @xmath266 if the estimator is unbiased , i.e. @xmath267 , taking into account that @xmath268 and @xmath269 , we obtain the constraint @xmath270 on the other hand , the variance is given by @xmath271 where we have taken into account that @xmath272 . by minimizing the previous expression with the constraint ( [ eq : constraint_appc ] ) , one obtains @xmath273 therefore , one obtains : @xmath274 @xmath275	this paper considers the problem of compact source detection on a gaussian background . we make a one - dimensional treatment ( though a generalization to two or more dimensions is possible ) . two relevant aspects of this problem are considered : the design of the detector and the filtering of the data . our detection scheme is based on local maxima and it takes into account not only the amplitude but also the curvature of the maxima . a neyman - pearson test is used to define the region of acceptance , that is given by a sufficient linear detector that is independent on the amplitude distribution of the sources . we study how detection can be enhanced by means of linear filters with a scaling parameter and compare some of them that have been proposed in the literature ( the mexican hat wavelet , the matched and the scale - adaptive filters ) . we introduce a new filter , that depends on two free parameters ( biparametric scale - adaptive filter ) . the value of these two parameters can be determined , given the a priori _ pdf _ of the amplitudes of the sources , such that the filter optimizes the performance of the detector in the sense that it gives the maximum number of real detections once fixed the number density of spurious sources . the new filter includes as particular cases the standard matched filter and the scale - adaptive filter . then , by construction , the biparametric scale adaptive filter outperforms these filters . the combination of a detection scheme that includes information on the curvature and a flexible filter that incorporates two free parameters ( one of them a scaling ) improves significantly the number of detections in some interesting cases . in particular , for the case of weak sources embedded in white noise the improvement with respect to the standard matched filter is of the order of @xmath0 . finally , an estimation of the amplitude of the source ( most probable value ) is introduced and it is proven that such an estimator is unbiased and it has maximum efficiency . we perform numerical simulations to test these theoretical ideas in a practical example and conclude that the results of the simulations agree with the analytical ones . methods : analytical - methods : data analysis - techniques : image processing
introduced by topologists @xcite , @xmath0-betti numbers have been generalized to various contexts like groups , groupoids etc . alain connes and dimitri shlyakhtenko @xcite introduced @xmath0-betti numbers for subalgebras of finite von neumann algebras , with the purpose to obtain a suitable notion for arbitrary ii@xmath1-factors and in the hope to get a nice homological invariant for them . unfortunately , as of now there are only very few concrete calculations of them . the most advanced computational result so far is due to andreas thom @xcite who proved that the @xmath0-betti numbers vanish for von neumann algebras with diffuse center . to allow more computable examples , he also introduced a continuous version of the first @xmath0-betti number @xcite which turns out to be much more manageable than its algebraic counterpart . the first continuous @xmath0-betti number is defined as the von neumann dimension of the first continuous hochschild cohomology of the von neumann algebra @xmath2 with values in the algebra of operators affiliated with @xmath3 . the word ` continuous ' here means that we restrict attention to derivations which are continuous from the norm topology on @xmath2 to the measure topology on the affiliated operators . so far only vanishing results were obtained about the first continuous @xmath0-betti number : it has been shown to vanish for ii@xmath1-factors with cartan subalgebras , non - prime ii@xmath1-factors @xcite as well as for ii@xmath1-factors with property ( t ) , property @xmath4 and finitely generated ii@xmath1 factors with nontrivial fundamental group @xcite . the last result is due to a compression formula for the first continuous @xmath0-betti number ( * ? ? ? * theorem 4.10 ) . the hope placed upon @xmath0-betti numbers for group von neumann algebras was to be able to connect them with @xmath0-betti numbers of groups , thus obtaining a powerful invariant which would be able to distinguish free group factors , thus solving a long - standing problem in operator algebras . in fact , the attempt to do this can be formulated in a very concrete way using generators of the @xmath0-cohomology of the group ring @xmath5 of the free group or some other subalgebras of @xmath6 generated by free elements . one possible choice of generators is to consider the so - called voiculescu s free difference quotients @xcite . andreas thom posed a natural question in @xcite , whether these derivations possess continuous extensions to operators from @xmath6 to @xmath7 ; a positive answer to this question would solve the free factor isomorphism problem . in the present paper we answer this question in the negative ; in fact , we show that the first continuous @xmath0-cohomology of free group factors vanishes ; in particular , they can not be distinguished by this invariant . this also suggests that the invariant might be altogether trivial , i.e. that the first continuous @xmath0-cohomology might in fact vanish for all ii@xmath1-factors . the result is established in several steps . first , we focus on the free group with three generators @xmath8 and show that the canonical derivations which `` derive in direction of a free generator '' can not be extended to the group von neumann algebra . this is shown by analyzing their values on some specific elements for which the spectrum of the resulting operators can be calculated using free probability theory . to derive the vanishing of the whole continuous cohomology , we have to use certain automorphisms of the free group factors . hereby we make use of certain weak mixing properties relative to a subalgebra ; intuitively speaking , we are using the fact that there are enough automorphisms to move our derivations around ; thus , the existence of one continuous non - inner derivation would automatically guarantee that all derivations of @xmath9 are extendable , which yields a contradiction . finally , we make use of the compression formula to extend the result from a single free group factor to all of them . the author thanks thomas schick and andreas thom for helpful discussions and useful suggestions . in this section we set up the notation and briefly recapitulate the theory of non - commutative integration and the theory of @xmath0-betti numbers for von neumann algebras . we consider finite von neumann algebras @xmath2 , @xmath10 etc . with separable preduals . we always endow them with a fixed faithful normal tracial state ( usually denoted by @xmath11 ) and consider them in the corresponding gns representation @xmath12 . if @xmath13 is a finite von neumann algebra , then there is an induced a faithful normal tracial state on the von neumann algebraic tensor product @xmath14 of @xmath10 with its opposite algebra ; abusing notation slightly , we will still denote it by @xmath11 . we let @xmath15 be the algebra of closed densely defined operators on @xmath12 affiliated with @xmath10 . we equip @xmath15 with the _ measure topology _ , defined by the following two - parameter family of zero neighbourhoods : n ( , ) = \{au(n)\| p(n ) : < , ( p^ ) < } , , > 0 . with this topology , @xmath15 is a complete ( * ? ? ? * theorem ix.2.5 ) metrizable ( * ? ? ? * theorem 1.24 ) topological vector space and the multiplication map ( a , b)ab u(n)u(n ) u(n ) is uniformly continuous when restricted to products of bounded subsets ( * ? ? ? * theorem 1 ) . convergence with respect to the measure topology is also referred to as _ convergence in measure _ and denoted by @xmath16 . if @xmath17 and @xmath18 is its source projection , we denote @xmath19 . of course , we also have @xmath20 , where @xmath21 is the target projection of @xmath22 . here and in the sequel @xmath23 denotes the algebric tensor product over @xmath24 . we freely identify @xmath2-@xmath2-bimodules with @xmath25-modules . for @xmath26 we equip @xmath27 with the @xmath2-@xmath2-bimodule structure @xmath28 all @xmath2-@xmath2-sub - bimodules of @xmath27 inherit this bimodule structure . let @xmath4 be a discrete group with its natural left action @xmath29 on itself . its von neumann algebra @xmath30 is equipped with the natural faithful normal tracial state @xmath31 . notice that the gns representation of @xmath32 with respect to @xmath11 coincides with @xmath33 . let @xmath34 be an algebra and @xmath35 an @xmath34-@xmath34-bimodule . we recall that a linear map @xmath36 is called a _ derivation _ if it satisfies @xmath37 and that a derivation is called _ inner _ if there exists a vector @xmath38 such that @xmath39 consider a free group @xmath40 generated by a set @xmath41 . for @xmath42 , a derivation @xmath43 is defined uniquely by the properties @xmath44 here we prove several lemmas which will help us to analyse convergence in measure of some particular operators . the following lemma tells us that we can analyse convergence in measure `` locally '' . [ lemma : measure - conv - local - crit ] let @xmath45 be a sequence . then @xmath46 if and only if [ eq : conv - in - measure - to - zero ] > 0(_[^2,+)(a_n^ * a_n ) ) 0,n . suppose that is satisfied . for every @xmath47 set @xmath48 . we obtain @xmath49 and @xmath50 thus , @xmath46 . on the other hand , let @xmath51 and suppose that is false . then , extracting a subsequence if needed , we may assume that @xmath52 suppose now that there exists a sequence of projections @xmath53 such that @xmath54 then for sufficiently big @xmath55 we get @xmath56 . putting @xmath57 , we obtain @xmath58 for @xmath59 , which satisfies @xmath60 , obtaining a contradiction . [ lemma : no - convergence - to - zero ] let @xmath61 be a sequence of selfadjoint elements such that @xmath62}(a_n)\right ) \to 0,\quad n\to\infty.\ ] ] then for every nonzero projection @xmath18 @xmath63 from the assumption of the lemma it immediately follows that @xmath64 setting @xmath57 , we get @xmath65 . now , if @xmath66 , then there exists an @xmath67 such that @xmath68 . thus for such @xmath55 we obtain that the projections @xmath69 satisfy @xmath70 . it follows that @xmath71 and hence by lemma [ lemma : measure - conv - local - crit ] @xmath72 . on the other hand , if @xmath73 , then by boundedness of @xmath74 @xmath75 this proves the lemma . in @xcite connes and shlyakhtenko introduced @xmath0-betti numbers in the general setting of tracial @xmath76-algebras ; if @xmath2 is a finite von neumann algebra and @xmath77 is any weakly dense unital @xmath78-subalgebra its @xmath0-betti numbers are defined as @xmath79 here the dimension function @xmath80 is lck s extended von neumann dimension ( * ? ? ? * chapter 6 ) . this definition is inspired by the well - known correspondence between representations of groups and bimodules over finite von neumann algebras , and it extends the classical theory by means of the formula @xmath81 whenever @xmath4 is a discrete countable group . in @xcite it is shown that the @xmath0-betti numbers also allow the following cohomological description : @xmath82 where @xmath83 denotes the algebra of operators affiliated with @xmath3 . it is a classical fact @xcite that the ext - groups above are isomorphic to the hochschild cohomology groups of @xmath34 with coefficients in @xmath84 , where the latter is considered as an @xmath34-bimodule with respect to the actions @xmath85 in particular , the first @xmath0-betti number can be computed as the dimension of the right @xmath3-module @xmath86 here @xmath87 denotes the space of derivations from @xmath34 to @xmath84 and @xmath88 denotes the space of inner derivations . these purely algebraically defined @xmath0-betti numbers have turned out extremely difficult to compute in the case when @xmath34 is @xmath2 itself . besides finite - dimensional algebras , the only computational result known in this direction vanishing for von neumann algebras with diffuse centre ( see ( * ? ? ? * corollary 3.5 ) and ( * ? ? ? * theorem 2.2 ) ) . it is therefore natural to consider variations of the definitions above that take into account the topological nature of @xmath2 . the continuous version of the first @xmath0-cohomology module was introduced by andreas thom in @xcite , where one restricts attention to those derivations @xmath89 which are closable from the norm topology to the measure topology . note that when @xmath34 is norm closed these are exactly the derivations that are norm measure topology continuous by the closed graph theorem . we denote the space of closable derivations by @xmath90 , the continuous cohomology by @xmath91 and by @xmath92 the corresponding continuous @xmath0-betti numbers ; i.e. @xmath93 notice that by continuity of multiplication on @xmath84 , @xmath90 is naturally a right @xmath84-module . the first continuous @xmath0-betti number satisfies the following compression formula analogous to ( * ? ? ? * theorem 2.4 ) : [ thm : compression - formula ] let @xmath2 be a @xmath94-factor with trace - state @xmath11 and let @xmath95 be a non - zero projection . then @xmath96 . although the extended von neumann dimension is generally not faithful , enlarging the coefficients from @xmath3 to @xmath84 has the effect that @xmath97 if and only if @xmath98 vanishes ( * ? ? ? * corollary 3.3 and theorem 3.5 ) . in particular , in order to prove that @xmath99 one has to prove that every derivation from @xmath34 into @xmath84 is inner . an analogous statement holds for continuous @xmath0-cohomology : [ prop : vanishing - cohomology - vs - vanishing - betti ] let @xmath2 be a finite von neumann algebra . we have @xmath100 if and only if @xmath101 . we will be interested in continuity of certain derivations . to understand it , we recollect some useful notions and properties here . let @xmath77 be a weakly dense @xmath78-subalgebra and @xmath102 be a derivation . let @xmath103 [ lemma : complete - lattice ] @xmath104 is a complete sublattice of @xmath105 . if @xmath106 , the right @xmath84-submodule generated by @xmath107 contains the derivation @xmath108 , where @xmath109 is the target projection of @xmath22 . if now @xmath110 , then @xmath111 for obvious reasons , and @xmath112 because @xmath113 is continuous and @xmath114 . if now @xmath115 is an orthogonal family of projections with sum @xmath116 , then @xmath117 and therefore the series @xmath118 converges uniformly in measure to @xmath119 . as sums of uniformly convergent series of continuous maps are continuous , @xmath120 is continuous . we call the unique supremum of @xmath104 the _ continuity projection _ of @xmath121 . let @xmath89 be a derivation and @xmath122 be an automorphism ; by slight abuse of notation , we still denote by @xmath123 the induced automorphism @xmath124 . the map @xmath125 is then a derivation @xmath126 . if @xmath121 is continuous , then @xmath127 is obviously continuous as well . notice that for @xmath128 we have @xmath129 the following observation is easy , but very useful . [ lemma : invariant - projection ] if @xmath130 is a subgroup of automorphisms of @xmath2 and @xmath131 for @xmath132 , then the continuity projection of @xmath121 is @xmath133-invariant . let @xmath116 be the continuity projection of @xmath121 . then @xmath120 is continuous , and therefore @xmath134 is continuous as well . if for some @xmath132 we have @xmath135 , then @xmath136 has @xmath116 as a proper subprojection , contradicting maximality . let @xmath137 be a trace - preserving automorphism of @xmath138 , where @xmath139 is the canonical trace . it induces automorphisms of @xmath140 obtaining by decomposing @xmath140 as a free product with respect to the subalgebras @xmath141 , @xmath142 , @xmath143 generated by @xmath144 , @xmath145 and @xmath146 using the free product decomposition @xmath147 . we denote @xmath148 @xmath149 @xmath150 we get actions of the groups @xmath151 , @xmath152 , and @xmath153 on @xmath140 . the following definition of the relative mixing property appeared in ( * ? ? ? * definition 2.9 ) . we will use a strengthening of it . let @xmath154 be a trace - preserving inclusion of finite von neumann algebras and let @xmath155 be a trace - preserving action of a group @xmath156 on @xmath2 such that @xmath157 for all @xmath158 . the action @xmath123 is called weakly mixing relative to @xmath10 if for every finite set @xmath159 and for every @xmath47 there is a @xmath158 such that @xmath160 an action @xmath123 which is weakly mixing relative to @xmath161 is called weakly mixing . when @xmath10 is pointwise fixed by the action , we get the following result resembling the classical equivalent characterisations of weakly mixing actions ( cf . * proposition d.2 ) ) . [ prop : weak - mixing - conditions ] let @xmath154 be a trace - preserving inclusion of finite von neumann algebras @xmath155 be a trace - preserving action such that @xmath162 for all @xmath158 and @xmath163 . then each of the following conditions implies the next one : 1 . @xmath123 is weakly mixing relative to @xmath10 ; 2 . for every @xmath164 there exists a sequence @xmath165 such that for every @xmath166 @xmath167 , @xmath168 , @xmath169 . every finite - dimensional invariant subspace of @xmath2 is contained in @xmath10 ; 4 . for every action @xmath170 of @xmath156 on a finite von neumann algebra @xmath171 @xmath172 the proof of ( * ? d.2 ) goes through with small modifications . i ) @xmath173 ii ) is immediate . to show ii ) @xmath173 iii ) , take a finite - dimensional invariant subspace @xmath174 . consider the space @xmath175 ; it is also finite - dimensional and @xmath156-invariant . but in view of ii ) we have for every @xmath176 and @xmath166 that @xmath177 it means that @xmath178 weakly as @xmath179 . as @xmath180 is finite - dimensional , this implies convergence to zero in norm , and as the @xmath156-action is norm - preserving , this means that @xmath181 . thus , @xmath182 , thus @xmath183 lies in @xmath10 . to show iii ) @xmath173 iv ) , take @xmath184 and view it as a hilbert schmidt operator @xmath185 . then the image of @xmath186 is contained in @xmath2 , and the operator @xmath187 is trace - class and commutes with the @xmath156-action . taking its spectral projection , we obtain a finite - dimensional @xmath156-invariant subspace of @xmath2 , which is necessarily contained in @xmath10 by iii ) . thus , the image of @xmath186 lies in @xmath10 , and therefore @xmath188 . [ lemma : weak - mixing - free - product ] let @xmath189 and @xmath10 be finite von neumann algebras and @xmath190 a trace - preserving weakly mixing action of a group @xmath156 on @xmath189 . then the action @xmath191 of @xmath156 on the free product von neumann algebra @xmath192 is weakly mixing relative to @xmath10 . by a standard density argument it is enough to check the relative weak mixing condition on the algebraic free product @xmath193 , which is the algebra generated by @xmath189 and @xmath10 inside @xmath2 . it is spanned by @xmath10 and alternating products of elements from @xmath194 und @xmath195 ; without loss of generality we may and will assume that the operator norms of all factors are bounded by 1 . moreover we have @xmath196 and @xmath197 for @xmath198 . thus , for @xmath199 we get @xmath200 @xmath201 thus , @xmath202 is spanned by the alternating products of elements from @xmath194 and @xmath195 , and it s enough to check the weak mixing property for such alternating products . given a finite set @xmath203 of them , let @xmath204 be all factors from @xmath205 occuring in the products . by the weak mixing property we find a @xmath158 such that @xmath206 the above formulae for @xmath207 then imply for two elements @xmath208 that @xmath209 unless @xmath210 @xmath211 where @xmath199 , @xmath212 , @xmath213 , @xmath214 , @xmath215 , @xmath216 with @xmath217 , @xmath218 . in this case we get @xmath219 which proves the statement . using the existence of weakly mixing actions on an arbitrary finite von neumann algebra ( e.g. bernoulli actions ( * ? ? ? 2.4 ) ) and proposition [ prop : weak - mixing - conditions ] , we obtain [ lemma : perm - invariant ] consider the actions of @xmath151 , @xmath152 , and @xmath153 on @xmath140 described above . then 1 . every @xmath220-invariant element @xmath221 is contained in @xmath222 ; 2 . every @xmath223-invariant element @xmath221 is contained in @xmath224 . let @xmath225 be a free group on three generators . we will naturally view @xmath226 as a subgroup of @xmath8 . it s well - known that @xmath227 contains a copy of @xmath228 , and we well fix such a copy . we will first construct a particular sequence @xmath233 $ ] such that @xmath234 , but @xmath235 . we set @xmath236\ ] ] and consider @xmath237 the elements @xmath238 are free . indeed , the family @xmath239 is free , because @xmath240 is a free family which is itself free from @xmath144 . therefore the elements @xmath241 are free and hence form a freely independent family of unitaries @xmath242 . consider the elements @xmath243 the spectral density of this operator can be computed using free probability theory . indeed , @xmath244 is an instance of a scaled random walk operator on a free group , and therefore we can use the formula for the cauchy transform for its spectral density from ( * ? ? ? * example 3.4.5 ) : @xmath245 using the stieltjes inversion formula , we get the spectral density of @xmath244 : @xmath246,\\ 0 & \text{otherwise}. \end{cases}\ ] ] we see that @xmath247}\left(\frac{h_n}{\sqrt[4]{n}}\right)\right ) = \int_{-\eps\sqrt[4]{n}}^{\eps\sqrt[4]{n } } \frac{n\sqrt{(2n-1)-x^2}\,dx}{n^2-x^2}\leqslant \frac{2\eps\cdot n^{5/4}\sqrt{2n-1}}{n^2-\eps^2\sqrt{n}}\to 0,\quad n\to\infty.\ ] ] in view of lemma [ lemma : no - convergence - to - zero ] , for every nonzero projection @xmath248 we get @xmath249{n } } \cdot p \not\xrightarrow{m } 0,\quad n\to\infty.\ ] ] in particular , @xmath250{n}}\right ) \not\xrightarrow{m } 0,\quad n\to\infty,\ ] ] although @xmath251{n}$ ] converges to @xmath252 in norm , because @xmath253 . thus , @xmath254{n}$ ] satisfies the required properties , and the derivation @xmath255 is not continuous . now we have to show that the continuity projection @xmath21 of @xmath121 is equal to @xmath252 . the derivation @xmath255 is invariant under the actions of the groups @xmath256 and @xmath257 . by lemma [ lemma : invariant - projection ] we get that @xmath21 is invariant under @xmath256 and @xmath257 , hence by corollary [ lemma : perm - invariant ] , i ) it belongs to @xmath222 . now , for @xmath258 we infer that @xmath259 is continuous . the support projection of @xmath260 is equal to @xmath261 , and by maximality of @xmath21 it follows that @xmath262 . therefore , @xmath263 or @xmath264 , but the latter case is impossible because @xmath255 is not continuous . for obvious symmetry reasons , the statement of proposition [ prop : no - continuous - extension ] is also true for the derivations @xmath265\to \ms u(l\mb f_3\htens l\mb f_3^\op)\ ] ] uniquely determined by the conditions @xmath266 resp . @xmath267 is is well - known that for every nonzero projection @xmath248 the derivations @xmath232 , @xmath268 , @xmath269 are not inner . the derivations @xmath255 , @xmath270 , @xmath271 freely generate the module @xmath272 . the restriction of @xmath121 to @xmath274 is a derivation and therefore can be uniquely written as a combination of @xmath255 , @xmath270 , @xmath271 : @xmath275 for some @xmath276 . as inner derivations are continuous , after subtracting an inner derivation we may assume that @xmath277 and @xmath278 the right @xmath84-module generated by @xmath279 contains @xmath280 , where @xmath281 is the target projection of @xmath282 . as the right @xmath84-action preserves continuity , we may assume that @xmath283 and that @xmath121 has the form @xmath284 multiplying from the right with @xmath285 and using proposition [ prop : no - continuous - extension ] , we deduce @xmath286 . thus , @xmath287 ; reasoning symmetrically , we infer @xmath288 . we also observe that for every @xmath289 , the elements @xmath281 and @xmath290 are uniquely determined . now , let @xmath291 analogously to lemma [ lemma : complete - lattice ] , we are going to prove that @xmath292 is a complete lattice . indeed , if @xmath293 , then @xmath294 because @xmath295 . now , if @xmath296 and @xmath297 are derivations corresponding to @xmath298 resp . @xmath299 , then @xmath300 , being the support projection of @xmath301 , corresponds to the derivation @xmath302 . completeness of @xmath292 is proven as follows : if let @xmath303 be an orthogonal family in @xmath292 with corresponding derivations @xmath304 with elements @xmath305 . by the observation above , @xmath306 . but then @xmath307 and therefore as in lemma [ lemma : complete - lattice ] the series @xmath308 converges uniformly in measure to a derivation @xmath121 having the supremum of @xmath309 as the corresponding projection @xmath281 . thus , @xmath292 is a complete lattice . let @xmath116 be its maximal element . in view of the equalities @xmath310 and @xmath311 for @xmath312 and @xmath313 we get by maximality that @xmath116 is invariant under @xmath314 and @xmath257 . therefore by corollary [ lemma : perm - invariant ] , i ) we deduce that @xmath315 . now , for @xmath258 we get that @xmath316 is continuous . the support projection of @xmath317 is equal to @xmath318 , and by maximality of @xmath116 it follows that @xmath319 . therefore , @xmath320 or @xmath321 . if @xmath321 , we are given a continuous derivation of the form @xmath322 for @xmath323 we obtain @xmath324 as @xmath325 is continuous , proposition [ prop : no - continuous - extension ] implies that @xmath326 thus using corollary [ lemma : perm - invariant ] , ii ) , we get that @xmath327 , where @xmath328 is generated by @xmath144 and @xmath145 . consider the restriction of the derivation @xmath121 to @xmath329 . subtracting an inner derivation and multiplying with a suitable element of @xmath330 from the right , we may assume that the derivation @xmath331 is continuous for some nonzero projection @xmath332 . thus , the continuity projection of @xmath333 is nonzero . arguing as in proposition [ prop : no - continuous - extension ] , we deduce that the continuity projection of @xmath333 is equal to 1 . for symmetry reasons , @xmath334 is continuous as well . therefore the module of continuous derivations @xmath335 is two - dimensional , and hence the first continuous @xmath0-betti number of @xmath329 is equal to @xmath336 : @xmath337 but then by the compression formula ( theorem [ thm : compression - formula ] ) we obtain @xmath338 which contradicts the result of proposition [ prop : no - continuous - extension ] that the submodule generated by @xmath255 consists of discontinuous derivations . in particular , the answer to the question of andreas thom in @xcite is negative : voiculescu s free difference quotients do nt have continuous extensions to @xmath6 . our result also allows to ask the following question . m. f. atiyah . elliptic operators , discrete groups and von neumann algebras . in _ colloque `` analyse et topologie '' en lhonneur de henri cartan ( orsay , 1974 ) _ , pages 4372 . astrisque , no . 3233 . soc . france , paris , 1976 . jean - louis loday . , volume 301 of _ grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences]_. springer - verlag , berlin , second edition , 1998 . appendix e by mara o. ronco , chapter 13 by the author in collaboration with teimuraz pirashvili . wolfgang lck . , volume 44 of _ ergebnisse der mathematik und ihrer grenzgebiete . 3 . folge . a series of modern surveys in mathematics [ results in mathematics and related areas . 3rd series . a series of modern surveys in mathematics]_. springer - verlag , berlin , 2002 . d. v. voiculescu , k. j. dykema , and a. nica . , volume 1 of _ crm monograph series_. american mathematical society , providence , ri , 1992 . a noncommutative probability approach to free products with applications to random matrices , operator algebras and harmonic analysis on free groups .	we prove that the first continuous @xmath0-cohomology of free group factors vanishes . this answers a question by andreas thom regarding continuity properties of free difference quotients and shows that one can not distinguish free group factors by means of first continuous @xmath0-betti number .
testing the isotropy of the speed of light serves as a sensitive test of special relativity and lorentz invariance . the classic experiment to test the isotropy of the speed of light uses a michelson interferometer and was first performed by a.a . michelson in berlin ( 1880 ) and potsdam ( 1881 ) , germany . he was later joined by e.w . morley to perform an improved experiment in cleveland , ohio ( 1887 ) @xcite . modern michelson - morley type experiments use electromagnetic resonators to probe for lorentz invariance violations and are generally based on comparing the resonance frequencies of two similar orthogonal resonators while either actively rotating the setup or relying solely on earth s rotation @xcite . in case of a linear resonator , a relative frequency change is most generally described by @xmath1 , where @xmath2 denotes a relative change in the speed of light in vacuum along the optical path , @xmath3 denotes a relative change in the length of the optical path , and @xmath4 denotes a relative change in the index of refraction along the optical path . the magnitude of the different types of lorentz violations , all three of which can occur in the case of spontaneous lorentz symmetry breaking @xcite , depend on the composition of the material the resonator is made of . comparing the eigenfrequencies of two similar resonators made of the same material as has been done in all previous reported modern michelson - morley experiments makes it impossible to distinguish between the different types of lorentz violation @xcite and due to the substraction of the different types an overall lorentz violating signal could even be suppressed or canceled . however , the material dependency makes it possible to distinguish between the different types of lorentz violations by using dissimilar electromagnetic resonators . in the past , we have combined results of an experiment performed in our laboratory in berlin , germany , consisting of linear optical resonators made of fused silica with mirrors made of bk7 with the results of an experiment performed by stanwix in perth , australia , consisting of whispering gallery microwave resonators made of sapphire in order to give separate bounds on the different types of lorentz violations @xcite . we note that since the experiments have not been optimized for this kind of comparison and have not been synchronized timewise , not all in principle obtainable information of such a combined experiment could be resolved . in our new setup , we compare the eigenfrequencies of two orthogonal evacuated linear optical fused - silica resonators with the eigenfrequency of a monolithic linear optical sapphire resonator in which the light is propagating in matter . we thus can directly compare light propagation in vacuum and matter and search for possible violations of the isotropy of the speed of light , which indicate lorentz violations in matter . moreover , this setup enables us in principle to give independent , simultaneous bounds on a broad range of lorentz invariance violating coefficients in the maxwell and dirac sector in the framework of the standard - model extension ( sme ) @xcite . in what follows we will give an overview of our new setup containing a monolithic optical sapphire resonator and two crossed evacuated optical fused - silica cavities followed by the first results obtained . we will also discuss improvements of this setup by cooling down the sapphire resonator to lhe temperatures . in the end we will give a brief outlook on current improvements of this setup and the next generation of modern michelson - morley type experiments . we have realized a new type of a combined experiment in our laboratory in which we can compare the resonance frequency of a monolithic linear optical sapphire resonator with the resonance frequency of two evacuated linear optical cavities made of fused silica as used in our previous experiment @xcite while actively rotating all resonators in a michelson - morley configuration on an air bearing turntable once every 45 s. figure [ fig : schema ] shows a schematic of the sapphire monolithic resonator . sapphire is a uniaxial crystal ( @xmath5 ) . the resonator was fabricated such that the crystal symmetry axis ( c axis ) lies perpendicular to the resonator axis . the eigenpolarizations are therefore linearly polarized , parallel to the crystal axis with extraordinary index @xmath6 at 1064 nm , and perpendicular to it with ordinary index @xmath7 . the modes are split in frequency due to the birefringence . the plane and convex ( radius of curvature 1 m ) end faces of the standing - wave resonator were coated with high - reflection ( hr ) coatings for a center wavelength of 1064 nm . the base material was a high - purity sapphire single - crystal rod ( hemex ultra , crystal systems , salem , ma ) . the monolithic sapphire resonator features a finesse of about @xmath8 , corresponding to a linewidth of 200 khz . the round trip loss inside the resonator is on the order of 600 ppm , although the loss due to absorption should only be on the order of @xmath9 ppm / cm as measured by calorimetry . this leads to the conclusion that most of the losses are caused by flaw coatings . the incoupling efficiency of the monolithic sapphire resonator is less than @xmath10 . for thermal shielding the sapphire resonator is placed inside a cryostat which , however , we operated at room temperature and pumped to a pressure of @xmath11 . it offers optical free beam access through windows ( see fig . [ fig : schema ] ) . a nd : yag laser at 1064 nm is frequency stabilized to one of the tem@xmath12 modes of the resonator using a modified pound - drever - hall method in transmission ( modulation index of 3.75 and demodulating at third harmonic of the modulation frequency of 444 khz ) . the incoupling light is first sent through a polarization maintaining optical fiber for mode cleaning and 20 mw of light power are impinging on the resonator s front side , resulting in @xmath13 nw of transmitted light which is sensed for the pound - drever - hall lock using an avalanche photodiode . fractions of the laser beams of the laser stabilized to the sapphire resonator and the laser stabilized to one of the fused - silica resonators are overlapped on a fast photodiode to generate a beat note at the difference frequency @xmath14 . the long term frequency drift of the beat note is normally @xmath15 in marked contrast to the drift of @xmath16 of the beat note between the two lasers stabilized to the fused - silica cavities @xcite . this is due to the fact that there is no common mode rejection as in the case of the two crossed fused - silica cavities fabricated into one monolithic block . additionally , the eigenfrequencies of the monolithic sapphire resonator feature a relatively high fractional temperature dependence at room temperature of @xmath17 resulting from its coefficient of thermal expansion ( cte ) and the temperature dependance of the index of refraction at a wavelength of @xmath18 of sapphire in contrast to @xmath19 for the fused - silica cavities resulting only from its cte . + the frequency stability of the beat note is characterized by the allan deviation after removing the irrelevant long term drift by using the residuals of a linear plus quadratic fit ( see fig . [ fig : alavar ] ) . there seems to be a flicker floor of @xmath20 normalized to the laser frequency of 282 thz between 4 s and 100 s. we assume thermal noise within the monolithic sapphire to be the origin of this flicker floor , although this can not be affirmed by theory , as there is no complete theory on the thermal noise level of a monolithic optical resonator so far . however , we also cooled the monolithic sapphire resonator down to 4.2 k. the improvement in frequency stability at cryogenic temperatures make the assumption of thermal noise limiting effects reasonable ( see sec . [ sec : cryo ] ) . the vacuum chamber ( cryostat ) containing the monolithic sapphire resonator was placed on a breadboard containing all necessary optics . the breadboard itself was mounted to the rotating part of the existing setup above the vacuum chamber containing the crossed fused - silica resonators ( see fig . [ fig : schema ] ) and thus represents a second new level within the setup . the sapphire resonator axis is orientated parallel to one of the fused silica s resonator axis and thus orthogonal to the resonator axis of the other fused - silica cavity . except for these modifications are there no further changes of the existing setup and all measures implemented in our previous experiment @xcite to reduce systematics connected with active rotation also apply for the monolithic sapphire resonator . ten days of comparison of the resonance frequency of the monolithic sapphire resonator operated at room temperature with the one of the orthogonal orientated fused - silica cavity while actively rotating at a chosen period of 45 s were performed in march 2009 . this corresponds to more than @xmath21 turntable rotations . preparations for rotating measurements with the monolithic sapphire resonator cooled down to 4.2 k are currently under way ( see sec . [ sec : cryo ] ) . the analysis of the beat note with respect to anisotropy signals characterizing lorentz invariance violations follows the same procedure as in our previous experiment @xcite . no significant anisotropy signal was found fixed to a sidereal frame ( see fig . [ fig : results ] ) . using the obtained sidereal modulation amplitudes we can conclude an upper limit for the anisotropy of the relative difference of the speed of light in vacuum and matter ( sapphire ) of @xmath22 ( one standard deviation ) . a detailed analysis within the framework of the sme has not been done , since the dependence of the index of refraction of sapphire in the optical region on lorentz violating coefficients of the photonic and fermionic sector has not been completely worked out yet . however , mller @xcite has already described a recipe for deriving this dependency . further measurements with the monolithic sapphire resonator operated at room temperature have not been performed due to measurable residual systematic effects on the order of @xmath23 acting on its eigenfrequencies at once and twice the rotation frequency . however , we note , that this anisotropy signal was fixed to the laboratory frame and averaged out when analyzing it with respect to a sidereal frame ( see fig . [ fig : results ] ) , which is the one of interest within the framework of the sme . the origin for anisotropy signals fixed to the laboratory are generally constant gradients ( temperature , electromagnetic smog , gravity , etc . ) that modulate the beat note at twice the turntable rotation rate . the signals fixed to the laboratory in this case can be fully assigned to systematic effects influencing the eigenfrequencies of the monolithic sapphire resonator , since our previous experiments using the fused - silica cavities showed no significant signal @xcite . the high round - trip loss rate and the high relative temperature dependency of the eigenfrequencies of the monolithic sapphire resonator can most likely explain the residual systematic effects . for example , the sensitivity of the monolithic sapphire resonator towards translational displacements of the incoupling beam which causes a change of the circulating light power inside the resonator and thus its temperature has been measured to be @xmath24 at the timescale corresponding to twice the rotation rate . since the resonator is free beam coupled with no active pointing stabilization , it is possible that the residual systematic effects fixed in the laboratory are caused by small deformations of the breadboard carrying the optics for the monolithic sapphire resonator system , while the setup is turning . and @xmath25 are most prone to constant systematic effects ( note the different scale ) . the mean values and standard errors ( one sigma ) are @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , @xmath33 , @xmath34 , @xmath35 ( all values @xmath36).,title="fig:",scaledwidth=100.0% ] + before continuing the measurements , we decided to minimize the residual systematic effects , which is most easily done by lowering the sensitivity to temperature changes of the eigenfrequencies of the sapphire resonator . this can be achieved by cooling down the sapphire resonator to cryogenic temperatures , e.g. , 4.2 k. at these temperatures , the temperature sensitivity of the eigenfrequencies should decrease to @xmath37k as suggested by previous experiments using cryogenic sapphire resonators @xcite and thus pointing instabilities or laser light power fluctuations , etc . should have less impact on the overall measurement . to enable operation at cryogenic temperatures , we changed the mounting of the monolithic sapphire resonator in the cryostat which first was just used as a vacuum chamber only ( see sec . [ sec : extension ] ) . at cryogenic temperatures an improvement of more than one order of magnitude in frequency stability for the eigenfrequencies of the monolithic sapphire resonator can be seen in the allan deviation of the beat note ( see fig . [ fig : alavar ] ) . furthermore , a reduction in the beat note long term frequency drift down to 50 hz / s was observed . a comparison with a third independent frequency reference , namely an actively temperature stabilized ule cavity , reveled an even lower long term frequency drift for the monolithic sapphire resonator of less than 0.1 hz / s , where most of the drift can be assigned to some residual relaxation processes within the sapphire crystal due to the thermal cycling . preparations to perform rotating measurements with the monolithic sapphire resonator cooled down to lhe temperatures are currently under way . at cryogenic temperatures the measurement sensitivity will increase by more than one order of magnitude due to the better frequency stability of the sapphire resonator as compared to the room temperature case while the sensitivity to residual systematics will be lowered . we have set up a michelson - morley experiment which is capable of comparing light propagation in matter and vacuum . the derived limit for an anisotropy of the speed of light in vacuum in comparison with light propagation in matter ( sapphire ) is @xmath38 ( preliminary ) . the measurement is currently limited by the frequency stability of the monolithic sapphire resonator operated at room temperature . rotating measurements with the sapphire resonator cooled to lhe temperatures will soon be performed . we are also planing the next generation of a modern michelson - morley experiment , in which we will overcome the momentarily limiting thermal noise limit @xcite by employing cryogenically cooled linear evacuated optical sapphire cavities . with these new cavities we should increase our measurement sensitivity for lorentz invariance violations by about three orders of magnitude as compared to our previous experiment @xcite and thus we will be able to probe the anisotropy of the speed of light in vacuum in the @xmath39 regime . the improvement in measurement sensitivity will result from the potentially more than two orders of magnitude better frequency stability of the new sapphire cavities as compared to the fused - silica cavities used in our previous experiment @xcite . moreover , we will combine our new experimental setup with cryogenic whispering gallery microwave resonators , which will be delivered and operated by the university of western australia @xcite , where light is propagating in matter and that feature a frequency stability in the mid @xmath40 region . with the proposed co - rotating microwave and optical resonator setup we will be able to probe for possible lorentz violations that are either frequency dependent ( microwave vs. optical ) , rely on the electromagnetic mode ( whispering gallery vs. linear ) , and/or depend on the medium the light is propagating in ( matter vs. vacuum ) . we thus will also be able to set first limits on a broad range of lorentz violating coefficients within the framework of the recently formulated extended sme @xcite .	we present a michelson - morley type experiment for testing the isotropy of the speed of light in vacuum and matter . the experiment compares the resonance frequency of a monolithic optical sapphire resonator with the resonance frequency of an orthogonal evacuated optical cavity made of fused silica while the whole setup is rotated on an air bearing turntable once every 45 s. preliminary results yield an upper limit for the anisotropy of the speed of light in matter ( sapphire ) of @xmath0 , limited by the frequency stability of the sapphire resonator operated at room temperature . work to increase the measurement sensitivity by more than one order of magnitude by cooling down the sapphire resonator to liquid helium temperatures ( lhe ) is currently under way .
searching and staying up - to - date with scholarly literature is an essential part of scientific research . with the advent of the world - wide web ( www ) and the evolution of electronic publishing , a powerful environment was created to open the vast universe of scientific literature on a world - wide scale . in early 1994 the www had become sophisticated enough to allow the search of electronic resources via `` web forms '' . over the past decade , this environment matured into an unavoidable and indispensable fact of life and tools have emerged in it that have become a crucial ingredient in scientific research . being able to search vast amounts of data electronically obviously facilitates the review process , but applying advanced technologies such as pattern recognition to the electronic data or by allowing nested searches , one is able to produce results that are unattainable in conventional , non - electronic ways . in that sense one can argue that the available electronic search tools on the www even further scientific research . using a straight - forward search engine ( google , yahoo , msn , altavista , ... ) results in thousands of documents , ranked by some sophisticated algorithm . even with the advanced versions of these tools , we still find ourselves awash in information . to search the electronic , scholarly literature , scientists need to be able to zoom in on bibliographic data using additional descriptors and search logic . what scholarly tools are available for specialists in astronomy and ( astro)physics ? the principal bibliographic services are the nasa astrophysics data system ( nasa ads ) , google scholar , inspec and the astronomy and astrophysics abstracts . important additional resources are the science citation index ( isi web of science ) , scopus and zetoc . although these tools allow researchers to zoom in on the scholarly literature , they do not offer additional tools to determine the most popular or most cited papers in a given subject . especially for staying up - to - date , it is essential to be notified of the most popular and most cited papers . late 2003 , the ads introduced the _ myads _ service @xcite , a fully customizable newspaper covering ( journal ) research for astronomy , physics and/or the arxiv e - prints . this service will give the user an overview of the most recent papers by his / her favorite authors , and the most recent , most cited and most popular papers in a particular subject area . additionally , the user will see an overview of citations to his / her papers . between a fifth and a quarter of all working astronomers already subscribe to myads @xcite . _ myads - arxiv _ is a fully customizable , open access , virtual journal , covering the most important papers of the past week in physics and astronomy . in other words , for the specialist , the myads - arxiv service provides a one stop shop for staying up - to - date in his / her field of interest . myads - arxiv is based on the existing services of the nasa astrophysics data system ( see @xcite and wikipedia ) and the arxiv e - print repository ( see @xcite @xcite and @xcite ) . the ads repository is completely synchronized with the arxiv e - prints system . each night ads bibliographic records are created for all the e - prints that were newly added . the references are also extracted from the e - prints and matched against existing records in the ads . thus we add lists of references to the bibliographic records and we use these references to maintain citation statistics . both of these elements are used in the myads - arxiv service . the service also uses readership information from both the ads and the arxiv to compute the most popular papers in a subject area . by this continuous influx of information , myads - arxiv provides the subscriber with a service that is as up - to - date as your morning newspaper . the service provides a weekly overview , which offers the unique view on what is happening in a field , and a daily notification . why is this view unique ? it is * fully customizable * because you specify the queries that determine the results . myads - arxiv is an * open access * service , because it is _ totally free _ ( no subscription costs ) . furthermore , it is a * virtual journal * , because the overview is a regularly appearing collection of scholarly papers , that are only available in electronic format ( the vast majority of e - prints have not yet been published as journal papers ) . last but not least , myads - arxiv covers the * most important * papers of the past week . this follows from concordance between e - printed and published papers , and citation statistics : in astronomy and physics , the most important papers are submitted as e - prints first @xcite . the url for the service is * http://myads.harvard.edu*. from here , you can set up your account and specify the queries for the myads - arxiv service . a maximum of 2 subject queries and 1 author query can be specified . the results page is like an an automatically generated newsletter . for each of your subject queries , you will get an overview of the `` recent '' , `` most popular '' and `` most cited '' papers . the `` recent '' papers are the newly ( i.e. since the previous query ) added entries in the e - print database that match your query . the `` most popular '' papers are found by looking at the also - read statistics for the top 100 of all papers that match your query and the `` most cited '' papers are obtained from the reference lists of the papers that match your query , published in the previous three months . the link to this newsletter is public , so it can be shared with colleagues . there is also a daily alerting service , showing you the latest e - prints in the categories of your choice , sorted according to a query specified by you ( or e - print number if you did not specify one ) . entries that match you query are preceded by an asterisk . why is myads - arxiv so unique and powerful ? it is because of the combination of the following two factors : the search capabilities of the myads - arxiv service guarantees the proper selection of bibliographic records for your queries , and the quality of the e - prints guarantees the relevance and importance of the bibliographic records . the unique machinery that powers the queries consists of _ reference resolving _ ( associating reference strings with existing bibliographic records , see @xcite ) , _ bibcode matching _ ( setting up the e - print / paper concordance ) , _ second order operators _ ( operations on results lists ) and _ also - read statistics _ ( `` people who read paper x , also read paper y '' , see @xcite ) . by associating reference lists in newly added papers with already existing records , we keep citation statistics up - to - date , which allows us to generate lists of `` most cited papers '' . with the reads statistics , we can construct a list of `` most popular papers '' , using the second order operators . the bibcode matching procedure is only relevant in cases where the preprint already appeared in a journal . titles , abstracts and author lists are indexed , so that they are available for searching . the quality of the e - prints is reflected in the observation : the most important papers in astronomy and physics appear as e - prints on the arxiv . this fact is illustrated by figure 1 ( left ) , which shows for a number of important astronomy and physics journals , the fraction of e - printed papers for the top 100 most cited papers , during the period of 1992 through 2005 . over 90% appears as e - print first ( for month . notices ras and nucl . b , it is even 100% ) . just the effect of primacy through early access is not enough to explain the fraction of e - printed papers in the top 100 most cited papers ( at a given moment ) . according to @xcite there is an effect called `` self - selection bias '' that results in a further increase of citation rates for e - printed papers . `` because papers in the arxiv are not refereed ... this suggests that authors self - censor or self - promote , or that for some reason the most citable authors are also those who first use the new publication venue '' @xcite . the papers are not cited more because they are read more , they are cited more and read more because of their quality . figure 1 ( right ) illustrates the fact that e - printed papers are cited and read more than papers that did not appear as e - print . the rule `` better searches give better results '' most definitely applies to the myads - arxiv service . based on the sophisticated search capabilities of the ads , myads - arxiv will provide the most powerful results for those who are able to characterize a field in a couple of keywords and key phrases . additionally , the use of `` simple logic '' allows a user to really zoom in on a specific research field . the myads - arxiv service is unique in the world of electronic libraries and publishing . no other electronic newsletter or alerting service produces a view on the scholarly physics and astronomy literaturei as comprehensive as this service . from the harvard - smithsonian center for astrophysics press release in spaceref.ca ( article number 16658 ) on the myads - arxiv service : `` it s the best thing since two pieces of sliced bread were assembled to make a sandwich , '' said paul ginsparg , professor of physics and information science at cornell university ( april 17 , 2005 ) . ginsparg , p. ( 2001 ) . `` creating a global knowledge network '' , in electronic publishing in science ii . proceedings of joint icsu press / unesco conference , paris , france . available at : http://arxiv.org/blurb/pg01unesco.html kurtz , michael j. , eichhorn , guenther , accomazzi , alberto , grant , carolyn , demleitner , markus , henneken , edwin , murray , stephen s. ( 2005b ) . `` the effect of use and access on citations '' , information processing and management , vol . 41 , issue 6 , p. 1395 - 1402	the myads - arxiv service provides the scientific community with a one stop shop for staying up - to - date with a researcher s field of interest . the service provides a powerful and unique filter on the enormous amount of bibliographic information added to the ads on a daily basis . it also provides a complete view with the most relevant papers available in the subscriber s field of interest . with this service , the subscriber will get to know the lastest developments , popular trends and the most important papers . this makes the service not only unique from a technical point of view , but also from a content point of view . on this poster we will argue why myads - arxiv is a tailor - made , open access , virtual journal and we will illustrate its unique character . the ads is funded by nasa grant nng06gg68 g .
the first 500 myr after the big bang mark the current frontier in our exploration of cosmic history . understanding when and how the first galaxies started to form , how they grew their stellar mass and eventually turned into the diverse population of galaxies we see today is one of the most intriguing and challenging questions of modern observational astronomy . this is the main science driver for the director s discretionary time @xmath7 frontier field program ( hff ; e.g. * ? ? ? the hff will make use of lensing magnification of 4 - 6 foreground clusters to probe the ultra - faint galaxy population as early as 400 - 500 myr after the big bang . furthermore , the hff additionally creates six deep parallel blank field pointings in order to mitigate the uncertainties of lensing magnification and cosmic variance . while great progress has been made recently in probing galaxy build - up out to @xmath8 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , beyond @xmath1 , our understanding of galaxies is still very limited due to small number statistics . consequently the evolution of the cosmic star - formation rate density from @xmath1 to @xmath2 is still uncertain . the analysis of the full hubble ultra - deep field 09/12 ( hudf09/12 ) data and of two fields from the cosmic assembly near - infrared deep extragalactic legacy survey ( candels ) revealed a rapid decline of the sfrd by @xmath9 in only 170 myr from @xmath1 to @xmath2 ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * but see also ellis et al . the two detections of @xmath10 galaxies in the cluster lensing and supernova survey with hubble ( clash ; * ? ? ? * ; * ? ? ? * ) have not changed this broad picture of a steeper decline compared to lower redshift trends . by adding up to twelve additional very deep sightlines , the hff program will be the prime dataset to clarify the sfrd evolution at @xmath11 before the advent of the james webb space telescope _ ( jwst)_. furthermore , given the power of lensing clusters ( see * ? ? ? * ) , the hff program will also provide a unique dataset to study resolved morphologies of very high - redshift , multiply imaged galaxies ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and will likely result in galaxy samples bright enough for spectroscopy ( e.g. * ? ? ? * ; * ? ? ? * ) . it may even be possible to probe the faint - end cutoff of the high - redshift ultra - violet ( uv ) luminosity functions with the hff dataset once all observational uncertainties and biases are under control @xcite . results on @xmath12 galaxies have been reported using partial hff data from the first observing epochs ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and very recently also from the full dataset of a2744 @xcite . the majority of these analyses to date have been limited , however , to the presentation of possible candidates only . the recent analysis of the complete dataset over abell 2744 by @xcite provided the first multiply imaged @xmath2 galaxy candidate identified from the hff program ( see also * ? ? ? the candidate jd1 is found to be a triply imaged source with an intrinsic apparent magnitude of only @xmath1329.9 mag , comparably faint as the previous @xmath2 galaxies identified in the deepest data over the hudf @xcite . the locations of all three multiple images of jd1 are consistent with the prediction of the cluster lensing maps for a @xmath14 source , which significantly decreases the likelihood of this source being a lower redshift contaminant . in this paper we make use of the complete hff dataset of the first cluster , abell 2744 , and its parallel field in order to search for additional @xmath2 galaxy candidates and to derive the first constraints on the star - formation rate density of @xmath2 galaxies based on hff data . in particular , we will discuss the effect of shear- and position - dependent completeness for high - redshift galaxy catalogs . this proves to be very important , yet has been largely overlooked so far . this paper is organized as follows : in section [ sec : data ] , we describe the dataset and sample selection . a detailed description of our completeness simulations and how shear affects the selection volume of galaxies is given in section [ sec : completeness ] . our results on the @xmath2 star - formation rate densities are presented in section [ sec : results ] , before summarizing in section [ sec : summary ] . throughout this paper , we adopt @xmath15 kms@xmath16mpc@xmath16 , i.e. @xmath17 , consistent with the most recent measurements from planck @xcite . magnitudes are given in the ab system @xcite , and we will refer to the hst filters f435w , f606w , f814w , f105w , f125w , f140w , f160w as @xmath18 , @xmath19 , @xmath20 , @xmath21 , @xmath22 , @xmath23 , @xmath24 , respectively . the hff program images each cluster / blank field for 140 orbits split over seven filters with the acs and wfc3/ir cameras . these filters are @xmath18 , @xmath19 , @xmath20 , @xmath21 , @xmath22 , @xmath23 , and @xmath24 . in this paper , we use the fully reduced version 1 hff dataset of abell 2744 and its parallel field provided by stsci . ] these images were calibrated , cosmic - ray cleaned , background corrected , astrometrically aligned , and drizzled to the same output frames . in particular , we use the images drizzled at 60 mas pixel scale . the final mosaics provided by stsci also include all ancillary data available over these fields in the same filters from additional programs . of particular importance is the frontier field uv imaging program ( go13389 , pi : siana ) which adds 16 orbits of acs data over the parallel field ( split over @xmath18 and @xmath19 ) . for the cluster field , we create a weighted combination of the individually provided epoch 1 and 2 acs images using the weightmaps , which adds the pre - existing data over this cluster ( go11689 , pi : dupke ) . the final @xmath25 depth of the images in empty regions of sky is @xmath26 mag as measured in circular apertures of 04 diameter . for more detailed information on these data , see koekemoer et al . ( 2014 , in preparation ) and visit the frontier field webpage at stsci . galactic extinction is accounted for by adjusting zeropoints for each hst filter using a milky way extinction curve @xcite and @xmath27 @xcite . this only results in minor corrections of @xmath28 mag in the wfc3/ir filters and up to 0.05 mag in the @xmath18 filter . gravitational lens models for all hff clusters were produced by five teams using different methods . these are made available through the frontier field webpage on mast . it is important to note that all these models are only based on ancillary data taken before the hff campaign , and they are expected to improve and converge with the additional constraints from the many faint multiple images found in the hff data @xcite . for details on the models see , e.g. , @xcite . here we use the five models that also released both components of the shear tensor which allow us to compute the radial and tangential magnification factors in order to be able to properly estimate the selection volume of high redshift galaxies ( see section [ sec : shear ] for more details ) . this includes the models of bradac et al . ( e.g. , * ? ? ? * ) , merten et al . ( e.g. , * ? ? ? * ) , zitrin et al . ( e.g. , * ? ? ? * ) , and williams et al . ( e.g. , * ? ? ? * ) . the results shown in the remainder of this paper are based on the lensing map provided by zitrin et al . ( zitrin - nfw ) for a2744 . however , our results on the overall number densities of @xmath2 galaxies do not change significantly when considering other magnification maps , consistent with the findings of @xcite . one significant concern with the data obtained over the cluster fields is the intra - cluster light ( icl ) which significantly increases the background and limits the detectability of faint galaxies . the brightness of the icl lies @xmath29 mag above the surface brightness limit of the hff data over a large part of the cluster field ( see * ? ? ? * ) and thus significantly limits the direct detectability of faint sources . furthermore , for abell 2744 the critical curve for lensing @xmath2 galaxies runs partially through the icl , which may significantly reduce the chance of finding highly magnified galaxies in standard sextractor catalogs due to blending with the icl and spurious detections . in order to mitigate some of this effect , we subtract the icl using a 25 wide median filter . when filtering , we exclude the cores of bright sources in order to minimize over - subtraction around bright galaxies or stars . the median subtracted images are then fed to sextractor @xcite to produce source catalogs using standard parameters . we found this procedure to result in somewhat more reliable catalogs and flux estimates for faint , small sources compared to running the standard sextractor background subtraction on the original images . however , in future analyses , it may be possible to improve upon our treatment of the icl using more sophisticated modeling and subtraction accounting for the icl and bright cluster galaxies simultaneously . this may likely result in even more complete catalogs of high - redshift sources toward the cluster center . whatever method is used , however , both the cluster galaxies themselves and the increased background due to the icl result in reduced search volumes of high redshift galaxies in cluster images ( see later section [ sec : completeness ] and figure [ fig : posdepcompleteness ] ) . similar to previous selections , we identified galaxies at @xmath30 by exploiting the spectral break shortward of ly@xmath31 due to inter - galactic hydrogen . red @xmath32 colors and non - detection in shorter wavelength filters are the key features used in the selection . in order to directly compare the hff sample with previous analyses ( e.g. * ? ? ? * ) , we restrict the search here to galaxies with @xmath4 , which selects sources at @xmath33 . we identified sources in a @xmath34 image constructed from the @xmath24 and @xmath23 images and measured photometry with sextractor run in dual image mode . all images were psf - matched to the @xmath24 point - spread function . colors were measured in small kron apertures ( kron factor 1.2 ) , typically 02 radius and total magnitudes were derived from larger elliptical apertures using the standard kron factor of 2.5 , with an additional correction to total fluxes based on the encircled flux measurements of stars in the @xmath24 band . based on these catalogs , we applied the same selections as we used previously in @xcite : @xmath35 @xmath36 furthermore , sources were required to be detected in @xmath24 and @xmath23 with @xmath37 in each and at least @xmath38 in one of the bands . the @xmath39 for each candidate source was computed following @xcite as @xmath40 , with @xmath41 the flux in band @xmath42 and @xmath43 the associated uncertainty . sgn(@xmath41 ) is equal to 1 if @xmath44 and @xmath45 if @xmath46 , and the summation is over the @xmath18 , @xmath19 , @xmath20 , and @xmath21 bands . the limit of @xmath47 efficiently excludes lower redshift contaminants while only reducing the selection volume by a small amount ( 20% ; see also * ? ? ? * ; * ? ? ? * ) . , @xmath22 , @xmath23 , and @xmath24 . as can be clearly seen , both source positions are affected by diffraction spikes from a nearby star and a bright cluster galaxy . to provide the most reliable estimates , we perform manual flux measurements . . ] applying the above selection criteria to the publicly released hff data of a2744 , we identify two candidates , a2744-jd1a and a2744-jd1b . these are two images of a single , triply imaged source , independently discovered earlier by ( * ? ? ? * see also ishigaki et al . the third image ( jd1c in zitrin et al . ) lies very close to a bright foreground source and is not present in our catalogs despite aggressive deblending parameters used in our sextractor runs . visual inspection indicates that it is a viable source . however , it is not included in the rest of this paper , as our effective volume simulations account for sources lost due to photometric scatter or blending with neighbors , as is the case for jd1c . the two detected images of jd1 lie on either side of the @xmath2 critical curve at ( ra , dec)@xmath48 ( 00:14:22.20 , @xmath4930:24:05.3 ) and ( 00:14:22.80 , @xmath4930:24:02.8 ) and are shown in figure [ fig : stamps ] ( see also fig 1 in * ? ? ? * ) . for both sources the photometry is heavily affected by diffraction spikes , in the first case caused by a nearby bright star and for the second source by a bright galaxy . nevertheless , it is clear that both sources are real . while both images satisfy the color selection criteria in our standard catalog , the sextractor photometry is likely unreliable due to the diffraction spikes . we therefore performed manual aperture photometry to confirm the color measurements and total magnitudes . in particular , in our manual measurement , we estimate the sky value in a small annulus around the source , excluding pixels which are obviously affected by the nearby diffraction spikes . photometry is then measured in small , circular apertures of increasing size up to 06 diameter . finally , we use the encircled energy of stars to correct the fluxes to total as given in the wfc3 handbook @xcite . using this approach , we find total magnitudes of @xmath50 and @xmath51 for the two images jd1a and jd1b , respectively , and colors @xmath52 and @xmath53 . these measurements are consistent with the photometry from @xcite where these images are discussed in detail and the source s photometric redshift is determined to be @xmath54 . this is confirmed by the lensing geometry which satisfactorily predicts the location of the three images only if the source lies at @xmath11 . the magnification of the two images is @xmath9 as predicted by @xcite . however , the full range of allowed magnification factors also predicted by other frontier field lensing models is @xmath55 . while uncertain , this source is likely to be of comparable brightness to the faintest @xmath2 candidates found in the xdf / hudf12 dataset @xcite , i.e. , the @xmath3 source xdfj-38126243 @xcite or the @xmath56 candidate udf12 - 4265 - 7049 @xcite . the same search for galaxies with @xmath4 in the parallel field of the hff cluster a2744 did not result in any candidate @xmath2 galaxy . while we do find a number of high - quality sources with colors within @xmath57 of this cut , these galaxies most likely lie at slightly lower redshift @xmath58 and will be discussed in a future paper . our selection function simulations which we discuss in the next sections do statistically account for sources lost due to photometric scatter , and we therefore proceed with zero @xmath2 galaxy candidates from the parallel field of a2744 . for @xmath2 galaxies at fixed apparent magnitude behind abell 2744 ( blue lines ) . the critical curve for lensing @xmath2 galaxies based on the zitrin - nfw model is shown in orange . the magnitude distribution of simulated galaxies is assumed to be flat for @xmath59 mag and the completeness is normalized to areas of @xmath60 ( 15% of the image ) , where the absolute completeness is @xmath61 . the relative completeness over much of the cluster center is significantly reduced due to the increased background . however , lower completeness is also found around the critical curve even in the absence of bright foreground sources . this is due to the sheared morphologies of galaxies . this effect has been largely ignored in lf analyses behind lensing clusters so far . it may be possible to increase the source completeness with the use of more sophisticated modeling and subtraction of the intra - cluster light and bright foreground galaxies as well as with the adoption of a detection smoothing kernel adapted to the expected shear at a given location within the image . nonetheless , it will not result in the same completeness levels expected in ultra - deep blank fields . ] in the next sections we show the importance of position - dependent source blending and shear on the completeness and selection efficiency of highly - magnified , high - redshift candidates . while faint @xmath2 galaxies are only marginally resolved with @xmath62 , they do have a finite size of @xmath63 kpc and are not point sources ( e.g. * ? ? ? * ; * ? ? ? this has important implications for the completeness of galaxy selections around the critical curves of lensing clusters . the limited surface brightness sensitivity of hst leads to a significant reduction of the selection efficiency for the most highly magnified and sheared sources ( see e.g. , * ? ? ? this effect has so far been largely ignored in previous determinations of selection volumes and uv lfs behind lensing clusters due to the computational challenges involved ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , with few exceptions . for instance , @xcite discuss how the signal - to - noise boost from lensing depends on a galaxy s two dimensional profile due to shear . @xcite used completeness simulations which account for the mapping of galaxy images from the source plane to the image plane using the lens models . this has recently also been incorporated in new determinations of the uv lfs behind the hff clusters ( e.g. * ? ? ? * ; * ? ? ? as we show below it is crucial to include lensing shear and magnification - dependent completeness when deriving lensed number densities of galaxies . we do this by taking shear into account to first order using the shear tensor to compute the tangential and radial magnification as well as the direction of the shear angle ( see e.g. * ? ? ? we therefore rely on hff lensing models which provide the two components @xmath64 and @xmath65 of the shear tensor . these are second order derivatives of the lensing potential @xmath66 , such that @xmath67 , @xmath68 ( see e.g. * ? ? ? * for a proper derivation ) . the results in the remainder of this paper are all based on the zitrin - nfw model . using different models has no significant impact on the overall number densities of @xmath2 sources , even though the predicted magnification for individual sources can show a wide range ( see also * ? ? ? based on the shear tensor , we derive the shear angle @xmath69 at each location in the image , as well as the tangential ( @xmath70 ) and radial shear factors ( @xmath71 ) , which can be computed based on the convergence @xmath72 and shear @xmath73 maps provided by the lens models : @xmath74 using these three quantities , we can estimate the effect of high magnifications on the selection function and completeness of galaxies . we follow standard procedures for blank fields and insert artificial galaxies with different light profiles , sizes , luminosities , and redshifts into the original science images . after re - running our detection algorithm with the same parameters as for the original images , the completeness @xmath75 is simply given by the fraction of sources that are detected and observed at magnitude @xmath76 . the only difference compared to non - cluster field completeness simulations is that we apply the position - dependent shear to the artificial galaxies before inserting these into the images . for computational efficiency we limit the tangential and radial shear factors to @xmath77 . our estimates therefore become unreliable above @xmath78 ( where the completeness is overestimated ; see section 3.3 ) . this only affects a small fraction of the image plane , however ( @xmath135% ; see also * ? ? ? the resulting completeness and effective selection volumes depend on the assumed properties of the simulated galaxy population . in particular , the galaxy size and morphological profile distributions as well as the intrinsic color distributions are important parameters of such simulations ( see e.g. * ? ? ? * ) . here , the color distributions are set according to the luminosity dependent distribution of uv continuum slopes as measured by bouwens et al . ( 2013 ; see also * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? in order to account for non - regular morphologies observed in star - forming galaxies at high redshift , lbg completeness simulations often rely on scaling actual observed lbgs ( e.g. , at @xmath79 ) and redshifting them accounting for the difference in cosmological angular diameter distance and for intrinsic size evolution . in the case of simulations where lensing magnification is taken into account such redshifting is not possible due to the insufficient resolution of actual @xmath79 galaxy observations with hst to reliably reproduce higher redshift , highly magnified sources . we therefore use another common approach adopting idealized galaxy light profiles . in practice , we use a 50% mix of exponential disks and devaucouleur profiles ( corresponding to sersic profiles with @xmath80 or @xmath81 respectively ; sersic 1968 ) . as shown later , this assumption has a small , but noticeable effect on the resulting completeness . the intrinsic size distribution of galaxies is chosen according to a log - normal distribution with mean evolving as @xmath82 as is consistent with most studies of lbg size evolution at @xmath83 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? in particular , the size distributions are normalized at @xmath84 where we set the peak of the distribution at @xmath85 kpc and we assume a constant width of @xmath86 ( consistent with the distribution of halo spin parameters ) . the simulated light profiles are then sheared according to the lens model at the position where they are inserted , before convolving them with the wfc3/ir psf . since we are simulating the light profiles of magnified sources , it is important to also account for any trend in size with mass or luminosity . smaller galaxy sizes at lower luminosities are sometimes used to argue that lensing shear has no effect on galaxy completeness ( e.g. * ? ? ? however , both the mass - size and the luminosity - size relations at high redshift are found to be very shallow following @xmath87 as measured for @xmath88 lbgs by @xcite similar to the luminosity scaling @xmath89 found by @xcite . these measurements are completely consistent with the surprisingly constant size scaling for late type galaxies at all redshifts @xmath90 seen in the candels dataset ( @xmath91 ; see * ? ? ? * ) and there is no convincing evidence for a change in these scaling relations at higher redshifts ( but see * ? ? ? this suggests that a galaxy magnified by a factor @xmath92 is _ intrinsically _ only @xmath93 smaller than a non - lensed galaxy observed in the field . in order to account for this size scaling in our shear simulations over the cluster field , we scale our assumed size distribution from the blank field by @xmath94 before inserting galaxies in the image . this thus corresponds to an assumed scaling of the size distribution of @xmath95 . ) corresponding to just 15% of the area of the image in this cluster , where the completeness is about 80% . the values are computed for galaxies at _ fixed apparent _ magnitude @xmath59 under the assumption that galaxy sizes scale as @xmath96 ( dark blue dots and line ) . calculations are only shown up to @xmath97 , above which our estimates start to become unreliable ( overestimated ) due to the use of 3wide bins to compute the position - dependent completeness and due to our limiting the shear factors for computational efficiency . a representative gray errorbar on the left shows the 1@xmath98 dispersion in the relation across the image . while lensing preserves surface brightness , highly sheared sources are spread out over many pixels resulting in a lower detection probability . the assumed size distribution has thus a significant impact on the expected completeness as shown by the gray dashed line where no size scaling with luminosity was assumed . even in the absence of shear , however , magnified sources have reduced completeness due to blending with foreground galaxies and intra - cluster light as shown with the dashed red line . the faint blue lines show the impact of different galaxy profiles . the upper line shows the result using only @xmath81 sersic profiles , while the lower curve is for exponential discs with @xmath80 . ] for @xmath2 galaxies as a function of magnification factor @xmath99 and @xmath24-band magnitude behind abell 2744 . the different curves show different observed @xmath24 magnitude bins with decreasing luminosity ( bright to faint ) . the colors represent different light profiles ( @xmath80 blue , @xmath81 red ) . as seen in the previous figure , the impact of the assumption on the light profile is noticeable , but relatively small . this figure demonstrates that the reduced relative completeness for highly magnified galaxies with @xmath100 mag is mostly dominated by blending with foreground galaxies , while shear impacts sources at @xmath101 mag , i.e. about 1.5 mag from the nominal completeness limit of the data . ] figure [ fig : posdepcompleteness ] shows the relative detection completeness for galaxies in the observed magnitude range @xmath59 mag as a function of position in the cluster field . the completeness is normalized to the median found in areas of the image with @xmath60 ( 15% of the image ) , where the absolute completeness is @xmath61 . while the relative completeness decreases significantly around brighter sources in the field due to blending , it is clearly apparent that the completeness is also reduced around the critical curve of the cluster where no bright foreground sources are present . in those areas of the image , the main reason for the reduced completeness is shear and magnification . even though lensing conserves surface brightness , a source which is highly magnified above the survey detection limit is spread over many more pixels than a non - sheared source at the same observed magnitude , reducing its s / n and detection probability . the relative completeness as a function of magnification averaged over the whole cluster field is shown in figure [ fig : magdepcompleteness ] . as expected , we find a significant decrease in completeness toward higher magnification . even though the scatter is significant in this relation , a source magnified by @xmath102 has on average a @xmath103 lower chance of being detected and included in a high - redshift catalog compared to a source which is only magnified by @xmath60 . however , as already pointed out , magnification - dependent completeness is present even when ignoring shear and magnification , simply due to blending with bright cluster galaxies closer to the critical curves ( red dashed line in fig . [ fig : magdepcompleteness ] ) . the shear adds to the incompleteness on top of this by a factor @xmath104 . figure [ fig : magdepcompleteness ] also shows that the size scaling does have a significant impact on the derived completeness relation . using our default scaling of @xmath95 , we find that very highly magnified sources are up to a factor @xmath105 more complete than assuming no size scaling at all . the large discrepancy between these two estimates , however , shows that accurate size scaling relations are necessary to accurately compute the selection volumes of high - redshift galaxies , adding to the uncertainties in lfs estimated from cluster fields . one possible strategy for mitigating these uncertainties is to use a differential technique to derive the relative normalization of the uv lf in various redshift bins from their relative surface densities ( see e.g. , * ? ? ? this is based on the assumption that the sizes and surface brightness profiles of galaxies in different redshift intervals are largely self - similar vs. luminosity . however , these assumptions in deriving the lf evolution need to be properly tested and calibrated . figure [ fig : magdepcompleteness ] also shows the impact of our assumption on the distribution of galaxy light profiles . if we simulate only sersic @xmath80 exponential disk profiles , the resulting completeness is slightly reduced ( by @xmath106 at @xmath92 ) compared to the steeper @xmath81 light profile . clearly , the impact of a size - luminosity relation is thus significantly more important , and it will be crucial to accurately determine this with future observations . finally , figure [ fig : completenessfmag ] shows the magnification - dependent completeness as a function of observed magnitude and galaxy profile . this demonstrates that the drop in completeness at high magnification seen in figure [ fig : magdepcompleteness ] is mostly driven by sources with @xmath107 mag , i.e. @xmath131.5 mag brighter than our 5@xmath98 detection limit . for sources with observed magnitudes brighter than this , the reduced completeness at higher magnification is mostly dominated by blending with the bright foreground cluster galaxies . we stress that the completeness estimates derived here only apply to galaxy catalogs using standard source detection algorithms . it may be possible to increase the source completeness around the critical curves with the use of a smoothing kernel adapted to the expected shear . furthermore , our calculations assume idealized light profiles . clumpy substructure in galaxies may further increase their detection probability . quantifying these effects is beyond the scope of this paper , however . nevertheless , whatever detection algorithm is used , it is clear that accounting for a positional dependence of the completeness is crucial for any luminosity function or star - formation rate density analysis behind lensing clusters , which has so far been largely overlooked in the literature . candidate images in the a2744 hff data assuming the @xmath2 uv lf from @xcite . _ top _ differential number counts per wfc3/ir field - of - view and per unit magnitude _ bottom _ cumulative number counts . dark red lines correspond to the cluster field ( assuming the magnification map zitrin - nfw ) and the dark gray lines show the parallel blank field predictions . the thick solid lines are the expected numbers based on full simulations of the selection efficiency and completeness , which also include shear for the cluster field . the dot - dashed lines are idealized predictions assuming a selection efficiency of @xmath108 and a selection volume of @xmath109 for comparison . the dashed line for the cluster field shows the prediction when ignoring the position - dependent completeness and assuming no shear in the simulations ( as is often done in the literature ) . clearly , the magnification - dependent completeness results in a significant reduction of expected images at faint magnitudes in the cluster field . the total number of expected images is comparable in both fields , in contrast to that predicted with idealized assumptions . the absence of @xmath2 galaxy candidates in the parallel field is indicated by the 1@xmath98 upper limit in dark gray in the top panel , while the detection of two images ( of the same galaxy ) in the cluster field is shown as dark red square . ] galaxies as a function of _ intrinsic _ , de - lensed magnitude . as in the previous figure , dark red lines show the cluster field and dark gray lines correspond to the blank field . the thick solid lines are the expected numbers based on full simulations of the selection efficiency and completeness , which also include shear for the cluster field . by comparison to the dashed line ( corresponding to a simulation without shear ) it is clear that the reduced completeness due to lensing shear and foreground blending at high magnification significantly reduces the search volume at @xmath110 mag , which somewhat limits the power of clusters to probe the ultra - faint galaxy population . ] clearly , the reduced completeness at high magnification also somewhat reduces the power of lensing clusters to probe deeper down the lf than ultra - deep blank fields . in this section , we estimate how this affects the expected number of @xmath2 galaxy candidates in the a2744 hff data set . using the previous simulations we compute the magnification - dependent selection efficiency , @xmath111 . this is given by the fraction of inserted galaxies at magnification @xmath99 with redshift @xmath112 and observed magnitude @xmath76 , which are both detected and satisfy our color selection criteria . this therefore accounts both for completeness at a given observed magnitude , as well as for photometric scatter which statistically removes galaxies from our lbg color selection box . using this selection function , we can compute the number of expected galaxy images ( double - counting multiple images ) in bins of magnitude for a given uv lf @xmath69 from @xmath113 where @xmath114 is the observed , magnified luminosity of a source , @xmath99 is the magnification and @xmath115 is the image solid angle ( i.e. observed pixels ) which is magnified by @xmath99 . @xmath116 is the cosmological volume per unit solid angle and redshift . the same equation also holds for blank fields , where @xmath117 everywhere . from the above equation it is clear that ignoring the position - dependent completeness and the reduction of selection efficiency due to shear in cluster fields typically results in higher expected numbers than may actually be present for a given lf this is demonstrated in figure [ fig : nexpz10 ] , where we show the number of expected galaxy images for both the cluster field and the parallel blank field of a2744 . the lf for this figure is taken from the analysis of the full candels - deep and xdf / hudf09 dataset . this is still uncertain due to the small number of candidates and @xcite therefore derive two possible @xmath2 uv lfs based on the previous data , one in which they assume evolution in @xmath118 relative to the @xmath1 lf and one where only the normalization @xmath119 is evolving . both derivations of this lf have an extremely steep faint - end slope @xmath120 . for fig . [ fig : nexpz10 ] , we show the results for the lf which evolves only in @xmath121 , which fits the candels and xdf / hudf09 datasets better . in addition to the curve resulting from the full selection function simulation , we also show an idealized prediction assuming @xmath108 and integrating the volume over @xmath109 . this overpredicts the expected surface density distribution of candidates by a factor @xmath105 for bright galaxies for both pointings . for the cluster field , we further show the expected number of sources if shear- and magnification - dependent completeness due to blending are ignored ( as is often done in the literature ) . including shear reduces the total expected number of sources in the cluster field by a factor 1.6@xmath122 for this assumed lf . figure [ fig : nexpz10mint ] shows the differential number counts expected for @xmath2 galaxy candidates as a function of _ intrinsic _ , unlensed magnitude . due to the average magnification in the cluster field , the number counts peaks at about 0.5 - 1 magnitude fainter than in the parallel field . however , our simulations show that accounting for magnification - dependent completeness and shear significantly limits the power of lensing clusters to probe galaxies fainter than @xmath123 mag , i.e. intrinsically much fainter than the hubble - ultra deep field . idealized calculations show that the cumulative galaxy number counts are expected to be larger behind a lensing cluster than in the field if the effective slope of the lf is steeper than @xmath124 in which case the reduced observed solid angle due to lensing is more than compensated for by the large abundance of faint , lensed galaxies ( e.g. * ? ? ? given our assumed uv lf with a faint - end slope of @xmath120 the cluster field would thus be expected to show a significantly larger number of high - redshift sources at all magnitudes . however , once we include magnification - dependent completeness , the cluster field in fact shows a very similar total number of expected images of @xmath2 galaxy candidates as the blank field ( within @xmath12515% ) . for the best - fit lf evolving in @xmath119 from @xcite shown in figure [ fig : nexpz10 ] , we predict to detect 0.46 images in the cluster field and 0.49 @xmath2 sources in the parallel . for an lf evolving in @xmath126 instead , we predict 1.3 images in the cluster and 1.1 sources in the parallel field . if these numbers are similar for all the other five hff clusters , the frontier field program is thus expected to find between 6 to 14 new @xmath2 galaxy candidates assuming the two different @xmath2 uv lfs of @xcite are representative . we stress that these numbers depend strongly on the exact evolution of the uv lf at @xmath11 ( see also * ? ? ? nevertheless , at @xmath2 alone , the hffs are likely to more than double the number of reliable lbg candidates known to date . we now combine the first hff cluster and blank field around a2744 to derive a new , independent estimate of the cosmic sfrd . from figure [ fig : nexpz10 ] it is clear that the two images of jd1 behind a2744 which satisfy our selection criteria will result in a higher cosmic sfrd at @xmath2 than we previously determined in the candels - deep and xdf / hudf12 data . we estimate the hff constraint on the @xmath2 cosmic sfrd from the total expected number of galaxy images per wfc3/ir field relative to the earlier @xmath2 uv lf estimate by @xcite . in particular , we use their parametrization for @xmath119-only evolution and search for the normalization , which reproduces two predicted images in the cluster field . with the assumed schechter function parameters of @xmath127 mpc@xmath128mag@xmath16 , @xmath129 mag , and @xmath120 , we predict a total of only 0.46 galaxy images in the cluster field . considering the cluster field alone , finding two images therefore requires a higher normalization , @xmath119 , by a factor @xmath130 compared to the xdf / hudf12 lf . such an increase would , however , result in a total of 2.2 predicted galaxies in the hff parallel blank field , which is marginally inconsistent with not finding any candidate with @xmath4 . we combine the two constraints from the hff cluster and parallel field by multiplying the poissonian probabilities of finding 2 or 0 sources in the two fields , respectively , for a given uv lf normalization @xmath119 . this results in a combined best - fit of @xmath131 mpc@xmath128mag@xmath16 , which is completely consistent , but @xmath132 dex higher than found in the ultra - deep fields . using this lf normalization , we estimate a cosmic sfrd from the a2744 fields . this is done by integrating the lf to obtain the luminosity density , which we then convert to a sfrd using the standard conversion between uv luminosity and sfr @xcite . no correction for dust extinction was applied , consistent with the expectation for very little dust in these galaxies at @xmath11 . while we adopt the widely used approach for deriving the sfr , we note that the conversion of uv luminosity to sfr assumes both an initial mass function ( imf ) as well as a star - formation history . if galaxies are significantly younger than 100 myr or if their star - formation is very bursty ( as predicted by some models ; e.g. * ? ? * ; * ? ? ? * ) , the standard conversion factor may need to be corrected ( see e.g. * ? ? ? * ) . the investigation of these alternatives goes beyond the scope of this paper , however . when integrating the uv lf down to @xmath133 , which corresponds to a sfr limit of 0.7 @xmath134 , we obtain a sfrd of @xmath135 m@xmath136yr@xmath16mpc@xmath128 . this is shown in figure [ fig : sfrdevol ] , where we also plot the previous estimates for comparison . our approach to estimate the sfrd , which is an integrated quantity , is not sensitive to the exact magnification of the candidate source . i.e. even if the magnification at the exact source position presented in @xcite were underestimated and the galaxy intrinsically had a sfr of less than our limit of 0.7 @xmath134 , the sfrd value derived here is still valid . as long as the magnification map produces an accurate differential number count distribution ( which is marginalized over the image plane ) our approach of seeking the best - fitting normalization to the uv lf and integrating this to a fixed sfr limit produces accurate results . while the new constraint from the a2744 hff fields is somewhat higher than the previous ultra - deep field constraints , it is consistent with the rapid decline across @xmath1 to @xmath2 that is predicted by theoretical models . in particular fig [ fig : sfrdevol ] also shows the average sfrd evolution of a series of semi - analytical / empirical models @xcite and from sph simulations @xcite as well as the sfrd from the illustris simulation @xcite . where necessary , we shifted the theoretical models to account for our use of a salpeter imf when converting the uv luminosity to sfr . all theoretical models agree on a very rapid decline in the cosmic sfrd by @xmath137 from @xmath1 to @xmath2 when limited at @xmath138 @xmath134 , indicating that a rapid build - up of galaxies above this limit is a generic prediction of any model of galaxy formation ( see also * ? ? ? nevertheless , given the still large errorbars , a more gradual decline in the sfrd as empirically estimated based on the uv lf evolution across @xmath139 ( see gray line in fig [ fig : sfrdevol ] ) can still not be completely ruled out . if the faint - end slope of the uv lf does not steepen further at higher redshift @xcite or if the escape fraction stays constant , this more gradual decline may be necessary for galaxies to complete reionization in agreement with the high optical depth measurement by wmap ( e.g. * ? ? ? * ; * ? ? ? note , however , that the rapid decline in the observed sfrd may simply be a consequence of our fixed detection limit in luminosity and is likely still compatible with a more gradual evolution of the _ total _ sfrd . this is supported both by the higher sfrd estimates of gamma ray burst counts ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , which are sensitive to the total sfrd , and by simulations ( e.g. , compare with * ? ? ? this paper presented a first estimate of the cosmic sfrd at @xmath2 based on frontier field data . in particular , we show that extensive completeness simulations including source blending and lensing shear close to the critical curve are crucial for any analysis of cluster data . we find a significantly lower completeness at higher magnification than for comparable blank field searches at a fixed observed magnitude ( figures [ fig : posdepcompleteness ] and [ fig : magdepcompleteness ] ) . this can be ascribed to several effects : blending with bright foreground cluster galaxies , higher background due to intra - cluster light , but also due to shear spreading out highly magnified sources over many pixels . sources at high magnification are on average only 70% complete in the a2744 image compared to a blank field even when the effect of shear is ignored ( due to blending with foreground sources and the icl ) . shear further reduces the completeness at @xmath102 by @xmath104 . however , the exact completeness at high magnification sensitively depends on the assumed size distribution for very faint sources below the detection limit of current blank field data ( see fig [ fig : magdepcompleteness ] ) . this effect therefore adds to the overall uncertainty of lf and sfrd estimates from cluster lensing fields . this position - dependent completeness has often been overlooked in the literature ( but see * ? ? ? * and recently ishigaki et al . 2014 , atek et al . however , it has important consequences on the expected number of high - redshift candidates seen behind lensing clusters compared to blank fields . in figure [ fig : nexpz10 ] , we show that the reduced completeness results in a similar number of source images predicted for the a2744 cluster and parallel field , very different from what is commonly assumed . following previous blank field studies , we search the hff a2744 cluster and parallel field data for @xmath2 galaxy candidates using a criterion @xmath4 and non - detections at shorter wavelength . while no candidates are found in the parallel field , we find two images of the same source lensed by the cluster ( previously identified in * ? ? ? * ) which both satisfy our selection criteria . combining the one multiply imaged candidate over the cluster field with the null detection in the parallel field , we derive a cosmic sfrd at @xmath2 which is consistent , but @xmath132 dex higher than found earlier in the ultra - deep blank fields ( see figure [ fig : sfrdevol ] ) . not surprisingly , this independent measurement based on the first completed hff cluster does not allow us to significantly rule out different possible scenarios for the sfrd evolution between @xmath1 and @xmath2 . the combination of these new results with all other estimates from the literature remain consistent with a rapidly declining sfrd as is predicted by cosmological simulations and dark - matter halo evolution in @xmath5cdm . the completed multi - cluster hff dataset will allow to further increase the sample size of galaxies at @xmath2 and to significantly tighten this first estimate of the cosmic sfrd . once biases due to magnification - dependent incompleteness are taken into account , the hff survey will be a key dataset to study the galaxy population at @xmath11 before the advent of the @xmath140 .	we search the complete hubble frontier field dataset of abell 2744 and its parallel field for @xmath0 sources to further refine the evolution of the cosmic star - formation rate density ( sfrd ) between @xmath1 and @xmath2 . we independently confirm two images of the recently discovered triply - imaged @xmath3 source by zitrin et al . ( 2014 ) and set an upper limit for similar @xmath0 galaxies with red colors of @xmath4 in the parallel field of abell 2744 . we utilize extensive simulations to derive the effective selection volume of lyman - break galaxies at @xmath2 , both in the lensed cluster field and in the adjacent parallel field . particular care is taken to include position - dependent lensing shear to accurately account for the expected sizes and morphologies of highly - magnified sources . we show that both source blending and shear reduce the completeness at a given observed magnitude in the cluster , particularly near the critical curves . these effects have a significant , but largely overlooked , impact on the detectability of high - redshift sources behind clusters , and substantially reduce the expected number of highly - magnified sources . the detections and limits from both pointings result in a sfrd which is consistent within the uncertainties with previous estimates at @xmath2 from blank fields . the combination of these new results with all other estimates are also consistent with a rapidly declining sfrd in the 170 myr from @xmath1 to @xmath2 as predicted by cosmological simulations and dark - matter halo evolution in @xmath5cdm . once biases introduced by magnification - dependent completeness are accounted for , the full six cluster and parallel frontier field program will be an extremely powerful new dataset to probe the evolution of the galaxy population at @xmath6 before the advent of the jwst .
s. y. cheng and s .- t . yau proved in @xcite that a bounded strictly pseudoconvex domain in @xmath4 admits a complete negative scalar curvature khler - einstein metric . their arguments also extended to other types of domains , such as a pseudoconvex domain which is the intersection of pseudoconvex domains with @xmath5 boundary . many cases of domains in arbitrary complex manifolds are already dealt with in @xcite , and in @xcite . in @xcite n. mok and s .- t . yau proved the existence of a khler - einstein metric on strictly pseudoconvex domains under some hypotheses . these include , in particular , domains in stein manifolds and domains which admit a negative ricci curvature metric . this article considers the existence of a complete negative scalar curvature khler - einstein metric on a strictly pseudoconvex domain of an arbitrary complex manifold . [ thm : main ] suppose @xmath2 is a strictly pseudoconvex domain in @xmath6 . then @xmath2 admits an unique complete khler - einstein metric of negative scalar curvature if and only if @xmath7 is positive , i.e. admits an hermitian metric with positive curvature . strict pseudoconvexity means that the boundary @xmath8 has a positive cr structure . we consider the case in which the cr structure on @xmath9 is in addition normal , that is , admits transverse vector field preserving it . we prove that any strictly pseudoconvex manifold with a normal cr structure on @xmath9 must be a domain in a resolution of the sasaki cone @xmath10 of the natural sasaki structure on @xmath9 . this severely restricts the strictly pseudoconvex manifolds with a normal cr structure on the boundary . in particular , a domain in a stein manifold must be in @xmath4 and have a boundary diffeomorphic to @xmath11 with cr structure a deformation of the standard one . we consider the following : [ prob : bound - k - e ] which positive normal cr manifold @xmath12 is the conformal boundary of khler - einstein manifold ? there are far too many cr structures on even simple manifolds for this to be a tractable problem in general . this is true even for @xmath13 . it is shown in @xcite that there are inequivalent families of cr structures on all odd dimensional spheres with the number of deformation classes growing doubly exponentially with the dimension . in particular , @xmath14 has 68 inequivalent deformation classes . more interesting is that these cr structures have associated sasaki - einstein metrics . but in the present context , we prove here that they can not be the conformal boundary of a khler - einstein manifold . we prove that any simply connected normal cr manifold @xmath12 satisfying the topological condition for a compatible sasaki - einstein metric , @xmath15 and @xmath16 , are negative examples of [ prob : bound - k - e ] . thus not surprisingly , those @xmath12 which can be the boundary at infinity of a complete ricci - flat manifold as considered in @xcite are excluded . we are able to mostly answer problem [ prob : bound - k - e ] in dimension 3 , just leaving open some cases of finite quotients of @xmath17 . we give many examples in which theorem [ thm : main ] applies . this gives many examples with normal cr structures at infinity and otherwise . an easy case is that of negative holomorphic bundles over a compact complex manifold . [ cor : main ] let @xmath18 be a negative holomorphic bundle over a compact complex manifold @xmath19 . if @xmath20 on @xmath19 , then the disk subbundles @xmath21 admit unique complete khler - einstein metrics of negative scalar curvature . we also construct some examples on resolutions of hypersurface singularities and on some familiar resolutions of quotient singularities . let @xmath9 be a real @xmath22-dimensional manifold . a _ cr structure _ on @xmath9 is a pair @xmath23 consisting of a distribution @xmath24 of real @xmath25-dimensional hyperplanes and an almost complex structure @xmath26 on @xmath27 such that , if @xmath28 denotes the type @xmath29-vectors , the _ formal integrability _ condition holds : @xmath30\subseteq d^{1,0}.\ ] ] the _ levi form _ @xmath31 is defined by @xmath32\ \mod d$ ] , for @xmath33 . it is easy to check that ( [ eq : formal - int ] ) is equivalent to both @xmath34 + [ jx , y ] \in\gamma(d)$ ] and the vanishing of the nijenhuis tensor @xmath35-[x , y]-j\bigl([x , jy]-[jx , y]\bigr)=0,\quad x , y\in\gamma(d).\ ] ] note that the former condition implies @xmath36 is @xmath26-invariant , i.e. @xmath37 . we will always assume @xmath9 is orientable , so @xmath38 is a trivial real line bundle . then there is a 1-form @xmath39 with @xmath40 , and we may identify @xmath41 . note that @xmath36 is only defined up to a choice of orientation of @xmath38 and a positive conformal factor . the cr structure @xmath23 is _ strictly pseudoconvex _ if , for a given orientation of @xmath38 , the levi form @xmath36 is positive , i.e. @xmath42 for any non - zero @xmath43 . note that formal integrability does not in general imply integrability , that is that @xmath12 is a real hypersurface in a complex manifold . the analogue of the newlander - nirenberg theorem only holds for analytic cr manifolds @xcite . a _ finite manifold _ is a pair @xmath44 of complex manifolds with @xmath2 an open relatively compact submanifold of @xmath6 with smooth non - empty boundary @xmath1 . let @xmath45 be a defining function of @xmath46 . that is , @xmath45 is @xmath47 in a neighborhood of @xmath48 , @xmath49 , and @xmath50 on @xmath1 . let @xmath26 denote the complex structure of @xmath6 . the real @xmath22-dimensional manifold @xmath8 has the cr structure @xmath23 where @xmath51 and @xmath26 is restricted to @xmath27 . define a 1-form on @xmath9 @xmath52 where @xmath53 . then it is easy to see that @xmath54 , and the levi form is @xmath41 , which is a @xmath55-form on @xmath27 as follows from the comments after ( [ eq : nijen ] ) . we say that the pair @xmath44 is _ strictly pseudoconvex _ or that @xmath2 is a strictly pseudoconvex domain , if the induced cr - structure @xmath23 is . one can check that by altering @xmath45 , for instance considering @xmath56 for a constant @xmath57 , one may assume that @xmath45 is strictly plurisubharmonic on a neighborhood of @xmath58 . that is , @xmath59 is a positive @xmath55-form . a strictly pseudoconvex domain @xmath46 is a particular type of 1-convex manifold . a complex manifold @xmath60 is said to be 1-convex if there is a stein space @xmath61 , a proper holomorphic surjective mapping @xmath62 satisfying @xmath63 , and a finite set @xmath64 such that if @xmath65 the map @xmath66 is a biholomorphism . then @xmath61 is called the remmert reduction of @xmath60 and @xmath67 is called the exceptional set of @xmath60 . note that @xmath67 is the maximal compact analytic subvariety of @xmath60 , i.e. the union of all compact analytic subvarieties of dimension @xmath68 . let @xmath69 be a vector field on @xmath9 so that @xmath70 we extend @xmath26 to a @xmath55 tensor @xmath71 on @xmath9 by @xmath72 the cr structure @xmath23 on @xmath9 is _ normal _ if there is a @xmath73 satisfying ( [ eq : dir - sum ] ) whose flow preserves @xmath23 . or in other words @xmath74 suppose @xmath75 for an oriented contact form @xmath39 . so we may assume by changing by a conformal factor that @xmath39 is the unique 1-form with @xmath76 and @xmath77 , i.e. @xmath73 is the reeb vector field of @xmath39 . then @xmath23 is normal if and only if @xmath78 in fact , it is easy to check that ( [ eq : normal ] ) is equivalent to ( [ eq : nijen ] ) and @xmath74 . and these two conditions are equivalent to the integrability of the almost complex structure on the _ cone _ @xmath79 @xmath80 where @xmath81 is the radial coordinate on @xmath82 and @xmath83 . if the cr structure @xmath23 is positive and normal , then with @xmath73 positively oriented as above , we have a natural metric @xmath84 in this case @xmath9 has a special type of metric contact structure , known as a _ sasaki structure _ , which we denote by @xmath85 . see @xcite for more details . we denote by @xmath86 the automorphism group of the cr manifold @xmath12 and its lie algebra by @xmath87 . r. schoen @xcite proved the following result . the cr automorphism group of a strictly pseudoconvex cr manifold @xmath88 is proper unless @xmath2 is either @xmath11 or the heisenberg group @xmath89 with the standard cr structures . if @xmath88 is a compact strictly pseudoconvex cr manifold , then the cr automorphism group @xmath86 is compact unless @xmath90 with the standard cr structure , in which case the cr automorphism group is @xmath91 . it will be useful to consider the space of compatible sasaki structures on a normal strictly pseudoconvex cr manifold @xmath12 . see @xcite for more details . let @xmath12 be a strictly pseudoconvex cr structure . a vector field @xmath92 is _ positive _ if @xmath93 for an oriented contact form @xmath39 . we denote by @xmath94 the space of all positive elements of @xmath87 . it is not difficult to see that @xmath94 is isomorphic the space of sasaki structures compatible with @xmath23 , @xmath94 is an open convex cone in @xmath87 , and is invariant under the adjoint action of @xmath86 . let @xmath12 be a normal strictly pseudoconvex cr manifold . the _ sasaki cone _ @xmath95 is the moduli space of sasaki structures compatible with @xmath23 on @xmath9 . we have @xmath96 choose a maximal torus @xmath97 of rank @xmath98 , @xmath99 , with lie algebra @xmath100 . then if @xmath101 denotes the subspace of positive elements , we have @xmath102 where @xmath103 is the weyl group of @xmath104 for a maximal compact subgroup @xmath105 . of course , @xmath106 unless @xmath107 let @xmath108 be the lattice of _ integral _ elements , that is @xmath109 . and define @xmath110 . then every @xmath111 defines a quasi - regular sasaki structure , that is all the orbits of the reeb vector field @xmath73 close to give a locally free @xmath112-action . the @xmath112-action on @xmath9 extends to a locally free holomorphic @xmath113-action on @xmath10 . and @xmath10 is biholomorphic to the total space minus the zero section @xmath114 of a negative holomorphic orbibundle over a khler orbifold @xmath115 ( cf . let @xmath46 be a smooth strictly pseudoconvex domain in a khler manifold @xmath116 . and let @xmath45 be a plurisubharmonic defining function which is strictly plurisubharmonic on a neighborhood of @xmath1 . then @xmath117 is strictly plurisubharmonic near @xmath1 , and @xmath118 is the khler form of a metric near @xmath1 which in coordinates is @xmath119 computation gives @xmath120 where @xmath121 , @xmath122 and @xmath123 . it is also easy to see that @xmath124 thus since @xmath125 as @xmath126 , the metric @xmath127 is complete toward @xmath1 . therefore , _ a fortiori _ the metric @xmath128 with khler form @xmath129 is a complete khler metric on @xmath2 . we will consider the existence of a complete khler - einstein metric on @xmath2 , that is a khler metric @xmath130 with @xmath131 for convenience we will set @xmath132 . if @xmath130 is a complete khler metric on @xmath2 with khler form @xmath133 , suppose we have @xmath134 with @xmath135 then a solution to the monge - ampre equation @xmath136 provides a khler metric @xmath137 satisfying ( [ eq : k - e ] ) . equation ( [ eq : m - a ] ) on noncompact manifolds was extensively studied by s .- y cheng and s .- t . yau @xcite . see also @xcite . there it was proved that ( [ eq : m - a ] ) has a unique solution if @xmath138 and @xmath139 has bounded geometry . we use this method to find a complete solution to ( [ eq : k - e ] ) where @xmath46 is a strictly pseudoconvex domain in a khler manifold @xmath6 . with @xmath45 a defining function of @xmath2 and @xmath140 a khler form on @xmath6 we consider the complete metric with khler form @xmath141 if a line bundle @xmath142 is given by a system of charts and transition functions @xmath143 , then an hermitian metric on @xmath142 is given by a system @xmath144 of smooth positive functions on @xmath145 which satisfy @xmath146 on @xmath147 . in particular , we will use that any other hermitian metric @xmath148 on @xmath142 is of the form @xmath149 for @xmath150 . an holomorphic line bundle @xmath142 is _ positive _ if it has an hermitian metric @xmath151 such that the curvature of the associated chern connection , @xmath152 , satisfies @xmath153 , i.e. is a positive @xmath55-form . the following theorem is mostly due to s. y. cheng and s. t. yau @xcite . [ thm : k - e ] let @xmath44 be a strictly pseudoconvex finite manifold . then @xmath2 admits a complete khler - einstein metric of negative scalar curvature if and only if @xmath154 is positive . let @xmath151 be a positive hermitian metric on @xmath154 and let @xmath148 be any connection on @xmath155 . choose @xmath156 with @xmath157 for @xmath158 and @xmath159 for @xmath160 . set @xmath161 . consider the metric @xmath162 , which has positive curvature on @xmath163 . choose @xmath164 sufficiently small that this set contains the maximal compact analytic subset @xmath67 of @xmath2 , and choose a plurisubharmonic function @xmath165 on @xmath6 which is strictly plurisubharmonic away from @xmath67 . then @xmath166 has positive curvature on a neighborhood of @xmath46 for @xmath167 . suppose @xmath151 is a metric on @xmath155 with @xmath168 positive on a neighborhood of @xmath2 . then the volume form @xmath169 defines an hermitian metric on @xmath155 by @xmath170 for @xmath171-form @xmath172 . then @xmath173 , for some @xmath174 . so we have @xmath175 we define @xmath176.\ ] ] then for the metric @xmath177 we have that @xmath178 satisfies ( [ eq : dd - bar - ric ] ) . it is easy to see that @xmath179 . in fact one checks that @xmath180 then the proof in @xcite shows that ( [ eq : m - a ] ) has a unique solution @xmath181 . the proof follows from an application of the generalized maximum principle to formulae of @xcite to obtain the necessary _ a priori _ estimates . the converse is clear . since if @xmath182 is khler - einstein , then the curvature of ( [ eq : can - metric ] ) satisfies @xmath183 . the following uniqueness result is due to s .- t . yau and follows from a more general schwartz lemma . [ prop : unique ] let @xmath184 and @xmath185 be complete complete khler - einstein manifolds of negative scalar curvature , normalized to have equal einstein constants , @xmath186 . if @xmath187 is a biholomorphism , then @xmath188 . let @xmath189 be the respective volume forms , and define @xmath190 . then we have @xmath191 the arithmetic - geometric inequality applied to the second term on the right of ( [ eq : vol ] ) gives @xmath192 from which we have @xmath193 and it follows from the maximum principle as in ( * ? ? ? * lemma 1.1 ) that @xmath194 . applying the same argument to @xmath195 gives @xmath196 , from which we have @xmath197 and @xmath188 . let @xmath198 denote the group of biholomorophisms of @xmath2 , @xmath199 the group of isometries of @xmath139 , and @xmath200 , @xmath201 their respective lie algebras . we also have the following easy converse to proposition [ prop : unique ] . let @xmath202 be an isometry of khler - einstein strictly pseudoconvex manifold . then @xmath203 is a biholomorphism up to conjugation , i.e. @xmath204 where @xmath26 is the complex structure of @xmath2 . thus @xmath205 furthermore , if @xmath206 , then @xmath207 . first note that since @xmath139 has curvature asymptotic to constant @xmath208 holomorphic bisectional curvature it must be irreducible as a khler manifold . there are two , @xmath26 and @xmath209 , parallel complex structures on @xmath2 . since @xmath210 is irreducible , either @xmath204 , or the holonomy group @xmath211 . in other words , in the second case one can show the existence of three parallel complex structures @xmath212 satisfying the quaternionic identities . but in this case @xmath213 , and ( [ eq : isom - hol ] ) follows . suppose that @xmath214 . the following argument is due to s. kobayashi ( * ? ? ? iii,1 ) . define @xmath215 since @xmath216 and @xmath139 is khler , we have @xmath217 we have @xmath218 and from ( [ eq : a ] ) @xmath219 it follows that @xmath220 is symmetric . but since it is also skew - symmetric , we have @xmath221 . this implies @xmath222 , a contradiction . we consider the boundary behavior of the metric @xmath223 and the einstein metric @xmath224 of theorem [ thm : k - e ] , where @xmath117 for a defining function @xmath45 and @xmath182 is a khler metric on @xmath48 . first a straight forward calculation as in @xcite gives the christoffel symbols @xmath225 and the curvature @xmath226 of @xmath127 near the boundary of @xmath2 . @xmath227 here @xmath228 denotes the curvature and @xmath229 the covariant derivative with respect to @xmath230 . the optimal regularity and asymptotic behavior of the solution @xmath181 to ( [ eq : m - a ] ) of theorem [ thm : k - e ] was given in @xcite for @xmath231 . the proof works with only minor modifications to an arbitrary strictly pseudoconvex @xmath46 with the initial metric @xmath232 . an essential step is to find a defining function @xmath233 so that @xmath178 defined in ( [ eq : f ] ) vanishes to high order on @xmath1 . the following was first proved by c. fefferman @xcite for @xmath231 . there exists a defining function @xmath233 of @xmath46 so that @xmath178 given in ( [ eq : f ] ) satisfies @xmath234 we seek @xmath235 so that @xmath236 so that @xmath237 but since @xmath238 we may take @xmath239 , and ( [ eq : f-0 ] ) is satisfied . then the inductive argument in the proof of @xcite goes through with the operator @xmath240 substituting that used there . the results of @xcite on the asymptotic behavior of the solution @xmath181 to ( [ eq : m - a ] ) with defining function @xmath233 are valid in this situation . one can define hlder spaces @xmath241 with respect to the metric @xmath242 . then if @xmath178 given in ( [ eq : f ] ) vanishes to order @xmath243 , we have @xmath244 moreover , there is an asymptotic expansion of @xmath245 . there are @xmath246 , such that for @xmath247 @xmath248 and vanishes to order @xmath249 . let @xmath232 be either the metric @xmath242 or @xmath250 solving ( [ eq : k - e ] ) , i.e. the khler - einstein metric , on a strictly pseudoconvex @xmath2 , and let @xmath251 be the distance from a fixed point @xmath252 , then the curvature of @xmath232 satisfies @xmath253 thus metric @xmath130 is asymptotically of constant holomorphic sectional curvature @xmath208 and are _ asymptotically complex hyperbolic_(ach ) . if @xmath130 is the khler - einstein metric on a strictly pseudoconvex @xmath254 and @xmath45 is any defining function , then @xmath255 on @xmath256 , where the levi form @xmath36 is of course only defined up to a conformal factor . one could also consider the weaker condition that @xmath258 is represented by a positive @xmath55-form . this is _ a priori _ weaker assumption as the @xmath257-lemma does not generally hold on a 1-convex manifold . it remains whether this weaker assumption is a sufficient condition for theorem [ thm : k - e ] . the following is easy . let @xmath60 be a complex manifold . then @xmath60 satisfies the @xmath257-lemma if and only if for any holomorphic line bundle @xmath142 any @xmath259 is represented by @xmath260 for some hermitian metric @xmath151 . we make the following [ conj : weak - hyp ] a strictly pseudoconvex domain @xmath2 admits a complete khler - einstein metric if and only if there is a khler form @xmath261 . we will give some results that make the conjecture plausible . at least the following results will show that constructing a counterexample to conjecture [ conj : weak - hyp ] would be very difficult . we consider some examples of 1-convex surfaces due to m. coloiu @xcite on which the @xmath257-lemma does not hold , but nevertheless theorem [ thm : k - e ] applies . let @xmath60 be an 1-convex manifold with exceptional set @xmath67 . we denote by @xmath262 the subgroup of line bundles @xmath142 which are topologically trivial on @xmath60 and holomorphically trivial in a neighborhood of @xmath67 . [ prop : b - isom ] for a 1-convex manifold with exceptional set @xmath67 there is a group isomorphism @xmath263.\ ] ] let @xmath264 and @xmath265 be smooth curves in @xmath266 intersecting transversely of degrees @xmath267 and @xmath268 respectively . let @xmath269 be the blow up at each of the @xmath270 points @xmath271 of intersection @xmath272 . if @xmath273 are their strict transforms , then as divisors @xmath274 where @xmath275 are the exceptional divisors . then @xmath276 and @xmath277 . the first inequality , by a theorem of grauert , implies that @xmath278 is exceptional , and the second inequality implies that @xmath279 is 1-convex . and it turns out that @xmath278 is the entire exceptional set . see @xcite for details . in addition it is shown that @xmath280 and by the genus formula we have @xmath281 . thus @xmath282 since @xmath283 , we have @xmath284 one can show using the nakai - moishezon criterion that @xmath285 . clearly , @xmath286 . we need to show that a curve @xmath287 with @xmath288 does not intersect @xmath289 of the points @xmath290 when counted with the multiplicity of @xmath291 at each point . but by bzout s theorem @xmath292 where the sum is over the points of intersection of @xmath291 with @xmath264 . we have @xmath293 where @xmath294 denote the multiplicity of @xmath291 , respectively @xmath264 , at @xmath295 . and @xmath296 shows that @xmath291 can not intersect @xmath289 of the points @xmath290 when counted with the multiplicity . for any relatively compact strictly pseudoconvex domain @xmath0 we have @xmath297 . let @xmath298 be the space of smooth exact @xmath55-forms . the @xmath257-lemma hold on a manifold @xmath60 precisely when the map @xmath299 is surjective . as observed in @xcite @xmath300 provides a nontrivial cokernel of ( [ eq : dd - bar ] ) . in fact , let @xmath301 , and let @xmath151 be any hermitian metric on @xmath142 . then @xmath302 is a real @xmath55-form which is exact because @xmath142 is topologically trivial . but suppose @xmath303 . then the metric @xmath304 has curvature @xmath305 . thus the chern connection of @xmath148 is flat , and since @xmath60 is simply connected , there is a parallel section @xmath306 . since @xmath307 , @xmath203 is holomorphic and @xmath142 is trivial , a contradiction . this defines an injective map @xmath308 if @xmath0 is a sufficiently large relatively compact strictly pseudoconvex domain , then @xmath309 . and from proposition [ prop : b - isom ] and the above we have the there exist infinitely many , topologically distinct , 1-convex surfaces which contain strictly pseudoconvex domains which do not satisfy the @xmath257-lemma but nevertheless satisfy theorem [ thm : k - e ] . the above arguments lead to the following more general result which is perhaps worth mentioning . let @xmath60 be a complex manifold with @xmath310 . then @xmath60 satisfies the @xmath257-lemma if and only if @xmath311 . denote by @xmath312 the subgroup of @xmath313 of topologically trivial line bundles . then we have @xmath314.\ ] ] choose @xmath315 so that @xmath316 . let @xmath317 be an element with @xmath318 non - trivial . for any hermitian metric @xmath151 , @xmath319 . if @xmath320 , then there exists a metric @xmath148 with a flat connection on @xmath142 . but a flat connection corresponds to an @xmath321 . we have @xmath322 , which implies that @xmath323 is trivial . the following shows that a counterexample to conjecture [ conj : weak - hyp ] would have to have singular exceptional set . let @xmath44 be a strictly pseudoconvex finite manifold with khler form @xmath261 . if the exceptional set @xmath324 is smooth , possibly not connected , then @xmath155 admits an hermitian metric which is positive in a neighborhood @xmath325 of @xmath48 . let @xmath151 an hermitian metric on @xmath155 . since each connected component @xmath326 of @xmath67 is obviously khler , there exists an @xmath327 so that the metric @xmath149 satisfies @xmath328 we may assume that @xmath6 is 1-convex , and let @xmath329 be the remmert reduction . if @xmath165 is the pull - back by @xmath330 of a strictly plurisubharmonic function on @xmath61 , then we have @xmath331 then it is easy to see from ( [ eq : except - ddbar ] ) , ( [ eq : stron - pseu1 ] ) , ( [ eq : stron - pseu2 ] ) , and ( [ eq : stron - pseu3 ] ) that for @xmath57 sufficiently large the metric @xmath332 has curvature @xmath333 where @xmath325 is a relatively compact neighborhood of @xmath48 . assuming that the cr boundary @xmath58 of a strictly pseudoconvex manifold is normal has strong consequences on @xmath2 . first we mention an embedding result for 1-convex manifolds in @xcite which was generalized to complex spaces in @xcite . [ thm:1-conv - emb ] let @xmath60 be a 1-convex complex space . then @xmath60 is embeddable in @xmath334 if and only if there is a positive holomorphic line bundle on @xmath60 . given a positive line bundle @xmath142 on @xmath60 it shown that there is an @xmath335 so that for @xmath336 there finitely many sections @xmath337 so that @xmath338 is empty and the map @xmath339 restricts to an embedding on a neighborhood @xmath340 of the exceptional set @xmath67 of the remmert reduction @xmath341 . this is combined with the embedding @xmath342 of the stein space @xmath61 ( cf . @xcite ) gives an embedding @xmath343 this gives us another necessary and sufficient condition for a strictly pseudoconvex domain @xmath2 to admit a khler - einstein metric . [ cor : k - ample ] a strictly pseudoconvex domain @xmath2 admits a khler - einstein metric if and only if for some @xmath344 there are finitely many sections @xmath345 inducing an embedding of a neighborhood @xmath340 of the exceptional set of @xmath2 in @xmath346 . [ prop : nor - st - pseudo ] suppose @xmath2 is a strictly pseudoconvex manifold such that the induced cr structure on @xmath58 is normal . then the remmert reduction of @xmath2 is @xmath347 , where @xmath348 is the domain in the sasaki cone of @xmath9 , with its induced sasaki structure . note that @xmath349 , with the vertex , has a unique structure of a normal stein variety @xcite , and also an affine variety @xcite . we will consider @xmath10 as such with the addition of the vertex . in other words , @xmath0 is a domain in a resolution @xmath350 of the cone @xmath10 with exceptional fiber @xmath351 over @xmath352 . we may suppose that @xmath0 with @xmath60 a 1-convex manifold with remmert reduction @xmath62 . thus @xmath330 maps @xmath2 to the strictly pseudoconvex domain @xmath353 . we first prove the following . the action of @xmath86 extends to a holomorphic action on @xmath19 . since @xmath61 has finitely many isolated singular points , it has finite embedding dimension and there there is an embedding @xmath354 ( cf . let @xmath355 , and define @xmath356 we have @xmath357 , i.e. @xmath358 is annihilated by @xmath359 , so by the extension theorem of j. kohn and h. rossi @xcite the @xmath358 extend to holomorphic functions on @xmath48 . there are holomorphic functions @xmath360 with @xmath361 . denote @xmath362 and @xmath363 . so @xmath364 and @xmath365 with @xmath366 . we have @xmath367 . if @xmath368 is any function defined with @xmath369 with @xmath340 connected and vanishing on @xmath370 , then @xmath371 vanishes on @xmath372 so vanishes identically . therefore there is a neighborhood @xmath373 of @xmath9 with @xmath374 . now cover @xmath375 with finitely many neighborhoods @xmath376 for which @xmath377 for defining functions @xmath378 . let @xmath379 . if @xmath380 , then @xmath381 vanish on @xmath382 . thus @xmath383 . continuing this argument shows that @xmath384 . since @xmath385 maps @xmath61 biholomorphically onto its image , we can define @xmath386 by @xmath387 . let @xmath45 be a plurisubharmonic defining function of @xmath353 , i.e. @xmath388 . then @xmath389 is plurisubharmonic and takes the value 0 on @xmath390 , so @xmath391 for @xmath392 by the maximum principle . therefore we have @xmath393 . suppose @xmath394 . since @xmath395 vanishes on @xmath9 it must vanish identically . therefore @xmath396 , so @xmath86 acts on @xmath397 by biholomorphisms . suppose @xmath398 . then we can replace @xmath73 with an integral element in @xmath399 . this @xmath73 generates an @xmath112-action on @xmath397 . if @xmath45 is a defining function for @xmath0 , @xmath400 , so @xmath401 points inward at @xmath9 . if @xmath402 , then @xmath403 \times s \ni ( t , x)\longrightarrow \exp(-tj\xi)x\ ] ] is a diffeomorphism onto a neighborhood of @xmath9 in @xmath397 . we consider a complex structure on @xmath404 which is that of ( [ eq : cone - cst ] ) in the coordinate @xmath405 of @xmath406 ; that is , @xmath407 for @xmath83 . it is not difficult to see that ( [ eq : bound - diff ] ) is a biholomorphism between @xmath408\times s , i)$ ] and @xmath409 where @xmath410 is a neighborhood of @xmath9 in @xmath397 . since @xmath411 and @xmath19 are both normal stein spaces hartogs theorem implies that an holomorphic function on @xmath412 extends to @xmath10 and likewise for holomorphic functions on @xmath413 . therefore @xmath414 , and we have a biholomorphism @xmath415 . [ cor : stein - dom ] if @xmath2 is a strictly pseudoconvex stein domain with the cr structure on @xmath58 normal , then @xmath13 , the @xmath22 sphere with a transversal deformation of the standard cr structure . see section [ subsubsect : transv ] for an explanation of `` transversal deformation . '' since @xmath2 is a domain in the resolution @xmath416 , we must have @xmath417 . thus @xmath418 and @xmath10 is nonsingular . suppose @xmath419 . we have an action @xmath420 , with weights @xmath421 . one can get an equivariant coordinate system @xmath422 , i.e. @xmath423 , by a simple averaging argument . since @xmath424 maps @xmath2 into @xmath340 for large enough @xmath425 , we have @xmath426 . we have @xmath427 with the standard cr structure @xmath23 and @xmath73 is a cr reeb vector field . for @xmath428 there is a unique @xmath429 with @xmath430 , and if we define @xmath431 then if we set @xmath432 , @xmath433 . note that @xmath434 , so the euler vector field and also the reeb vector field is unchanged . and the contact forms are related by @xmath435 let @xmath0 be a strictly pseudoconvex domain in one of the 1-convex surfaces of section [ sec : com - dd - bar ] so that @xmath309 . then the cr structure on @xmath58 is not normal . suppose otherwise , then the remmert reduction is @xmath436 . let @xmath437 be a non - trivial element . by definition @xmath142 is holomorphically trivial in a neighborhood of the exceptional curve @xmath67 . so there is an @xmath438 on @xmath439 with @xmath440 . but obviously , @xmath441 , so @xmath442 implying that @xmath142 is holomorphically trivial . this proposition also simply follows from ( * ? ? ? 1 ) which give as a condition for @xmath443 that @xmath444 is injective . this proposition has an obvious generalization . let @xmath0 be a strictly pseudoconvex domain . if @xmath445 , then the cr structure on @xmath1 is not normal . a strictly pseudoconvex finite manifold @xmath2 is called a _ normal _ khler - einstein manifold if it has a complete khler - einstein metric and its conformal cr infinity is normal . it is well known @xcite , see also @xcite , that the khler cone @xmath349 over a sasaki manifold @xmath9 is an affine variety . the homomorphism @xmath446 with @xmath447 extends to @xmath448 , an algebraic action . the following result is evident when this is combined with theorem [ thm:1-conv - emb ] . a normal khler - einstein manifold is a domain in a quasi - projective variety . we have that @xmath0 where @xmath60 is a resolution of @xmath349 with @xmath58 . we compactify @xmath60 to @xmath449 as follows . choose @xmath450 . then @xmath10 is biholomorphic to @xmath114 where @xmath451 is an holomorphic orbibundle over @xmath115 . now define an orbifold @xmath449 by replacing @xmath452 with the total space of @xmath453 , so @xmath449 is @xmath60 with the divisor @xmath115 with positive normal bundle @xmath454\|_w$ ] added . ( the divisor @xmath455 is only @xmath456-cartier , but the following arguments work in the orbifold setting . see @xcite . ) let @xmath457 be a section vanishing on @xmath115 . let @xmath151 be a metric on @xmath458 $ ] with @xmath459 . in a neighborhood of @xmath115 identified with a neighborhood of the zero section of @xmath460 define @xmath461 , where @xmath462 is the fiber coordinate , and extend @xmath463 to all of @xmath458 $ ] . then @xmath464 is strictly plurisubharmonic near @xmath115 . since @xmath465 at @xmath115 , we can modify it away from @xmath115 to a plurisubharmonic function @xmath466 on @xmath60 . if we set @xmath467 , then @xmath468 is a metric on @xmath458 $ ] with @xmath469 and @xmath470 in a neighborhood @xmath19 of @xmath115 . then for sufficiently large @xmath471 , @xmath472 $ ] admits a metric with positive curvature . by the baily embedding theorem @xcite for sufficiently large @xmath473 , the sections of @xmath474 define an embedding @xmath475 . let @xmath2 be a normal khler - einstein manifold with cr infinity @xmath12 . then the action of the connected component @xmath476 extends to a holomorphic action on @xmath2 . thus there is an injection @xmath477 let @xmath478 $ ] be the very ample bundle as above . choose any @xmath450 , then as above the action of @xmath113 on @xmath10 is algebraic . let @xmath479 , so @xmath480 . if @xmath481 is the projection , then we have a section @xmath482 of @xmath483 @xmath484 , @xmath485 . and @xmath482 is a rational section of @xmath486 , where @xmath487 is again the projection . we know that @xmath482 is regular except perhaps on some @xmath488 where @xmath489 is an exceptional divisor of @xmath60 or on @xmath490 . but since @xmath491 is smooth , it is easy to see that @xmath482 is regular along @xmath492 . it is easy to see that @xmath482 has at most a pole of order @xmath493 along @xmath490 . so @xmath494 for @xmath485 . as above we have the embedding @xmath495 if we denote the action of @xmath113 on @xmath496 by @xmath497 , then for @xmath498 @xmath499 . so @xmath113 preserves @xmath500 . since @xmath67 is nowhere dense , it is easy to see that @xmath501 is preserved by the action of @xmath113 on @xmath496 . for each @xmath111 the @xmath112 action extends to @xmath60 . these actions generate @xmath97 . since all maximal tori are conjugate , @xmath502 generates @xmath100 for all maximal tori . so the action of every maximal torus @xmath97 extends to @xmath2 . when @xmath86 is compact every element of @xmath476 is contained in a maximal torus so the result follows . otherwise , @xmath12 is the round sphere @xmath503 and @xmath504 the complex hyperbolic space , and the theorem follows . every normal khler - einstein manifold @xmath2 , with fixed cr reeb vector field , has a natural infinite dimensional space of deformations parameterized by basic functions on @xmath58 with sufficiently small 2nd derivatives . by basic we mean invariant under the reeb action of @xmath73 . these correspond to _ transversal deformations _ of the cr structure on @xmath9 . as above @xmath505 , so we may take @xmath506 as the defining function of @xmath2 . the cr distribution @xmath54 , where @xmath507 . here @xmath39 is the unique 1-form with @xmath508 and @xmath509 . if @xmath510 is a basic function , which we may take as a function on @xmath10 , then set @xmath511 . then @xmath512 defines the boundary @xmath513 of a domain in @xmath10 with defining function @xmath514 . one can check that @xmath513 is naturally diffeomorphic to @xmath9 , one has @xmath515 , and this alters the cr structure on @xmath9 by @xmath516 if @xmath517 is nowhere zero , this defines another normal strictly pseudoconvex cr structure on @xmath9 . this is equivalent to the new levi form @xmath518 being positive . equations ( [ eq : cont - def ] ) and ( [ eq : compl - def ] ) define the deformed cr structure on @xmath9 . alternatively , one may fix the complex structure on @xmath10 deform the domain via @xmath511 . figure [ fig : transverse ] shows the respective domains @xmath411 and @xmath519 . another type of deformation will be of interest in the final section which we call _ transversal deformations of the second kind_. let @xmath520 , then we have the cr structure on @xmath9 @xmath521 with @xmath71 given again by ( [ eq : compl - def ] ) . if @xmath522 , with @xmath523 , then the new cr structure is just a gauge transformation of @xmath23 in the reeb direction . we consider the problem of which strictly pseudoconvex cr manifolds @xmath9 are cr infinities of some khler - einstein manifold . which strictly pseudoconvex cr manifolds @xmath9 are conformal cr infinities of complete khler - einstein manifolds . by theorem [ thm : k - e ] this is equivalent to whether there exists a complex manifold @xmath2 with @xmath524 and @xmath7 positive . one necessary condition is that @xmath9 is embeddable , meaning there is a smooth embedding @xmath525 with the cr structure on @xmath9 induced from the complex structure of @xmath526 . conversely by a theorem of harvey and lawson @xcite @xmath58 for some @xmath2 if @xmath9 is embeddable . it is known @xcite that all strictly pseudoconvex @xmath12 are embeddable in dimension @xmath527 . but generic perturbations of the standard cr structure on @xmath17 are nonembeddable . we consider the more restricted problem [ prob : bound - k - e ] , considering only normal cr structures . we can prove a negative result in case @xmath9 has a normal cr structure @xmath23 . we will need a definition . we say that a normal cr manifold @xmath12 has _ property s - e _ if @xmath528 and for @xmath529 the transversal first chern class @xmath16 , i.e. represented by a positive @xmath55-form . if this property holds for some @xmath529 , then it holds for all of @xmath530 , so it is intrinsic to @xmath23 . if @xmath531 , then by conjugation with an element of @xmath532 we may suppose @xmath533 where @xmath534 is the lie algebra of a maximal torus of @xmath532 . the respective contact forms satisfy @xmath535 where @xmath536 , so @xmath537 recall that @xmath538 , where @xmath539 is with respect to the foliation generated by @xmath73 . thus from ( [ eq : con - kah ] ) with respect to the foliation of @xmath540 @xmath541 here the ricci forms are computed on @xmath23 . this is precisely the topological condition for the associated sasaki structure @xmath542 to admit a possible transversal deformation to a sasaki - einstein structure . sasaki manifolds satisfying this condition have been studied extensively ( cf . @xcite , ) mainly in order to construct new einstein manifolds . in terms of the khler cone @xmath10 property s - e is equivalent to the existence of an holomorphic @xmath171-form @xmath172 with @xmath543 , with @xmath544 , and satisfying @xmath545 where @xmath133 is the natural khler metric on @xmath10 for the sasaki manifold @xmath546 and @xmath151 is invariant under the action generated by @xmath73 and @xmath547 . see @xcite . if @xmath9 is not simply connected then one may have to take @xmath172 to be multivalued , i.e. it defines a section of @xmath548 for some @xmath549 . we have @xmath550 defining an hermitian metric on @xmath551 via ( [ eq : hol - form ] ) @xmath552 for @xmath171-form @xmath553 . it is easy to see that the associated connection @xmath554 on @xmath551 is flat . and we have the surjective homomorphism to the holonomy group @xmath555 , whose image is a finite group @xmath556 . if @xmath557 , then the singularity @xmath558 is _ gorenstein_. and if @xmath559 , @xmath558 is _ @xmath456-gorenstein_. [ thm : poss - inf ] let @xmath12 be a normal strictly pseudoconvex cr manifold with property s - e such that the singularity @xmath349 is gorenstein , e.g. @xmath560 . then @xmath12 is the cr infinity of a complete khler - einstein manifold @xmath2 only if @xmath13 , with @xmath23 a transversal deformation of the standard cr structure and @xmath231 . one can show if @xmath12 has property s - e , then @xmath561 must be finite . this follows from the existence of a sasaki structure with positive ricci curvature and an application of meyer s theorem . in most applications one will have @xmath12 simply connected . by proposition [ prop : nor - st - pseudo ] @xmath2 is a resolution @xmath562 . we have a holomorphic n - form @xmath172 on @xmath10 satisfying ( [ eq : hol - form ] ) . it is a result of @xcite and @xcite that @xmath352 is a rational singularity if and only if it has a small neighborhood @xmath340 with @xmath563 and this easily follows from ( [ eq : hol - form ] ) . moreover , it is a consequence of ( [ eq : rational ] ) that for any resolution @xmath562 the form @xmath172 extends to a holomorphic form on @xmath2 ( cf . we recall a definition . let @xmath60 be a normal @xmath456-gorenstein variety . then @xmath60 has _ canonical singularities _ if for every resolution @xmath564 one has @xmath565 with @xmath566 for each exceptional divisor @xmath326 . the equality in ( [ eq : can - sing ] ) means linear equivalence . it is sufficient to check ( [ eq : can - sing ] ) for one resolution , and we have each @xmath566 . by corollary [ cor : k - ample ] for some @xmath567 @xmath568 is ample in a neighborhood of the exceptional set @xmath67 , and we have @xmath569 . if @xmath570 , then @xmath571 is a meromorphic function which is holomorphic on @xmath572 . by the riemann extension theorem @xmath466 extends holomorphically to @xmath573 on @xmath411 . so @xmath574 , and @xmath575 is a constant multiple of @xmath576 . therefore , we must have @xmath577 , @xmath578 . since @xmath2 is smooth we must have @xmath231 and the rest follows as in the proof of corollary [ cor : stein - dom ] . theorem [ thm : poss - inf ] is a global result . if @xmath12 satisfies the assumptions of the theorem and in addition is sasaki - einstein , then the ansatz in section [ subsect : diag - quot ] , due to e. calabi , constructs an incomplete khler - einstein metric with cr infinity @xmath12 . note also that the examples in [ subsect : hirz - jung ] and [ subsect : diag - quot ] show that gorenstein assumption in the theorem in necessary . we consider some cases in which theorem [ thm : k - e ] is easily applicable . the following easy result will be helpful in some of the cases that follow . [ prop : quasi - proj ] suppose @xmath60 is a projective manifold and @xmath579 is a smooth divisor ( @xmath60 may have orbifold singularities along @xmath115 ) with @xmath580 > 0 $ ] , for some @xmath344 , and @xmath458\|_w > 0 $ ] . then @xmath581 is 1-convex and @xmath582 . let @xmath18 be an holomorphic bundle with a hermitian metric @xmath151 . recall , that @xmath583 has a unique chern connection @xmath584 which is compatible with @xmath151 and @xmath585 . in a holomorphic local frame @xmath586 the connection form is @xmath587 . we have the curvature @xmath588 given by @xmath589 this condition was called weakly positive by p. griffiths @xcite , as it is not strong enough to ensure the properties of a positive line bundle such as kodaira vanishing . although , a weaker condition than in [ defn : pos ] was called weakly positive by h. grauert @xcite . [ prop : neg - pseudo ] suppose the metric @xmath151 has negative curvature . then the disk bundles @xmath594 , for @xmath595 , are strictly pseudoconvex . in fact , @xmath596 is a positive @xmath55-form outside the zero section . choose a local holomorphic frame @xmath586 , with fiber coordinates @xmath597 , so that @xmath598 at @xmath590 . then at @xmath590 we have @xmath599 while @xmath600 the two middle terms on the right of ( [ eq : psh - r ] ) vanish . thus ( [ eq : neg - cur ] ) and ( [ eq : psh - r ] ) show that at @xmath601 we have @xmath602 . if @xmath583 is an arbitrary holomorphic vector bundle and @xmath142 is a negative line bundle on @xmath19 , then the curvature of @xmath603 is @xmath604 thus for sufficiently large @xmath605 @xmath603 is negative . associated to a vector bundle @xmath18 is the bundle of projective spaces @xmath606 with fibers @xmath607 . let @xmath608 be the universal bundle of lines in @xmath583 . the hermitian metric @xmath151 on @xmath583 defines a natural metric on @xmath142 . we will compute the curvature @xmath609 of this metric on @xmath142 in terms of the curvature @xmath610 of @xmath583 . this was proved in @xcite . we prove it here as it is important to what follows . let @xmath586 be a local holomorphic frame of @xmath583 , and let @xmath611 be fiber coordinates . then denote @xmath612 . we have @xmath613 rearranging terms in ( [ eq : curv - her ] ) we get @xmath614 where @xmath615 is the fubini - study metric on the fibers and @xmath616 one can choose a frame @xmath586 so that the connection form @xmath617 vanishes at @xmath618 . then @xmath619 , so we have the following . consider the compactification @xmath622 of @xmath583 . so we have @xmath623 let @xmath624 be the divisor at infinity , that is @xmath625\in \mathbb{p}(\mathbf{e}\oplus{\mathbb{c}})\}=\mathbb{p}(\mathbf{e})$ ] . clearly , @xmath626 $ ] and @xmath627 restrict to the hyperplane bundle on each fiber @xmath628 . thus @xmath626\otimes\mathbf{l } = \tilde{\pi}^(\mathbf{f})$ ] for a line bundle @xmath629 on @xmath19 . one can check that the normal bundle @xmath630\|_{d_\infty } \cong\mathbf{l}^{-1}\|_{d_\infty}$ ] , thus @xmath631 is trivial . the projection @xmath632 induces an injection on the picard group , therefore @xmath626=\mathbf{l}^{-1}$ ] . in particular , suppose @xmath18 is a negative bundle . further , suppose that if @xmath6 denotes the total space of @xmath583 , @xmath633 . this can be seen to be equivalent to @xmath634 . for if @xmath635 with @xmath636 negative , then @xmath637 is a negative form in @xmath638 where @xmath639 . by ( [ eq : can - bund2 ] ) , proposition [ prop : neg - taut ] , and the above comments we have the following . [ cor : neg - bund ] let @xmath6 be the total space of a negative holomorphic bundle @xmath18 such that @xmath641 , equivalently @xmath642 , then the strictly pseudoconvex tubular neighborhoods @xmath643 of the zero section admit unique complete khler - einstein metrics , with normal cr infinity the sphere bundle @xmath644 . a polynomial @xmath645 $ ] is _ weighted homogeneous _ with weights @xmath646 and degree @xmath647 if @xmath648 here we assume that @xmath649 . we assume that @xmath650 is smooth away from @xmath651 . then it is well known , see @xcite , that @xmath652 has a natural sasaki structure with reeb vector field generating the action @xmath653 and the cr structure of @xmath9 satisfies property s - e precisely when @xmath654 , _ loc . cit._. the codimension of the singular set of @xmath60 is @xmath655 , so @xmath60 is normal . and @xmath60 is easily seen to be gorenstein with holomorphic form given on @xmath656 by adjunction by @xmath657 where @xmath658 for @xmath659 . therefore we have the following . we have @xmath662 . and by proposition [ prop : nor - st - pseudo ] any strictly pseudoconvex domain with cr infinity @xmath9 must be the domain @xmath663 in a resolution @xmath664 . examples of such resolutions are easy to find by taking blow - ups or more generally weighted blow - ups . the weight @xmath665 defines a grading @xmath666=\oplus_{k\geq 0 } { \mathbb{c}}[z_0,\ldots , z_n]_k$ ] . we define @xmath667_k \}$ ] . the _ weighted blow - up _ @xmath668 is constructed similarly to the usual but with the weighted grading , and the exceptional fiber @xmath669 , the weighted projective space . and , of course , one obtains the usual blow - up with @xmath670 . if @xmath671 is the strict transform , then we have the adjunction formula for the canonical bundle @xmath672 provided @xmath673 does not contain a divisor singular along a singular set of @xmath674 . see @xcite for more details . consider the hypersurface @xmath675 with @xmath676 . we consider a series of blow - ups of @xmath60 . blowing up gives @xmath677 where @xmath678 is the strict transform of @xmath60 in @xmath679 , the blow - up of @xmath680 at the origin . then @xmath678 is covered with affine neighborhoods @xmath681 . take for example @xmath682 which has coordinates @xmath683 and @xmath330 is given by @xmath684 . thus if @xmath685 , then @xmath686 . so if @xmath687 , then @xmath688 . it is elementary to check that this is a non - singular hypersurface and similarly for @xmath689 . we have @xmath690 where @xmath691 , and this hypersurface has a singular point at the origin unless @xmath692 or @xmath693 . repeating the procedure we get a resolution @xmath664 , @xmath694 , if @xmath695 or @xmath696 . denote by @xmath326 the strict transform of the exceptional set of the @xmath697 blow - up . then if follows from ( [ eq : adjunct ] ) that @xmath698 in order to prove that @xmath660 we will compactify @xmath449 and employ a lemma of h. grauert . let @xmath699 and set @xmath700 and @xmath701 . then we have @xmath702 , where @xmath703 is the weighted projective space , and @xmath704 is weighted homogeneous with respect to these weights . let @xmath705 . if @xmath706 are homogeneous coordinates on @xmath703 , then we have added @xmath707 to @xmath680 . let @xmath708 . let @xmath709 be the above resolution of @xmath61 given by resolving @xmath710 . we will prove that @xmath711 + t[e_\infty ] > 0\ ] ] on @xmath709 for @xmath712 sufficiently large . note that @xmath629 is not a cartier divisor unless @xmath713 . we will use the following due to h. grauert @xcite . [ lem : grauert ] a line bundle @xmath142 on a compact complex space @xmath60 is positive if and only if for every irreducible compact nowhere discrete analytic subspace @xmath714 there is an holomorphic section @xmath203 of @xmath715 , for some @xmath98 , with a zero on @xmath716 but not vanishing entirely . let @xmath717 be an hypersurface which is tangent to the line @xmath718 at @xmath719 to at least order @xmath720 . if @xmath721 , then one can check that @xmath722 , where @xmath723 is the strict transform of @xmath27 . assume that @xmath466 is algebraic , so it extends to a rational function on @xmath61 with a pole along @xmath724 . if @xmath425 is sufficiently large , then @xmath725 gives an holomorphic section of @xmath629 with @xmath726 , for some @xmath727 . it is not very difficult to check that , for various such @xmath27 , the condition of lemma [ lem : grauert ] is satisfied for @xmath728 not contained in @xmath724 . suppose @xmath729 . the above argument gives a section @xmath203 of @xmath730 with @xmath731 . if @xmath732 is sufficiently divisible by @xmath733 and @xmath734 , then there are many rational functions @xmath735 with @xmath130 not vanishing along @xmath724 . we have @xmath736 , and various choices of @xmath27 and @xmath130 give the required section of @xmath737 . let @xmath738 with @xmath739 . then @xmath60 can be resolved similar to example 1 but by taking blow - ups with weight @xmath740 . one can repeatedly blowing up the unique singular with this weight @xmath741 times . and if @xmath742 or @xmath743 , this ends in a smooth resolution @xmath744 . one can check using arguments as in example 1 that @xmath660 . let @xmath745 with @xmath746 . again @xmath60 can repeatedly blowing up with weight @xmath747 the unique singular point at each step @xmath748 times . if @xmath742 or @xmath749 , this ends in a smooth resolution @xmath750 . and again similar arguments show that @xmath660 . we will consider examples of @xmath751 khler manifolds where the metric converges to that of the complex hyperbolic space @xmath752 . more precisely , suppose @xmath753 is a finite group . by making a conjugation , we may assume that @xmath754 . assume that @xmath755 acts freely away from @xmath756 , where @xmath757 is the unit ball . the khler manifold @xmath139 is _ locally strongly ach _ , of order @xmath758 , if there is a compact set @xmath759 , a ball @xmath760 , and a biholomorphism @xmath761 , such that if @xmath182 is the hyperbolic metric of @xmath762 , latexmath:[\[\label{eq : met - conv } also , one may consider the weaker condition that @xmath165 is merely a diffeomorphism and the complex structure @xmath26 of @xmath2 converges to @xmath766 of @xmath764 as in ( [ eq : met - conv ] ) . see @xcite for some interesting rigidity results for such strongly ach manifolds . especially considering proposition [ prop : st - ach ] , it is an interesting question whether there are similar rigidity results for _ locally _ strongly ach manifolds . [ prop : st - ach ] let @xmath139 be a khler manifold with khler form @xmath767 , and assume that @xmath768 is biholomorphic to @xmath769 , with @xmath755 as above . then there is a locally strongly ach khler - einstein metric @xmath130 on @xmath2 which is of order @xmath770 for all @xmath771 , with convergence including all derivatives . that is ( [ eq : der - conv ] ) holds for all @xmath772 . by proposition [ prop : nor - st - pseudo ] @xmath2 is a domain in a resolution @xmath773 . since this a resolution of a rational singularity the @xmath257-lemma holds . thus @xmath774 admits an hermitian metric @xmath151 so that @xmath775 is a khler form . if @xmath776 , then @xmath777 and @xmath778 is a meromorphic section of @xmath779 , which incidentally must have poles on the exceptional divisors . we have a metric on @xmath779 , also denoted by @xmath151 , with @xmath780 . let @xmath781 , and for @xmath782 define a cut - off function @xmath783 @xmath784 let @xmath463 be the metric on @xmath785 , where @xmath67 is the exceptional set , with @xmath786 . define the metric @xmath787 which for @xmath788 sufficiently large @xmath789 is positive . since @xmath790 also defines an hermitian metric on @xmath791 , we can define @xmath792 by @xmath793 where @xmath794 for @xmath795 . in particular , we have ( [ eq : omega0-ricci ] ) . the defining function @xmath796 where @xmath797^{\frac{n}{n+1 } } \bigl(r^2 -1 \bigr)$ ] has @xmath798 in a neighborhood of @xmath799 . we may modify @xmath45 on @xmath800 so that @xmath801 on @xmath2 . then on @xmath802 ( [ eq : f ] ) becomes @xmath803=0,\ ] ] as is not difficult to check . therefore , when we solve ( [ eq : m - a ] ) to get the khler - einstein metric @xmath804 on @xmath802 we have @xmath805 where @xmath806 satisfies both ( [ eq : u - asym - hol ] ) and ( [ eq : u - asym ] ) . the conclusion follows from the observation that if @xmath251 is the distance with respect to the bergman metric and @xmath45 is any defining function , then @xmath807 for @xmath808 . let @xmath809 be relatively prime integers and consider the finite group @xmath810 generated by @xmath811 then @xmath812 has an isolated orbifold singularity at the origin , and its minimal resolution given by a hirzebruch - jung string is well known . see ( ? ? ? ii , 5 ) and @xcite for a description in terms of toric geometry . this minimal resolution @xmath813 has the following properties : the integers @xmath822 are given by the continued fraction expansion @xmath823 in other words , the @xmath822 are determined by the euclidean algorithm where we define @xmath824 , inductively @xmath825 note that the resolution @xmath813 is the unique minimal toric resolution . and one can retrieve the toric diagram , i.e. the stabilizers of the @xmath815 , from the @xmath822 as follows . suppose @xmath815 has stabilizer @xmath826 , with @xmath827 and @xmath828 coprime , then we have @xmath829 and from ( [ eq : stabil ] ) we obtain @xmath830 unambiguously since @xmath831 . if we denote by @xmath832 and @xmath833 the non - compact curves with stabilizers @xmath29 and @xmath834 respectively corresponding to the axis of @xmath835 , then the arrangement of curves is given in figure [ fig : resol ] . we have @xmath842 if and only if @xmath843 for @xmath844 , in which case proposition [ prop : st - ach ] gives a locally strongly ach khler - einstein metric on the domain @xmath845 , where @xmath846 , which is invariant under @xmath847 . it is known from d. calderbank and m. singer @xcite that the domains @xmath848 admit asd ( self - dual weyl curvature vanishes , @xmath849 . ) hermitian einstein metrics with negative scalar curvature . these metrics are toric but not khler . and furthermore , these metrics have the same cr - infinity as the khler - einstein metrics given here . it would be interesting to know how these metrics are related in say the moduli of einstein structures on @xmath848 . let @xmath851 be the group acting on @xmath4 with generator @xmath852 , @xmath853 . then the singularity @xmath850 has as resolution @xmath854 the total space of @xmath855 , with @xmath142 the tautological line bundle . then @xmath842 , if and only if @xmath856 , and in this case proposition [ prop : st - ach ] gives a locally strongly ach khler - einstein metric on the domain @xmath845 , where @xmath781 , which is invariant under @xmath857 . the metrics in this example are of cohomogeneity one , so one expects to find an explicit formula for the khler potential of @xmath133 . there is a simple formula for these metrics due to a. futaki @xcite . the 2-form @xmath858 is basic with respect to the @xmath113-action on @xmath859 and restricts to the fubini study metric on the quotient @xmath860 . define @xmath861 then the transversal ricci form , i.e. that of @xmath862 on @xmath860 , satisfies @xmath863 let @xmath864 . the technique of e. calabi @xcite is to consider metrics of the form @xmath865 it turns out to be easier to work in a _ momentum coordinate _ along the fiber . so set @xmath866 then ( [ eq : calabi ] ) becomes @xmath867 then its ricci form is computed in @xcite to be @xmath868 and the einstein equation @xmath869 is satisfied with @xmath870 if @xmath871 one retrieves the complex coordinate expression ( [ eq : calabi ] ) by integrating , for fixed @xmath872 , @xmath873 and one also obtains @xmath874 since ( [ eq : profile ] ) grows quadratically ( [ eq : moment - inv ] ) shows that the range of @xmath875 is finite . in fact , @xmath876 maps @xmath877 to @xmath878 . and one can show that it is a complete metric defined on @xmath879 . the differing radii arise from the ambiguity in the integral ( [ eq : moment - inv ] ) . it would be interesting to obtain a closed formula for the khler potential of the metric ( [ eq : calabi ] ) . there is such a formula for the ricci - flat metric in case @xmath880 on @xmath60 due to e. calabi @xcite . since the explicit formula involves integrals it is easier to see the strongly ach nature of the metric from proposition [ prop : st - ach ] . using the classification of normal cr structures on 3-manifolds in @xcite and @xcite we are able to mostly classify those normal cr 3-manifolds which bound khler - einstein manifolds and the unique khler - einstein surfaces which thus arise . the classification of normal cr structures on 3-manifolds follows from a classification of sasaki structures on 3-manifolds which in turn follows from a classification @xcite of vaisman metrics ( or locally comformally khler metrics with a parallel lee forms ) on compact surfaces . the riemannian product of a sasaki manifold with a circle is a vaisman manifold . @xmath9 is a seifert @xmath882-bundle over a riemann surface of genus @xmath883 , @xmath73 generates the @xmath882-action , and the vaisman manifold @xmath884 is a properly elliptic surface admitting two holomorphic circle actions.[it : pr - ellip ] @xmath9 is a seifert @xmath882-bundle over an elliptic curve , @xmath73 generates the @xmath882-action , and the vaisman manifold @xmath884 is a kodaira surface admitting two holomorphic circle actions.[it : kod ] @xmath9 is a finite quotient of @xmath17 , with holomorphic coordinates @xmath885 we have @xmath886 with @xmath887 , and the vaisman manifold @xmath888 is a hopf surface of class 1 where @xmath889 is generated by the contraction @xmath890.[it : hopf ] the classification of normal cr manifolds @xmath12 is more complicated as as it involves identifying and distinguishing the underlying cr structures of the above sasaki structures . this is solved for ( [ it : pr - ellip ] ) and ( [ it : kod ] ) by showing there are no other cr reeb vector fields , so other sasaki structures are deformations as in section [ subsubsect : transv ] . an above @xmath882-seifert bundle is an @xmath882 subbundle of a negative orbifold bundle @xmath142 over a riemann surface @xmath19 . suppose it has multiple fibers over @xmath891 of multiplicities @xmath892 . we can classify the multiple fibers by @xmath893 , where there is a neighborhood @xmath340 of @xmath894 and locally @xmath19 is the quotient of @xmath895 by the @xmath896-action generated by @xmath897 . if @xmath898 , that is cases ( [ it : pr - ellip ] ) and ( [ it : kod ] ) in theorem [ thm : sasak - class ] , then @xmath12 is the boundary of an unique khler - einstein manifold if and only if for each @xmath899 the @xmath822 produced by the euclidean algorithm of section [ subsect : hirz - jung ] satisfy @xmath843 . [ it : quasi - reg ] if @xmath900 , case ( [ it : hopf ] ) , then we have a biholomorphism @xmath901 , with @xmath902 finite and acting freely on @xmath903 . then @xmath904 is a rational singularity , thus the exceptional set of the minimal resolution @xmath905 is a tree of rational curves @xmath906 . this is the unique khler - einstein manifold with boundary @xmath12 if and only if @xmath907.[it : sphere ] for case ( [ it : quasi - reg ] ) @xmath9 is a seifert @xmath882 subbundle of a negative orbifold bundle @xmath142 over a riemann surface @xmath19 of genus @xmath908 . either we have a regular sasaki structure and @xmath142 is smooth , or there are multiple fibers and corresponding singularities @xmath909 each of these can be resolved as in section [ subsect : hirz - jung ] , altogether giving the unique minimal resolution @xmath416 . it is clear that @xmath910 is necessary from the adjunction formula @xmath911 . if @xmath832 denotes the zero section of @xmath142 , then @xmath912 by grauert s criterion for an exceptional curve ( * ? ? * p. 367 ) @xmath913 , so @xmath914 . therefore by assumption @xmath915 for each irreducible exceptional curve . and if @xmath916 denotes the exceptional set , the argument in the proof of the lemma on p. 347 of @xcite shows that @xmath917 . we have @xmath918 , minus the zero section . then , after a possible homothetic change of @xmath73 , the radial @xmath81 on @xmath10 is given by @xmath639 for an hermitian metric @xmath151 on @xmath142 . let @xmath919 be an holomorphic section which vanishes on @xmath920 . define the @xmath29-form on the total space of @xmath142 @xmath921 then the @xmath922-form @xmath923 satisfies @xmath924 and has a pole of order 1 on the zero section . strictly speaking , we need to take this local construction to the power @xmath925 to get a true meromorphic section @xmath926 . so @xmath927 is a meromorphic section with zeros along @xmath928 and poles of varying orders on the exceptional curves @xmath815 . then the arguments in example 1 above or in the proof of satz 4 in ( * ? ? ? * p. 367 ) show that @xmath929 . in case ( [ it : sphere ] ) , @xmath901 follows from theorem [ thm : sasak - class].[it : hopf ] . it is well known that @xmath835 has only rational singularities , and the properties of the exceptional curves follow from well known properties of rational surface singularities ( cf . @xcite ) . if the exceptional curves satisfy @xmath930 the arguments in part ( [ it : quasi - reg ] ) show what @xmath929 . wolf p. barth , klaus hulek , chris a. m. peters , and antonius van de ven . , volume 4 of _ ergebnisse der mathematik und ihrer grenzgebiete . 3 . folge . a series of modern surveys in mathematics [ results in mathematics and related areas . 3rd series . a series of modern surveys in mathematics]_. springer - verlag , berlin , second edition , 2004 . l. boutet de monvel . intgration des quations de cauchy - riemann induites formelles . in _ sminaire goulaouic - lions - schwartz 19741975 ; quations aux derives partielles linaires et non linaires _ , pages exp . centre math . , cole polytech . , paris , 1975 . charles p. boyer , krzysztof galicki , and santiago r. simanca . the sasaki cone and extremal sasakian metrics . in _ riemannian topology and geometric structures on manifolds _ , volume 271 of _ progr . _ , pages 263290 . birkhuser boston , boston , ma , 2009 . ngaiming mok and shing - tung yau . completeness of the khler - einstein metric on bounded domains and the characterization of domains of holomorphy by curvature conditions . in _ the mathematical heritage of henri poincar , part 1 ( bloomington , ind . , 1980 ) _ , volume 39 of _ proc . . pure math . _ , pages 4159 . soc . , providence , ri , 1983 . louis nirenberg . . american mathematical society , providence , r.i . , 1973 . expository lectures from the cbms regional conference held at the texas technological university , lubbock , tex . , may 2226 , 1972 , conference board of the mathematical sciences regional conference series in mathematics , no tadao oda . , volume 15 of _ ergebnisse der mathematik und ihrer grenzgebiete ( 3 ) [ results in mathematics and related areas ( 3)]_. springer - verlag , berlin , 1988 . an introduction to the theory of toric varieties , translated from the japanese . miles reid . young person s guide to canonical singularities . in _ algebraic geometry , bowdoin , 1985 ( brunswick , maine , 1985 ) _ , volume 46 of _ proc . pure math . _ , pages 345414 . soc . , providence , ri , 1987 . g. tian and s .- t . existence of khler - einstein metrics on complete khler manifolds and their applications to algebraic geometry . in _ mathematical aspects of string theory ( san diego , calif . , 1986 ) _ , volume 1 of _ adv . _ , pages 574628 . world sci . publishing , singapore , 1987 .	extending the results of s. y. cheng and s .- t . yau it is shown that a strictly pseudoconvex domain @xmath0 in a complex manifold carries a complete khler - einstein metric if and only if its canonical bundle is positive , i.e. admits an hermitian connection with positive curvature . we consider the restricted case in which the cr structure on @xmath1 is normal . in this case @xmath2 must be a domain in a resolution of the sasaki cone over @xmath1 . we give a condition on a normal cr manifold which it can not satisfy if it is a cr infinity of a khler - einstein manifold . we are able to mostly determine those normal cr 3-manifolds which can be cr infinities . we give many examples of khler - einstein strictly pseudoconvex manifolds on bundles and resolutions . in particular , the tubular neighborhood of the zero section of every negative holomorphic vector bundle on a compact complex manifold whose total space satisfies @xmath3 admits a complete khler - einstein metric .
the insurgence of critical electric fields in the process of gravitational collapse leading to vacuum polarization process @xcite has convinced us of the necessity of critically reexamining the gravitational and electrodynamical properties in neutron stars . in this light we have recently generalized the feynman , metropolis and teller treatment of compressed atoms to the relativistic regimes @xcite . we have so enforced , self - consistently in a relativistic thomas - fermi equation , the condition of @xmath0-equilibrium extending the works of v. s. popov @xcite , ya . b. zeldovich and v. s. popov @xcite , a. b. migdal et al . @xcite , j. ferreirinho et al . @xcite and r. ruffini and l. stella @xcite for heavy nuclei . thanks to the existence of scaling laws ( see @xcite and @xcite ) this treatment has been extrapolated to compressed nuclear matter cores of stellar dimensions with mass numbers @xmath2 or @xmath3 . such configurations fulfill global but not local charge neutrality . they have electric fields on the core surface , increasing for decreasing values of the electron fermi energy @xmath4 reaching values much larger than the critical value @xmath5 , for @xmath6 . the assumption of constant distribution of protons at nuclear densities simulates , in such a treatment , the confinement due to the strong interactions in the case of nuclei and heavy nuclei and due to both the gravitational field and the strong interactions in the case of nuclear matter cores of stellar sizes . in this article we introduce explicitly the effects of gravitation by considering a general relativistic system of degenerate fermions composed of neutrons , protons and electrons in @xmath0-equilibrium : this is the simplest nontrivial system in which new electrodynamical and general relativistic properties of the equilibrium configuration can be clearly and rigorously illustrated . we first prove that the condition of local charge neutrality can never be implemented since it violates necessary conditions of equilibrium at the microphysical scale . we then prove the existence of a solution with global , but not local , charge neutrality by taking into account essential gravito - electrodynamical effects . first we recall the constancy of the general relativistic fermi energy of each specie pioneered by o. klein @xcite . we subsequently introduce the general relativistic thomas - fermi equations for the three fermion species fulfilling relativistic quantum statistics , governed by the einstein - maxwell equations . the solution of this system of equations presents a formidable mathematical challenge in theoretical physics . the traditional difficulties encountered in proving the existence and unicity of the solution of the thomas - fermi equation @xcite are here enhanced by the necessity of solving the general relativistic thomas - fermi equation coupled with the einstein - maxwell system of equations . we present the general solution for the equilibrium configuration , from the center of the star all the way to the border , giving the details of the gravitational field , of the electrodynamical field as well as of the conserved quantities . we illustrate such a solution by selecting a central density @xmath7 , where @xmath8 g @xmath9 is the nuclear density . we point out the existence near the boundary of the core in the equilibrium configuration of three different radii , in decreasing order : @xmath10 corresponding to the vanishing of the fermi momentum of the electron component ; @xmath11 , @xmath12 corresponding to the vanishing of the fermi momentum of the proton component ; @xmath13 and @xmath14 corresponding to the radius at which the fermi momentum of neutrons vanishes : @xmath15 . we then give explicit expressions for the proton versus electron density ratio and the proton versus neutron density ratio for any value of the radial coordinate as well as for the electric potential at the center of the configuration . a novel situation occurs : the description of the pressure and density is not anylonger a local one . their determination needs prior knowledge of the global electrodynamical and gravitational potentials on the entire system as well as of the radii @xmath14 , @xmath12 and @xmath10 . this is a necessary outcome of the self - consistent solution of the eigenfunction within general relativistic thomas - fermi equation in the einstein - maxwell background . as expected from the considerations in @xcite , the electric potential at the center of the configuration fulfills @xmath16 and the gravitational potential @xmath17 . the implementation of the constancy of the general relativistic fermi energy of each particle species and the consequent system of equations illustrated here is the simplest possible example admitting a rigorous nontrivial solution . it will necessarily apply in the case of additional particle species and of the inclusion of nuclear interactions : in this cases however it is not sufficient and the contribution of nuclear fields must be taken into due account . we consider the equilibrium configurations of a degenerate gas of neutrons , protons and electrons with total matter energy density and total matter pressure @xmath18 where @xmath19 is the relativistic single particle energy . in addition , we require the condition of @xmath0-equilibrium between neutrons , protons and electrons @xmath20 where @xmath21 denotes the fermi momentum and @xmath22 is the free - chemical potential of particle - species with number density @xmath23 . we now introduce the extension to general relativity of the thomas - fermi equilibrium condition on the generalized fermi energy @xmath4 of the electron component @xmath24 where @xmath25 is the fundamental charge , @xmath26 is the coulomb potential of the configuration and we have introduced the metric @xmath27 for a spherically symmetric non - rotating neutron star . the metric function @xmath28 is related to the mass @xmath29 and the electric field @xmath30 ( a prime stands for radial derivative ) through @xmath31 thus the equations for the neutron star equilibrium configuration consist of the following einstein - maxwell equations and general relativistic thomas - fermi equation @xmath32 = - 4 \pi \alpha \hbar c \ , { \rm e}^{\nu/2 } { \rm e}^{\lambda } \bigg\ { n_p \nonumber \\ % % & & - \frac{{\rm e}^{-3 \nu/2}}{3 \pi^2}[\hat{v}^2 + 2 m_e c^2 \hat{v } - m^2_e c^4 ( { \rm e}^{\nu}-1)]^{3/2}\bigg\}\ , , \label{eq : grtf}\end{aligned}\ ] ] where @xmath33 denotes the fine structure constant , @xmath34 , @xmath35 and we have used eq . ( [ eq : electroneq ] ) to obtain eq . ( [ eq : grtf ] ) . it can be demonstrated that the assumption of the equilibrium condition ( [ eq : electroneq ] ) together with the @xmath0-equilibrium condition ( [ eq : betaeq ] ) and the hydrostatic equilibrium ( [ eq : tov ] ) is enough to guarantee the constancy of the generalized fermi energy @xmath36 for all particle species separately . here @xmath37 denotes the particle unit charge of the @xmath38-species . indeed , as shown by olson and bailyn @xcite , when the fermion nature of the constituents and their degeneracy is taken into account , in the configuration of minimum energy the generalized fermi energies @xmath39 defined by ( [ eq : olsoneq ] ) must be constant over the entire configuration . these minimum energy conditions generalize the equilibrium conditions of klein @xcite and of kodama and yamada @xcite to the case of degenerate multicomponent fluids with particle species with non - zero unit charge . if one were to assume , as often done in literature , the local charge neutrality condition @xmath40 instead of assuming the equilibrium condition ( [ eq : electroneq ] ) , this would lead to @xmath41 identically ( since there will be no electric fields generated by the neutral matter distribution ) implying via eqs . ( [ eq : betaeq ] ) and ( [ eq : tov ] ) @xmath42 thus the neutron fermi energy would be constant throughout the configuration as well as the sum of the proton and electron fermi energies but not the individual fermi energies of each component . in fig . [ fig:1 ] we show the results of the einstein equations for a selected value of the central density of a system of degenerate neutrons , protons , and electrons in @xmath0-equilibrium under the constraint of local charge neutrality . in particular , we have plotted the fermi energy of the particle species in units of the pion rest - energy . it can be seen that indeed the fermi energies of the protons and electrons are not constant throughout the configuration which would lead to microscopic instability . this proves the impossibility of having a self - consistent configuration fulfilling the condition of local charge neutrality for our system . this result is complementary to the conclusion of eq . ( 4.6 ) of @xcite who found that , at zero temperature , only a dust solution with zero particle kinetic energy can satisfy the condition of local charge neutrality and such a configuration is clearly unacceptable for an equilibrium state of a self - gravitating system . , where @xmath43 g @xmath9 denotes the nuclear density . we turn now to describe the equilibrium configurations fulfilling only global charge neutrality . we solve self - consistently eqs . ( [ eq : gab1 ] ) and ( [ eq : gab2 ] ) for the metric , eq . ( [ eq : tov ] ) for the hydrostatic equilibrium of the three degenerate fermions and , in addition , we impose eq . ( [ eq : betaeq ] ) for the @xmath0-equilibrium . the crucial equation relating the proton and the electron distributions is then given by the general relativistic thomas - fermi equation ( [ eq : grtf ] ) . the boundary conditions are : for eq . ( [ eq : gab1 ] ) the regularity at the origin : @xmath44 , for eq . ( [ eq : tov ] ) a given value of the central density , and for eq . ( [ eq : grtf ] ) the regularity at the origin @xmath45 , and a second condition at infinity which results in an eigenvalue problem determined by imposing the global charge neutrality conditions @xmath46 at the radius @xmath10 of the electron distribution defined by @xmath47 from which follows @xmath48 then the eigenvalue problem consists in determining the gravitational potential and the coulomb potential at the center of the configuration that satisfy the conditions ( [ eq : bound1])([eq : bound3 ] ) at the boundary . the solution for the particle densities , the gravitational potential , the coulomb potential and the electric field are shown in fig . ( [ fig : fig2 ] ) for a configuration with central density @xmath49 . in order to compare our results with those obtained in the case of nuclear matter cores of stellar dimensions @xcite as well as to analyze the gravito - electrodynamical stability of the configuration we have plotted the electric potential in units of the pion rest - energy and the gravitational potential in units of the pion - to - proton mass ratio . one particular interesting new feature is the approach to the boundary of the configuration : three different radii are present corresponding to distinct radii at which the individual particle fermi pressure vanishes . the radius @xmath10 for the electron component corresponding to @xmath50 , the radius @xmath12 for the proton component corresponding to @xmath51 and the radius @xmath14 for the neutron component corresponding to @xmath52 . the smallest radius @xmath14 is due to the threshold energy for @xmath0-decay which occurs at a density @xmath53 g @xmath9 . the radius @xmath12 is larger than @xmath14 because the proton mass is slightly smaller than the neutron mass . instead , @xmath54 due to a combined effect of the difference between the proton and electron masses and the implementation of the global charge neutrality condition through the thomas - fermi equilibrium conditions . for the configuration of fig . [ fig : fig2 ] we found @xmath55 km , @xmath56 km and @xmath57 where @xmath58 denotes the electron compton wavelength . we find that the electron component follows closely the proton component up to the radius @xmath12 and neutralizes the configuration at @xmath10 without having a net charge , contrary to the results e.g in @xcite . bottom panel : proton and electron coulomb potentials in units of the pion rest - energy @xmath59 and @xmath60 respectively and the proton gravitational potential in units of the pion mass @xmath61 where @xmath62.,title="fig : " ] @xmath9 . bottom panel : proton and electron coulomb potentials in units of the pion rest - energy @xmath59 and @xmath60 respectively and the proton gravitational potential in units of the pion mass @xmath61 where @xmath62.,title="fig : " ] normalized to its value at @xmath63 . bottom panel : electric field for @xmath64 normalized to its value at @xmath63 . we have shown also the behavior of the solution of the general relativistic thomas - fermi equation ( [ eq : grtf ] ) for two different eigenvalues close to the one which gives the globally neutral configuration.,title="fig : " ] normalized to its value at @xmath63 . bottom panel : electric field for @xmath64 normalized to its value at @xmath63 . we have shown also the behavior of the solution of the general relativistic thomas - fermi equation ( [ eq : grtf ] ) for two different eigenvalues close to the one which gives the globally neutral configuration.,title="fig : " ] it can be seen from fig . [ fig : fig2 ] that the negative proton gravitational potential energy is indeed always larger than the positive proton electric potential energy . therefore the configuration is stable against coulomb repulsion . this confirms the results in the simplified case analyzed by m. rotondo et al . in @xcite . from eq . ( [ eq : olsoneq ] ) and the relation between fermi momentum and the particle density @xmath65 , we obtain the proton - to - electron and proton - to - neutron ratio for any value of the radial coordinate @xmath66^{3/2}\ , , \\ \frac{n_p ( r)}{n_n(r ) } & = & \left [ \frac{g^2(r ) \mu^2_n(r)-m^2_p c^4}{\mu^2_n(r)-m^2_n c^4 } \right]^{3/2}\ , , \end{aligned}\ ] ] where @xmath67 , @xmath68 and the constant values of the generalized fermi energies are given by @xmath69 a novel situation occurs : the determination of the quantities given in eqs . ( [ eq : nenpratio ] ) and ( [ eq : efi ] ) necessarily require the prior knowledge of the global electrodynamical and gravitational potential from the center of the configuration all the way out to the boundary defined by the radii @xmath10 , @xmath12 and @xmath14 . this necessity is an outcome of the solution for the eigenfunction of the general relativistic thomas - fermi equation ( [ eq : grtf ] ) . from the regularity condition at the center of the star @xmath45 together with eq . ( [ eq : nenpratio ] ) we obtain the coulomb potential at the center of the configuration @xmath70\ , , \end{aligned}\ ] ] which after some algebraic manipulation and defining the central density in units of the nuclear density @xmath71 can be estimated as @xmath72 ^ 2 } \bigg ] \nonumber \\ & \simeq & \frac{1}{2}\bigg[\frac{(3 \pi^2 \eta/2)^{2/3 } m_p}{(3 \pi^2 \eta/2)^{2/3 } m_\pi+ m^2_n / m_\pi}\bigg ] m_\pi c^2\ , , \end{aligned}\ ] ] where we have approximated the gravitational potential at the boundary as @xmath73 . then for configurations with central densities larger than the nuclear density we necessarily have @xmath74 . in particular , for the configuration we have exemplified with @xmath75 in fig . [ fig : fig2 ] , from the above expression ( [ eq : v0app ] ) we obtain @xmath76 . this value of the central potential agrees with the one obtained in the simplified case of nuclear matter cores with constant proton density @xcite . we have proved in the first part of this letter that the treatment generally used for the description of neutron stars adopting the condition of local charge neutrality , is not consistent with the einstein - maxwell equations and microphysical conditions of equilibrium consistent with quantum statistics ( see fig . [ fig:1 ] ) . we have shown how to construct a self - consistent solution for a general relativistic system of degenerate neutrons , protons and electrons in @xmath0-equilibrium fulfilling global but not local charge neutrality . although the mass - radius relation in the simple example considered here in our new treatment , differs slightly from the one of the traditional approaches , the differences in the electrodynamic structure are clearly very large . as is well - known these effects can lead to important astrophysical consequences on the physics of the gravitational collapse of a neutron star to a black hole @xcite . having established in the simplest possible example the new set of einstein - maxwell and general relativistic thomas - fermi equations , we now proceed to extend this approach when strong interactions are present @xcite . the contribution of the strong fields to the energy - momentum tensor , to the four - vector current and consequently to the einstein - maxwell equations have to be taken into account . clearly in this more general case , the conditions introduced in this letter have to be still fulfilled : the @xmath77-independence of the generalized fermi energy of electrons and the fulfillment of the general relativistic thomas - fermi equation @xcite . in addition , the generalized fermi energy of protons and neutrons will depend on the nuclear interaction fields . the fluid of neutrons , protons and electrons in this more general case does not extend all the way to the neutron star surface but is confined to the neutron star core endowed with overcritical electric fields , in precise analogy with the case of the compressed nuclear matter core of stellar dimension described in @xcite .	we present the self - consistent treatment of the simplest , nontrivial , self - gravitating system of degenerate neutrons , protons and electrons in @xmath0-equilibrium within relativistic quantum statistics and the einstein - maxwell equations . the impossibility of imposing the condition of local charge neutrality on such systems is proved , consequently overcoming the traditional tolman - oppenheimer - volkoff treatment . we emphasize the crucial role of imposing the constancy of the generalized fermi energies . a new approach based on the coupled system of the general relativistic thomas - fermi - einstein - maxwell equations is presented and solved . we obtain an explicit solution fulfilling global and not local charge neutrality by solving a sophisticated eigenvalue problem of the general relativistic thomas - fermi equation . the value of the coulomb potential at the center of the configuration is @xmath1 and the system is intrinsically stable against coulomb repulsion in the proton component . this approach is necessary , but not sufficient , when strong interactions are introduced . neutron star electrodynamics , general relativistic thomas - fermi treatment .
unlike pristine noble metal or graphite basal - plane electrodes used in classical electric double layer ( edl ) studies,@xcite lithium ion battery ( lib ) electrodes generally exhibit complex interfaces.@xcite both electron ( @xmath5 ) and li@xmath0 transport can occur inside lib electrodes . in addition , solid thin films , on the order 1 - 50 nm thick , are ubiquitous on lib electrode surfaces and can become part of the edl ( fig . [ fig1]a - h ) . the edl is critical because key battery processes like li@xmath0 incorporation kinetics ( including li@xmath0 desolvation ) , parasitic reactions , li - plating , and degradation - inducing phase transformations on electrode surfaces most likely initiate within it . indeed , `` solid electrolyte interphase '' ( sei ) films@xcite covering anodes are relied upon as kinetic barriers that prevent thermodynamic equilibrium in batteries ; they are critical for lib which operate outside the redox stability window of the organic solvent electrolytes used . as discussed below , the edl is also crucial in computational work because it effectively determines the electrode potential ( henceforth `` applied voltage '' or simply `` voltage '' ) in the simulation cell . our work examines edl associated with thin - film coated electrode surfaces , and emphasizes the importance of creating electrode interface models that exhibit consistent electrochemical activities for both li@xmath0 and @xmath5 . examples of solid films on electrode surfaces include li@xmath3co@xmath1 layers formed on pristine cathode oxide surfaces;@xcite cathode - coating films made of electrolyte decomposition products;@xcite sei films on anodes arising from reductive decomposition of liquid electrolyte components;@xcite artificial protective / passivating coatings,@xcite including atomic layer deposition ( ald ) layers@xcite which can undergo phase transformations at low voltages;@xcite ald layers between solid electrolytes and electrodes in all - solid state batteries;@xcite and even li@xmath3o@xmath3 films deposited on cathodes during fast discharge of li - air batteries,@xcite the re - oxidation of which is accompanied by significant voltage hysteresis and is a root cause of the lack of a suitable liquid electrolyte in li - air batteries . detailed atomic lengthscale understanding of the interfaces and edl associated with such thin films has remained elusive , partly due to challenges in imaging buried interfaces at sub - nanometer resolution . even edls associated with liquid organic electrolyte in batteries have arguably only recently received systematic experimental studies.@xcite modeling efforts have made much progress in elucidating the structures of solid - solid@xcite and solid - vacuum@xcite interfaces . however , voltage dependences , particularly in processes involving @xmath5 transfer like parasitic reactions and undesirable metal plating , have arguably received less theoretical attention . in this work , we apply electronic density functional theory ( dft ) calculations to investigate how voltages affect the structures and stoichiometries of thin solid films , their interfaces , and their edl at atomic lengthscales . in view of the complexities of lib electrodes , we have adopted simple model systems . the liquid electrolyte is omitted , although a few solvent molecules are included as frozen monolayers in some models to illustrate their huge impact on the voltage . au(111 ) surfaces are adopted as model cathodes , instead of transition metal oxides typically found in batteries . au does not alloy with li under voltages considered in this work and is a convenient inert electrode . these systems might be realized in ultra - high vacuum ( uhv ) settings ; they dovetail with the use of copper@xcite and gold@xcite electrodes in recent fundamental battery science experimental studies . the model thin films examined in this work , in order of increasing complexity , are li@xmath1po@xmath2 ( 010 ) ( fig . [ fig2]a),@xcite li@xmath3co@xmath1 basal plane ( fig . [ fig2]b),@xcite and li@xmath4mn@xmath3o@xmath2 ( 111 ) ( fig . [ fig2]c).@xcite these are coated on au(111 ) on one side and face a vacuum region on the other . li@xmath1po@xmath2 is a non - redox - active solid electrolyte . it illustrates the correlation between interfacial dipole densities and voltages . li@xmath3co@xmath1 is often found on as - synthesized cathode surfaces . in lib studies , there is disagreement@xcite concerning whether li@xmath3co@xmath1 dissolves upon soaking in electrolyte,@xcite is formed at high voltages,@xcite or is removed at @xmath64 v.@xcite in li - air batteries , li@xmath3co@xmath1 is generally accepted to be oxidized and removed above @xmath74.5 v.@xcite our goal is not to elucidate the detailed reaction mechanism , but to study the electronic and structural signatures accompanying voltage increase . redox - active limn@xmath3o@xmath2 ( spinel lmo ) is not used as nanometer - thick films in batteries , but its inclusion here helps illustrate the diversity of interfacial behavior ( fig . [ fig1 ] ) . like most lib cathode materials , lmo is a polaron conductor because mn can exhibit multiple redox states . it also conducts li@xmath0 , but is not a band ( `` metallic '' ) conductor . by classical electrostatics definition , the interior of lmo is not instantaneously at a constant potential.@xcite our lmo / au interface model provides a well - defined fermi level ( @xmath8 ) to demonstrate how voltage control is achieved at atomic lengthscales , and how this affects ec oxidative decomposition on lmo surfaces.@xcite out of necessity , we critically examine the voltage calibration method used in the theortical lib literature ( sec . [ voltage ] ) . the same experimental voltage governs li@xmath0 and @xmath5 motion . indeed , in cyclic voltametry , it is difficult to distinguish currents arising from faradaic li - insertion or parasitic processes that involve undesired @xmath5 transfer to the liquid . but what is `` voltage '' at atomic lengthscales ? the absolute electrochemical potential for a charged particle in an atomic configuration is the ( free ) energy needed to transfer it into a given phase from infinity ( see , e.g. , ref . ) . there can be differences whether the particle is a li@xmath0 or a @xmath5 . in dft modeling , the electronic voltage ( @xmath9 ) governing electronic motion , and ionic voltage ( @xmath9 ) governing li@xmath0 motion , can and must be distinguished . first we focus on @xmath9 . under constant voltage experimental conditions , potentiostats directly control the @xmath5 ( not li ) content . when the phase is a metallic conductor with a well - defined @xmath8 , the voltage asociated with transferring @xmath5 from infinity , multiplied by the electronic charge , is equal to the work function ( @xmath10 ) modulo a constant , just like in photovoltaics ( pv)@xcite and electrocatalytic systems.@xcite @xmath10=@xmath11 , where @xmath12 is the vacuum level and @xmath8 is set to zero herein without loss of generality . if the electrode is immersed in liquid electrolyte , @xmath10 calculated for the electrode covered by a sufficiently thick liquid layer , divied by the electronic charge , also gives @xmath9.@xcite @xmath9 applies even under out - of - equilibrium conditions . it is readily computed in our model systems because vacuum regions exist and qualify as `` infinity . '' in commercial batteries , there is no vacuum . nevertheless , our focus on @xmath10 is formally correct and emphasizes the synergy between batteries , fuel cells,@xcite pv,@xcite and metal - semiconductor interfaces.@xcite in those research areas , dipoles at interfaces are recognized as properties of paramount importance . at battery liquid - solid interfaces , edl s are the manifestations of complex surface dipole distributions associated with applied voltages ; it is in effect our computational potentiostat . based on our calculations , we propose the manipulation of surface dipoles as a novel electrode passivating strategy ( fig . [ fig1]b ) . li@xmath0 ions are only indirectly controlled by potentiostats ; they slowly redistribute in response to changes in electronic configurations . we define the li `` voltage '' ( @xmath13 ) as the ( free ) energy difference between li chemical potential in a system and the lithium metal cohesive energy per li divided by @xmath14 . this definition is widely used as the sole voltage estimate in the lib theory literature.@xcite at equilibrium , when the voltage is pinned at the redox potential of li - insertion reactions , @xmath9 must be equal to @xmath13 . thus @xmath13 should be considered a self - consistent criterion at interfaces . by itself , @xmath13 can give incorrect voltages when modeling parasitic processes induced by @xmath5 transfer , and it may not govern the dynamics of li@xmath0 insertion processes adequately . fig . [ fig3]a - d offers a preview of our results , pertinent to issues discussed above . the rest of the paper is organized as follows . 2 discusses and reconciles different definitions of voltage in dft calculations in more detail . 3 describes the methods used . the results are given in sec . 4 and sec . 5 summarizes the paper . a supporting information ( s.i . ) document provides further details about our three model systems , compares the modeling of battery interfaces with geochemical@xcite and other liquid - oxide interfaces,@xcite and discusses li@xmath3co@xmath1 oxidation thermodynamics . experimentalist readers are encouraged to skip to sec . [ results ] . dft calculations are conducted at constant number of electrons , not constant electrochemical potential . estimating and maintaining voltages in periodically replicated , condensed phase dft simulation cells have long been recognized as a challenge.@xcite recently we have applied an _ ab initio _ molecular dynamics ( aimd)-based li@xmath0 transfer free energy method to predict voltage dependences of redox processes at interfaces between liquid electrolytes and lithium - intercalated graphite.@xcite aimd has also been coupled with continuum approximations of liquid electrolytes to model constant voltage conditions on pristine electrodes.@xcite but present - day aimd simulation time scales are too short to extend to li@xmath0 motion _ inside _ lib electrodes or the surface films coating them except at significantly elevated temperatures . this work focuses on how applied voltages affect solid films . thus we model solid interfaces at zero temperature . with sufficient equilibration time and in the presence of li@xmath0 and @xmath5 reservoirs , the total number of electrons ( @xmath15 ) and li@xmath0 content ( @xmath16 ) in an electrode in a lib are governed by the chemical potentials for electrons ( @xmath17 ) and li@xmath0 ions ( @xmath18 ) , respectively . @xmath8 of a metallic electrode is @xmath17 modulo a constant . here we have not used @xmath19 or @xmath20 , the notations for total electrochemical potentials of @xmath5 and li@xmath0,@xcite precisely because constant - charge dft calculations and terminologies are being applied to infer constant - potential electrochemical properties . surface potential contributions to @xmath21 s should naturally be present in our simulation cells , which feature explicit interfaces . the electronic voltage of an atomic configuration which is band conductor , imposed by a potentiostat , not necessarily at equilibrium with respect to @xmath22 or the li@xmath0 occupation sites , is defined as @xmath23 if the li@xmath0/li(s ) reference is used.@xcite in the literature , values of 1.39 v and 1.44 v have been adopted ; the small discrepancy is a measure of possible systematic error . in the dft formulation , the @xmath5 configuration finds its ground state instantaneously . @xmath9 pertains to infinitesimal changes in @xmath15 , which leaves the system uncharged . note that refs . and deal with @xmath9 despite the fact that li@xmath0 is transferred , because the excess @xmath5 is left on the metallic electrode at its fermi level . as stated in ref . , the free energy of that graphite edge - plane system has yet to be optimized with respect to the surface li - content . in the present work , @xmath9 are always reported for locally optimized atomic configurations so that all forces are zero . in the `` lithium cohesive energy '' ( @xmath13 ) approach , discrete and matched numbers of @xmath5 and li@xmath0 are added / subtracted simultaneously : @xmath24 the last equality holds when the system approaches infinite size . here @xmath25 is the zero temperature total energy after optimization of all interior degrees of freedom . the @xmath26 k voltage is then given by @xmath27 where @xmath28 is the cohesive energy of lithium metal , predicted to be 1.57 ev using the dft / pbe functional . no net charge is introduced . there is an ambiguity concerning whether @xmath29 should be @xmath30 or @xmath31 . we have chosen the latter definition ( eq . [ mu_li ] ) . @xmath32 is used instead of @xmath18 to control @xmath22 because the former is the widely used convention , and because it is harder to control charged particles due to the periodic boundary condition adopted . in the literature , @xmath13 is generally averaged over a wider range of li content than in the present work.@xcite @xmath9 and @xmath13 are not necessarily equal . a transparent example is a pristine li metal slab in vacuum . @xmath13=0 by definition . yet the work function of li(100 ) is reported to be 2.93 ev.@xcite this translates into an electronic voltage of @xmath9=1.51 v ( eq . [ instant ] ) ! indeed , if the li surface is covered with a poor sei that allows @xmath5 transfer but li@xmath0 transfer is slowed to beyond experimental timescales the opposite of what a good sei does the lithium slab acts as an @xmath5 emitter , the voltage of which must be governed by its @xmath8 . for li(s ) to truly exhibit 0 v vs. the lib li@xmath0/li(s ) reference in @xmath5 transfer processes , net surface charge densities and electrode - electrolyte interface contributions must be accounted for . organic solvent decomposition can occur on uncharged li ( 100 ) surface under uhv conditions despite the relatively high ( 1.51 v ) potential , but in these reactions the metal surface is a reactant and li@xmath0 are produced . \(1 ) @xmath9=@xmath13 at equilibrium . under this condition , @xmath13 controls the li content . . [ relax ] should be interpreted as a self - consistency criterion . @xmath32=(@xmath10 - 1.37 ev+1.57 ev ) from eqs . [ instant ] , [ relax ] , and the pbe estimate of @xmath33 . identifying @xmath10 as @xmath17 , eq . [ mu_li ] implies that @xmath18 becomes a voltage - independent 0.20 ev . this reflects that li@xmath0 is in excess in an infinite reservoir . a related point has been made in the fuel cell literature.@xcite in that case h@xmath0 concentration at the interface , governed by the ph , is also strongly coupled to the predicted voltage . \(2 ) when @xmath13@xmath34@xmath9 , and the voltage is held fixed by the potentiostat , and li@xmath0 and @xmath5 sources are present , the li content should spontaneously decrease ( as permitted by kinetics ) so that @xmath13 rises towards @xmath9 . but @xmath13 does not always increase with decreasing @xmath22 . going back to the lithium solid example : if the potentiostat is set at 1.51 v in solution , the entire li slab dissolves . if instead an electrically disconnected li solid is dipped into an electrolyte , some li@xmath0 ionizes , leaving a negatively charge , possibly sei - covered , li surface . now it is the electronic voltage @xmath9 which decreases to 0 v vs. li@xmath0/li(s ) to coincide with @xmath13 . \(3 ) when @xmath13@xmath6@xmath9 and the voltage is held fixed , the li content should in general increase to lower @xmath13 . this may not occur in all cases . for example , when all available li@xmath0 insertion sites are filled in fully lithiated lic@xmath35 , the electrode becomes supercapacitor - like . @xmath9 now controls the net surface electronic charge,@xcite and @xmath13 ceases to matter . conducting simulations under such non - equilibrium conditions are crucial to understanding many phenomena in batteries . an example is the kinetics of metal plating@xcite at overpotentials an explicitly non - equilibrium phenomenon , a major safety issue in lib , but also a critical process for batteries featuring rechargeable metal anodes . this discussion highlights the primacy of @xmath9 when computing voltages . [ relax ] was originally applied to bulk crystalline electrode simulation cells , and @xmath13 has been identified as the redox potential.@xcite li should remain in the solid when the applied potential is lower than @xmath13 . in these special cases of simulation cells without interfaces , the li(s ) cohesive energy approach ( @xmath13 ) is _ neither consistent nor inconsistent _ with @xmath9 . this is because the electrostatic potential at any point in space is only defined to within a constant in the interior of a crystal;@xcite therefore neither @xmath9 nor @xmath8 is well - defined without specifying interfaces . thus metals terminated in different crystal facets exhibit different @xmath10s.@xcite much more significant changes in @xmath10 can be induced using adsorbed molecules.@xcite using dft+u or hybrid dft functional methods , @xmath13 has generally been predicted to be in good agreement with measured open circuit voltages in bulk transition metal oxide simulation cells . the errors are at most a few hundred mevs.@xcite this suggests that interfacial contributions may be small on most battery cathode surfaces . our work suggests that artificial interfaces can be potentially be engineered to give larger , beneficial effects . molecular adsorption effects will be emphasized in sec . [ li2co3_ec ] . dft calculations of interfaces are usually performed in finite - sized , periodically - replicated simulation cells kept at overall charge neutrality . in the literature , in dft calculations without ions in the electrolyte ( or without any electrolyte at all ) , the @xmath13 approach has therefore assigned a wide range of voltages to lib interfaces at charge neutrality . in other words , a continuous range of potentials - of - zero - charge ( pzc ) have been assigned by dft . yet simple `` classical '' electrodes such as pristine noble metals usually permit one pzc;@xcite at voltages away from pzc , the surface is charged . if liquid electrolyte is present , the electrode surface can exhibit non - zero charges , compensated by counterions in the electrolyte . in this work , we illustrate this effect by adding pf@xmath36 anions under uhv conditions . dft calculations are conducted using the vienna atomic simulation package ( vasp ) version 5.3@xcite and the pbe functional.@xcite modeling spinel li@xmath4mn@xmath3o@xmath2 requires spin - polarized dft and the dft+u augmented treatment@xcite of mn @xmath37 orbitals . the @xmath38 and @xmath39 values depend on the orbital projection scheme and dft+u implementation details ; here @xmath404.85 ev is chosen to taken from the literature.@xcite the charge state on mn ions are determined from the approximate local spin polarzation @xmath41 reported by the vasp code ; @xmath42=3/2 , 4/2 , and 5/2 are assigned to mn(iv ) , mn(iii ) , and mn(ii ) , respectively . for li@xmath1po@xmath2 and li@xmath3co@xmath1 , most calculations exclude spin - polarization . a few spin - polarized calculations are performed to confirm that this is adequate . a 400 ev planewave energy cutoff and a 3@xmath433@xmath431 monkhorst - pack grid are applied in all cases . increasing the cutoff or the @xmath44-grid changes the predicted absolute energies by at most 0.1 ev ; energy differences , which are the relevant quantities , are much smaller . our model systems are asymmetric slabs : li@xmath3co@xmath1 basal plane , li@xmath1po@xmath2 ( 010 ) , and limn@xmath3o@xmath2 ( 111 ) slabs on one side of 4-layer thick au(111 ) ( fig . [ fig2 ] ) . the standard dipole correction is applied to negate image interactions in the periodically replicated , charge - neutral simulation cells.@xcite au is a metallic conductor and its @xmath8 is well defined . work functions ( @xmath10 ) are the differences between @xmath8 and vacuum levels . two vacuum - surface interfaces exist per asymetric slab , and two vacuum levels , for bare and coated au(111 ) respectively , are obtained.@xcite four layers of au atoms do not give a completely converged @xmath10 for uncoated au(111 ) . therefore we have shifted the two computed @xmath10 of each system by the difference between the predicted au(111 ) @xmath10 , and the fully converged au(111 ) @xmath10=5.15 ev.@xcite the shift is at most 0.1 ev . we also report the net surface dipole density ( @xmath45 ) , which is the total dipole moment of the simulation cell in the direction perpendicular to the surface divided by the lateral surface area . the dipole moment is computed as the sum of all charges ( including electron density defined on a grid ) multipled by their displacement from a user - specified center of the cell , and is reported by the vasp code . @xmath45 should be independent of the cell - center in a charge - neutral simulation cell as long as the choice locates the artificial dipole layer@xcite inside the vacuum region . more details on the thin film models , and rationale for crystal facets chosen , are provided in the s.i . .[table1 ] electronic ( @xmath9 ) and ionic ( @xmath13 ) voltages of a li@xmath1po@xmath2(010)/au(111 ) interface cell as the number of li vacancies at the the inner surface of the li@xmath1po@xmath2 slab varies . exceptions : cases @xmath46 and @xmath47 involve removing a li from a middle layer or from the outer surface of li@xmath1po@xmath2 . @xmath13 is defined for incremental li subtraction , e.g. , the `` -1 '' value refers to the difference between the total energies of `` -1 '' and `` -2 . '' [ cols="^,>,>,>,>,>,>,>",options="header " , ] next we add ec to the lmo surface . if the system were a liquid - solid interface , the surface li content , net surface charge , and surface density of ec molecules which directly coordinate to available li ions should be self - consistently deduced using aimd simulations at finite temperature.@xcite here we confine ourselves to @xmath26 k and to adding one ec , yielding the surface used in ref . . @xmath48(li)=-4 in this model ( fig . [ fig9]b ) , and @xmath9=4.23 v , in reasonably agreement with @xmath13=4.19 v for this configuration ( table [ table4 ] ) . the voltage is within the experimental lmo operating range . except for the shift in @xmath9 , and hence in the vacuum level , the dos is almost unchanged from the case where no ec is present ( fig . [ fig10]a - b ) . with no anion on the surface , the edl structure should resemble fig . [ fig1]a . it is also important to elucidate the edl structure of a charged , polaron - conducting lmo surface , not considered previously.@xcite as in the li@xmath3co@xmath1 example , we impose a positive charge by optimizing the geometry of a pf@xmath36 anion on the lmo surface in the overal charge - neutral simulation cell ( fig . [ fig9]c ) . the dos ( fig . [ fig10]c ) reveals that pf@xmath36 retains its excess electron and is not oxidized.@xcite a net @xmath49 compensating charge is found in the lmo slab , on a second layer mn ion ( `` mn 14 '' ) below the lmo / vacuum interface . this layer contains mn(ii ) ions , exchanged with li@xmath0 in the persson group s reconstruction,@xcite which may be the reason a mn(iii ) ion there can lose an @xmath5 most readily . the dipole moment created by the mn(iv)-pf@xmath36 charge separation raises @xmath9 dramatically , from 4.25 v to 6.41 v. comparing the dos with ( fig . [ fig10]b ) and without ( fig . [ fig10]c ) the pf@xmath36 , an occupied @xmath50-orbital on `` mn 14 '' is indeed seen to be depopulated ; its orbital energy level now moves above @xmath8 . the edl in this redox - active material now resembles fig . [ fig1]h . a lower surface anion density will yield smaller shifts @xmath9 but will require a simulation cell with a much larger lateral surface area . finally , we re - examine ec decomposition on the @xmath48(li)=@xmath514 lmo slab , now with au(111 ) present . [ fig9]d depicts a partially decomposed ec molecule on this lmo surface . a h@xmath0 is transferred to the lmo slab without breaking any c - o bond in the ec . its energy is + 0.35 ev above the intact ec configuration ( fig . [ fig9]b ) , in good agreement with that of the reaction intermediate depicted in fig . 5c of ref . where the au slab was absent . a subsequent , c - o bond - breaking step is needed to render ec oxidation favorable.@xcite examining the spin states reveals that a mn(iv ) ion ( `` mn 16 '' ) immediately below the oxide surface gains an @xmath5 to become a mn(iii ) . consistent with this analysis , the dos shows an unoccupied mn @xmath50-orbital ( fig . [ fig10]b for intact ec ) becoming occupied ( fig . [ fig10]d for broken ec ) , moving from above @xmath8 to below it due to polaronic relaxation.@xcite some orbitals on the decomposed ec fragment reside near @xmath8 , suggesting further oxidation can readily occur . regardless of whether an @xmath5 is added to lmo from an oxidized ec , or is removed by imposing a @xmath49 charge due to a counter - anion nearby , only mn redox states change . charge localization on mn ions , accompanied by polaronic relaxation and a compensating pf@xmath36 or ec@xmath0 fragment nearby , ensures that @xmath5 and hole never access the fermi level which lies on au orbitals ( fig . [ fig10 ] ) . the au current collector only serves to establish @xmath8 , with which polaron formation in lmo must remain in equilibrium . therefore the effect of the applied voltage is indirect . increasing @xmath13 and @xmath9 consistently requires the loss of surface li@xmath0 ( table [ table4 ] ) , which leaves the surface with more oxygen ions that are under - coordinated , and increases its reactivity towards ec molecules . adsorbed pf@xmath36 ions also increases the voltage . since @xmath13 controls the li content at equilibrium in this voltage range ( li - sites are not fully depopulated yet ) , one can arguably ignore both the au slab and @xmath9 while performing simulations of parasitic reaction on such redox - active surfaces.@xcite as discussed above , the same reaction intermediate energetics is predicted with or without the au slab . however , @xmath9 also controls the pf@xmath36 surface density . anions in proximity of the electrode surface constitute spatial inhomogeneities or `` hot spots '' where parasitic or li@xmath0 insertion reactions may preferentially occur . indeed , when a pf@xmath36 is adsorbed near the ec , the deprotonated ec intermediate ( fig . [ fig9]e ) is found to be 0.1 ev more favorable than the intact ec , instead of 0.35 ev less favorable when the pf@xmath36 is absent . thus voltage - dependent anion adsorption can affect interfacial processes . if the cathode material is metallic rather than polaronic , the surface charge is more uniformly distributed,@xcite @xmath9 is more relevant , and qualitatively different behavior may be observed . both electrons and li@xmath0 can move in the surface films covering the electrodes in lithium ion battery ( lib ) , and in the electrodes themselves , in response to voltage variations . this makes lib rather unique among electrochemical devices . the structures of electric double layers ( edl ) , and the atomic lengthscale manifestations of voltages , become much more complex than on pristine noble metal electrodes . this work considers au(111 ) slabs coated with li - conducting solid electrolytes or cathode oxide materials under ultra high vacuum ( uhv ) conditions . the applied voltage is shown to be strongly correlated with the surface dipole density at the interface , reminiscent of photovoltaics studies . in relatively electrochemically inert thin films like li@xmath1po@xmath2 and li@xmath3co@xmath1 coated on au(111 ) , dipole moments responsible for voltage increase are established by creating negatively charged li@xmath0 vacancies at the interface between cathode surfaces and the thin solid films , compensated with positive charges on the cathode . the phosphate example highlights the crucial role of dipole density . the carbonate case emphasizes that organic solvent molecules like ethylene carbonate ( ec ) can dramatically modify the predicted voltage even in the absence of charged species ( salt ) . a corollary is that the poisson - boltzmann theory for liquid electrolytes , which ignores solvent dipole moments , may not quantitatively describe battery interfaces . with li@xmath4mn@xmath3o@xmath2 ( lmo ) cathode oxide thin films coated on au current collectors , charge transfer occurs at the outer lmo surface and involves changes in mn redox states . the au slab only indirectly determines the surface li content and the anion surface density . modeling of electrolyte decomposition on such electrode surfaces arguably does not require explicitly including the current collector . however , the voltage - dependent surface density of pf@xmath36 creates inhomogeneities and hot spots where fast electrolyte decomposition can occur . we also critically examine the definitions of `` applied voltage '' in dft calculations at interfaces . we distinguish two voltages : electronic ( @xmath9 ) , due to electronic motion , and ionic ( @xmath13 ) , due to li@xmath0 redistribution . in any atomic configuration , @xmath9 is well defined in a metallic electrode and should coincide with the experimental voltage imposed via a potentiostat . @xmath13 is a self - consistency critierion . at equilibrium , the system is pinned by li@xmath0 insertion redox reactions , and @xmath9=@xmath13 . it can be efficiently computed using the `` lithium metal cohesive energy '' method widely used in the battery theoretical literature , although most previous work neglects the possibility of net surface charges , compensated by ions inside the liquid electrolyte , which may affect parasitic reactions and li@xmath0 insertion kinetics . but out - of - equilibrium conditions are also critical to many battery - related phenomena such as sei formation and electroplating . here @xmath9 , which governs electron content at battery interfaces , is the correct definition of voltage . however , the work function - derived definition of @xmath9 does not fully account for spatial inhomogeneities . our solid state modeling considerations form the basis for future aimd simulations of thin film - coated electrode / liquid electrolyte interfaces . a supporting information ( s.i . ) document provides more details about the model systems used , reports calculations on li@xmath3co@xmath1(001)/au(111 ) interfaces and li@xmath3co@xmath1 bulk crystal decomposition thermodynamics considerations , and gives a brief comparison between aqueous interfaces and battery electrolyte interfaces . this material is available free of charge via the internet at http://pubs.acs.org . we thank steve harris , sang bok lee , john cumings , yue qi , peter feibelman , and ismaila dabo for useful discussions , and nitin kumar for performing prelminary co@xmath52 breakdown simulations . this work was supported by nanostructures for electrical energy storage ( nees ) , an energy frontier research center funded by the u.s . department of energy , office of science , office of basic energy sciences under award number desc0001160 . sandia national laboratories is a multiprogram laboratory managed and operated by sandia corporation , a wholly owned subsidiary of lockheed martin corporation , for the u.s . deparment of energy s national nuclear security administration under contract de - ac04 - 94al85000 . bard , a.j . ; abruna , h.d . ; chidsey , c.e . ; faulkner , l.r . ; feldberg , s.w . ; itaya , k. ; majda , m. ; melroy , o. ; murray , r.w . ; porter , m.d . ; soriaga , m.p . ; white , h.s . the electrode / electrolyte interface a status report . _ j. phys . chem . _ * 1993 * , _ 97 _ , 7147 - 7172 . hrmann , n.g . ; jckle , m. ; gossenberger , f. ; roman , t. ; forster - tonigold , k. ; naderian , m. ; sakong , s. ; gross , a. some challenges in the first - principles modeling of structures and processes in electrochemical energy storage and transfer . _ j. power sources _ * 2015 * , _ 275 _ , 531 - 538 . aurbach , d. ; gamolsky , k. ; markovsky , b. ; salitra , g. ; gofer , y. heider , u. ; oesten , r. ; schmidt , m. the study of surface phenomena related to electrochemical lithium intercalation into li@xmath4mo@xmath53 host materials ( m = ni , mn ) _ j. electrochem _ * 2000 * , _ 147 _ , 1322 - 1331 . saito , y. ; shikano , m. ; kobyayashi , h. state of charge ( soc ) dependence of lithium carbonate on lini@xmath54co@xmath55al@xmath56o@xmath3 electrode for lithium - ion batteries . _ j. power sources _ * 2011 * , _ 16 _ , 6889 - 6892 . song , s .- w . ; zhuang , g.v . ; ross , p.n . surface film formation on lini@xmath54co@xmath55al@xmath56o@xmath3 cathodes using attenuated total reflection ir spectroscopy . _ j. electrochem . soc . _ * 2004 * , _ 151 _ , a1162-a1167 . ostrovskii , d. ; ronci , f. ; scrosati , b. ; jacobsson , p. a ftir and raman study of spontaneous reactions occuring at the lini@xmath53co@xmath57o@xmath3 electrode / non - aqueous electrolyte interface _ j. power sources _ * 2001 * , _ 94 _ , 183 - 188 . eriksson , t. ; andersson , a.m. ; bishop , a.g . ; gejke , c. ; gustafsson , t. ; thomas , j.o . surface analysis of limn@xmath3o@xmath2 electrodes in carbonate - based electrolytes . _ j. electrochem . soc . _ * 2002 * , _ 149 _ , a69-a78 . simmen , f. ; hintennach , a. ; horisberger , m. ; lippert , t. ; novk , p. ; schneider , c.w . ; wokaun , a. aspects of the surface layer formation on li@xmath58mn@xmath3o@xmath2 during electrochemical cycling . _ j. electrochem . soc . _ * 2010 * , _ 157 _ , a1026 - 1029 . li , j.c . ; dudney , n.j . ; nanda , j. ; liang , c.d . artificial solid electrolyte interphase to address the electrochemical degradation of silicon electrodes . _ acs appl . . interfaces _ * 2014 * , _ 6 _ , 10083 - 10088 . jung , y.s . ; cavanagh , a.s . ; dillon , a.c . ; groner , m.d . ; george , s.m . ; lee , s.h . enhanced stability of licoo@xmath3 cathodes in lithium - ion batteries using surface modification by atomic layer deposition . _ j. electrochem . soc . _ * 2010 * , _ 157 _ , a75-a81 . jung , y.s . , cavanagh , a.s . , leah , r.a . , kang , s.h . , dillon , a.c . , groner , m.d . , george , s.m . ; lee , y.h . ultrathin direct atomic layer deposition on composite electrodes for highly durable and safe li - ion batteries . mater . _ * 2010 * , _ 22 _ , 2172 - 2176 . liu , y. ; hudak , n.s . ; huber , d.l . ; limmer , s.j . ; sullivan , j.p . ; huang , j.y . in situ transmission electron microscopy observation of pulverization of aluminum nanowires and evolution of the thin surface al@xmath3o@xmath1 layers during lithiation - delithiation cycles . _ nano lett . _ * 2011 * , _ 11 _ , 4188 - 4194 . woo , j.h . ; trevey , j.e . ; cavanaugh , a.s . ; choi , y.s . ; kim , s.c . ; george , s.m . ; oh , k.h . ; lee , s .- h . nanoscale interface modification of licoo@xmath3 by al@xmath3o@xmath1 atomic layer deposition for solid - state li batteries . _ j. electrochem . * 2012 * , _ 159 _ , a1120-a1124 . browning , j.f . ; baggetto , l. ; jungjohann , k.l . ; wang , y. ; tenhaeff , w.e . ; keum , j.k . ; wood , d.l . ; veith , g.m . in situ determination of the liquid / solid interface thickness and composition for li ion cathode limn@xmath60ni@xmath59o@xmath2 . _ acs appl . mater . interfaces _ * 2014 * , _ 6 _ , 18569 - 18576 . leenheer , a.j . ; sullivan , j.p . ; shaw , m. j. ; harris , c. t. a sealed liquid cell for in situ transmission electron microscopy of controlled electrochemical processes . _ j. microelectromech . syst . _ * 2015 * doi : 10.1109/jmems.2014.2380771 . haruyama , j. ; sodeyama , k. ; han , l. ; takada , k. ; tateyama , y. space - charge layer effect at interface between oxide cathode and sulfide electrolyte in all - solid - state lithium - ion battery . mater . _ * 2014 * , _ 26 _ , 4248 - 4255 . lepley , n.d . ; holzwarth , n.a.w . ; du , y.a . structures , li@xmath0 mobilities , and interfacial properties of solid electrolytes li@xmath1ps@xmath2 and li@xmath1po@xmath2 from first principles . b _ * 2013 * , _ 88 _ , 104103 . qian , d. ; hinuma , y. ; chen , h. ; du , l .- s . ; carroll , k.j . ; ceder , g. ; grey , c.p . ; meng , y.s . electronic spin transition in nanosize stoichiometric lithium cobalt oxide . soc . _ * 2012 * , _ 134 _ , 6096 - 6099 . chan , m.k.y . ; wolveton , c. ; greeley , j.p . first principles simulations of the electrochemical lithiation and delithiation of faceted crystalline silicon . _ j. am . _ , * 2012 * , _ 134 _ , 14362 - 14374 . wang , l. ; zhou , f. ; meng , y.s . ; ceder , g. first - principles study of surface properties of lifepo@xmath2 : surface energy , structure , wulff shape , and surface redox potential . _ b _ * 2007 * , _ 76 _ , 165435 . duan , y. ; sorescu , d.c . density functional theory studies of the structural , electronic , and phonon properties of li@xmath3o and li@xmath3co@xmath1 : application to co@xmath3 capture reaction . b _ * 2009 * , _ 79 _ , 014301 . garcia - lastra , j.m . ; myrdal , j.s.g . ; christensen , r. ; thygesen , k.s . ; vegge ; t. dft+u study of polaronic conduction in li@xmath3o@xmath3 and li@xmath3co@xmath1 : implications for li - air batteries . _ j. phys . c _ * 2013 * , _ 117 _ , 5568 - 5577 . shang , s .- l . ; hector , l.g . ; shi , s. ; qi , y. ; wang , yi . ; liu , z .- k . lattice dynamics , thermodynamics and elastic properties of monoclinic li@xmath3co@xmath1 from density functional theory . _ acta mater . _ * 2012 * , _ 60 _ , 5204 - 5216 . shi , s. ; qi , y. ; li , h. ; and hector , l.g . defect thermodynamics and diffusion mechanisms in li@xmath3co@xmath1 and implications for the solid electrolyte interphase in li - ion batteries . _ j. phys . c _ * 2013 * , _ 117 _ , 8579 - 8593 . chen , y.c . ; ouyang , c.y . ; song , l.j . ; sun , z.l . electrical and lithium ion dynamics in three main components of solid electrolyte interphase from density functional theory study . _ j. phys . c _ * 2011 * , _ 115 _ , 7044 - 7049 . several low - index li@xmath3co@xmath1 surfaces are considered by : bruno , m. ; prencipe , m. _ ab initio _ quantum - mechanical modeling of the ( 001 ) , ( @xmath6101 ) , and ( 110 ) surfaces of zabuyelite ( li@xmath3co@xmath1 ) _ sur . sci . _ * 2007 * , _ 601 _ , 3012 - 3019 . our basal plane requires a rotation about the @xmath62-axis , discussed below . working libs typically include binders and conductive carbon additives . even though the cathode oxides are li@xmath0 and polaron conductors and should achieve voltage equilibrium at long time scales , oxide electrodes that omit such additives for basic science study purposes tend to exhibit ohmic losses . see , e.g. , hirayama , m. ; ido , h. ; kim , k.s . ; cho , w. ; tamura , k. ; mizuki , j. ; kanno , r. dynamic structural changes at limn@xmath3o@xmath2/electrolyte interface during lithium battery reaction . soc . _ * 2010 * , _ 132 _ , 15268 - 15276 . klein , a. ; krber , c. ; wachau , a. ; suberlich , f. ; gassenbauer , y. ; harvey , s.p . ; proffit , d.e . ; mason , t.o . transparent conducting oxides for photovoltaics : manipulation of fermi level , work function and energy band alignment . _ materials _ * 2010 * , _ 3 _ , 4892 - 4914 . beljonne , d. ; cornil , j. ; muccioli , l. ; zannoni , c. ; bredas , j .- l . ; castet , f. electronic processes at organic - organic interfaces : insight from modeling and implications for opto - electronic devices . mater . _ * 2011 * , _ 23 _ , 591 - 609 . sai , n. ; gearba , r. ; dolocan , a. ; tritsch , j.r . ; chan , w.l . ; chelikowsky , j.r . ; leung , k. ; zhu , x.y . understanding the interface dipole of copper phthalocyanine ( cupc)/c-60 : theory and experiment . _ j. phys . lett . _ * 2012 * , _ 3 _ , 2173 - 2177 . churakov , s.v . ; labbez , c. ; pegado , l. ; sulpizi , m. intrinsic acidity of surface sites in calcium silicate hydrates and its implication to their electrokinetic properties _ j. phys . chem . c _ * 2014 * , _ 118 _ , 11752 - 11762 . letchworth - weaver , k. ; arias , t. a. joint density functional theory of the electrode - electrolyte interface : application to fixed electrode potentials , interfacial capacitances , and potentials of zero charge . b _ * 2012 * , _ 86 _ , 075140 . taylor , c.d . ; wasileski , s.a . ; filhol , j .- s . ; neurock , m. first principles reaction modeling of the electrochemical interface : consideration and calculation of a tunable surface potential from atomic and electronic structure . b _ * 2006 * , _ 73 _ , 165402 . viswanathan , v. ; hansen , h.a . ; rossmeisl , j. ; jaramillo , t.f . ; pitsch , h. ; norskov j.k . simulating linear sweep voltammetry from first - principles : application to electrochemical oxidation of water on pt(111 ) and pt@xmath1ni(111 ) . _ j. phys . c _ * 2012 * , _ 116 _ , 4698 - 4704 . leung , k. ; tenney , c.m . toward first principles prediction of voltage dependences of electrode / electrolyte interfacial processes in lithium ion batteries . _ j. phys . c _ , * 2013 * , _ 117 _ , 24224 - 24235 . xing , l. ; borodin , o. ; smith , g. ; li , w. density functional theory study of the role of anions on the oxidative decomposition reaction of propylene carbonate _ j. phys . chem . a _ * 2011 * , _ 115 _ , 13896 - 13905 . barile , c.j . ; barile , e.c . ; zavadil , k.r . ; nuzzo , r.g . ; gewirth , a.a . electrolytic conditioning of a magnesium aluminum chloride complex for reversible magnesium deposition . _ j. phys . c _ , * 2014 * , _ 118 _ , 27623 - 27630 . dudarev , s.l . ; botton , g.a . ; savrasov , s.y . ; humphreys , c.j . ; sutton , a.p . electron - energy - loss spectra and the structural stability of nickel oxide : an lsda+u study . b _ * 1998 * , _ 57 _ , 1505 - 1509 . see ref . . the coated au(111 ) work functions are almost indistinguishale from calculations where au(111 ) slabs are symmetrically coated on both sides . this verifies that no net charge resides on the bare au(111 ) surface . the spatially decomposed dos plots fig . [ fig3 ] , [ fig5 ] , [ fig7 ] , and [ fig9 ] follow refs . . they are constructed from the vasp procar output file which approximately decomposes @xmath5 spatial distributions onto individual atoms . contributions above @xmath63=0.07 and @xmath63=0.05 are included for li@xmath1po@xmath2/au and li@xmath3co@xmath1 plots . different criterion are used because @xmath5 are delocalized to different extent at different interfaces ; we stress all dos decomposition schemes are approximate . smith , g.d . ; borodin , o. ; russo , s.p . ; rees , r.j . ; hollenkamp , a.f . a molecular dyanmics simulation study of lifepo@xmath2/electrolyte interfaces : structure and li@xmath0 transport in carbonate and ionic liquid electrolytes . _ phys . chem . phys . _ * 2009 * , _ 11 _ , 9884 - 9897 . most monatomic anions like cl@xmath65 would lose an @xmath5 to the electrode under uhv conditions . fortunately , the highest occupied molecular orbital ( homo ) of pf@xmath36 lies far below the fermi level in all cases and permit this strictly illustrative test . transferring one @xmath5 to the lmo surface does change @xmath9 due to the small lateral surface area of the simulation cell . in the limiting case of one ec decomposition on an infinite lmo surface , such voltage changes should be minimal . we also note that the @xmath48(li)=@xmath514 voltage of @xmath66 v should in reality induce some loss of li from the _ bulk _ lmo regions . this effect has been omitted in ref . for simplicity ; it should make ec decomposition occur even more readily because it will produce more mn(iv ) ions in the near - surface region . jorn , r. ; kumar , r. ; abraham , d.p . ; voth , g.a . atomistic modeling of the electrodeelectrolyte interface in li - ion energy storage systems : electrolyte structuring . _ j. phys . c _ * 2013 * , _ 117 _ , 37473761 .	battery electrode surfaces are generally coated with electronically insulating solid films of thickness 1 - 50 nm . both electrons and li@xmath0 can move at the electrode - surface film interface in response to the voltage , which adds complexity to the `` electric double layer '' ( edl ) . we apply density functional theory ( dft ) to investigate how the applied voltage is manifested as changes in the edl at atomic lengthscales , including charge separation and interfacial dipole moments . illustrating examples include li@xmath1po@xmath2 , li@xmath3co@xmath1 , and li@xmath4mn@xmath3o@xmath2 thin - films on au(111 ) surfaces under ultrahigh vacuum conditions . adsorbed organic solvent molecules can strongly reduce voltages predicted in vacuum . we propose that manipulating surface dipoles , seldom discussed in battery studies , may be a viable strategy to improve electrode passivation . we also distinguish the computed potential governing electrons , which is the actual or instantaneous voltage , and the `` lithium cohesive energy''-based voltage governing li content widely reported in dft calculations , which is a slower - responding self - consistency criterion at interfaces . this distinction is critical for a comprehensive description of electrochemical activities on electrode surfaces , including li@xmath0 insertion dynamics , parasitic electrolyte decomposition , and electrodeposition at overpotentials . keywords : lithium ion batteries ; voltage prediction ; density functional theory ; computational electrochemistry epsf
a meta - problem in the theory of hyperplane arrangements starts with an algebraic object @xmath7 derived from an arrangement @xmath8 and determines the extent to which the intersection lattice @xmath9 determines @xmath7 up to isomorphism . the prototypical example of this associates to an arrangement @xmath10 the cohomology ring of the complement @xmath11 . the structure of this ring was determined by orlik and solomon @xcite it is determined by the combinatorics of no - broken - circuit subsets of the associated matroid . there is an equivariant version of the meta - problem where the arrangement @xmath8 is fixed by the action of a group @xmath12 acting linearly on @xmath13 . when this happens one wants to determine the structure of @xmath7 as a representation of @xmath12 . the prototypical example here is that of a real reflection group @xmath12 acting together with its reflection arrangement @xmath10 and one wants to understand the structure of @xmath14 as a representation of @xmath12 . for complex reflection groups this problem first studied by orlik and solomon @xcite and later by many others . of recent interest are the so - called zonotopal algebras of a hyperplane arrangement of holtz and ron @xcite . these were rediscovered in many disparate areas of mathematics and unified under the guise of power ideals by ardila and postnikov in @xcite ( which contains an extensive list of references ) . the zonotopal ideals of an arrangement @xmath10 are ideals in @xmath15 generated by powers of linear forms . specifically , the @xmath16th zonotopal ideal of @xmath8 is @xmath17 where @xmath18 is the number of hyperplanes in @xmath8 that do not contain @xmath19 . the quotient @xmath20 is the @xmath16th zonotopal algebra of @xmath8 . holtz and ron single out the cases @xmath21 as being of particular interest , and call these the internal , central and external zonotopal algebras of @xmath8 . the internal case @xmath22 exhibits dramatic subtleties in comparison to all other cases @xmath23 @xcite . when @xmath24 the zonotopal algebra no longer has structure determined by the lattice @xmath9 ( * ? ? * proposition 4 ) . the intermediary cases have not been studied to our knowledge . in this paper we study the internal zonotopal algebra of certain arrangements coming from complex reflection groups @xmath12 . for numerological reasons we do not study the naturally occurring reflection arrangement of @xmath12 , but instead its gale dual . we are specifically interested in the structure of this algebra as representations of @xmath12 . we summarize our main results here . let @xmath25 be the reflection arrangement of one of the monomial groups @xmath26 , consisting of @xmath1-by-@xmath1 permutation matrices whose entries are @xmath2th roots of unity ( @xmath27 and @xmath28 ) . let @xmath29 denote the gale dual of @xmath25 . then the degree @xmath30 component of @xmath31 is isomorphic to @xmath32 as a representation of @xmath12 , where @xmath33 is the cyclic subgroup generated by a coxeter element . we complement this result with several others on the structure of the internal zonotopal algebra @xmath31 . we study the type a case of the above theorem , when @xmath34 . based on our theorem above , it is perhaps unsurprising that the lie representation appears in type a , however our perspective makes obvious the `` hidden '' action of a larger symmetric group than is necessary on the lie representation , as observed by mathieu @xcite and robinson and whitehouse @xcite . for all @xmath0 , we give explicit generators of the internal zonotopal ideal and then use this description to prove finite generation in the sense of sam and snowden @xcite and representation stability as described by gan and li @xcite . we generalize a recurrence relation for the whitehouse representation which factorizes the regular representation of @xmath0 , @xmath27 . in type b , when @xmath35 , we use a grbner basis for the internal zonotopal ideal to compute an analog of decreasing trees , which are classic combinatorial objects . let @xmath8 be a central arrangement of hyperplanes in a complex finite dimensional vector space @xmath13 . we let @xmath36 denote the matroid of @xmath8 . recall that this is the simplicial complex on @xmath8 whose faces ( alias independent sets ) are collections of hyperplanes whose defining linear forms are linearly independent . maximal faces in @xmath37 are referred to as its bases . define a function @xmath38 by the rule @xmath39 and use this to define the ideal @xmath40 for @xmath41 the quotient @xmath42 has krull dimension zero , and so it is a finite dimensional complex vector space . the vector space dimension of this quotient is the _ degree _ of the quotient as a @xmath15 module . the ring @xmath15 is graded in the usual way , and we denote the @xmath2th graded piece of a graded module over this ring by @xmath43 . the ideal @xmath44 is graded , since it is generated by powers of linear forms , which are homogeneous . hilbert series _ of @xmath45 is the generating function for the dimensions of the graded pieces of this graded @xmath15-module . since @xmath46 for @xmath2 sufficiently large there is a least @xmath2 for which @xmath47 and we call this the _ top _ of quotient and denote it by @xmath48 to ease the proliferation of superscripts and subscripts . we will describe the hilbert series of @xmath45 in terms of the matroid of @xmath8 . to do so , we need the tutte polynomial of a matroid . we take the most expedient route . given a matroid @xmath37 with ground set @xmath49 , the _ rank _ of a subset @xmath50 of @xmath49 , is size of the largest independent set of @xmath37 contained in @xmath50 . the _ tutte polynomial _ of a matroid @xmath37 with ground set @xmath49 and rank function @xmath51 is the bivariate polynomial @xmath52 the tutte polynomial of a matroid is universal in the sense that any matroid invariant taking values in an abelian group satisfying a generalized deletion - contraction identity must be an evaluation of the tutte polynomial . for @xmath53 there is a short exact sequence relating @xmath45 to @xmath54 and @xmath55 , where @xmath56 is @xmath8 with one of its defining hyperplanes @xmath57 removed , and @xmath58 is the arrangement obtained by projecting @xmath8 into @xmath57 . when interpreted at the level of hilbert series this becomes a deletion - contraction relation and we have the following result . [ thm : ap ] let @xmath8 be a central , essential arrangement of hyperplanes in @xmath13 . let @xmath36 denote the matroid of the arrangement @xmath8 , which consists of @xmath2 hyperplanes . the hilbert series of @xmath45 is equal to 1 . @xmath59 if @xmath60 ; 2 . @xmath61 if @xmath62 ; 3 . @xmath63 if @xmath22 . the algebras @xmath45 occurring in the theorem are , respectively , referred to as the external , central and internal zonotopal algebras of @xmath8 by holtz and ron . the most subtle case of the theorem is the third @xcite . in the external and central cases , explicit bases of the _ macaulay inverse system _ of @xmath44 can be given @xcite in terms of independent sets of the matroid @xmath37 with a given external activity . in the internal case , the analogous results fail @xcite and there is no known canonical basis described by the matroid of @xmath8 . this is particularly striking for the top degree component of @xmath45 , whose dimension is @xmath64 the number of bases of @xmath37 with internal activity zero . the number @xmath64 is also the reduced euler characteristic of the independent set complex of @xmath37 . finding a combinatorial basis for @xmath65 or its macaulay inverse system is a tantalizing open problem . a smaller generating set for @xmath44 suffices than the one given when @xmath66 . to describe the smaller generating set , we recall that a _ line _ of the arrangement @xmath8 is a one - dimensional intersection of hyperplanes in @xmath8 . [ thm : line ] let @xmath8 be a central , essential arrangement of hyperplanes in @xmath13 . the ideal @xmath67 is equal to @xmath44 for @xmath66 . a comprehensive treatment of reflection groups in unitary spaces is lehrer and taylor s book . let @xmath13 be a complex vector space . a ( generalized ) reflection is an element of @xmath68 that fixes a hyperplane point - wise and has finite order . a complex reflection group is a finite subgroup of @xmath68 generated by reflections . such groups have been classified by shephard and todd , and serre and chevellay . there appears a single infinite family of groups @xmath69 , @xmath70 , called the monomial groups as well as @xmath71 exceptional groups . the monomial group @xmath69 consists of @xmath1-by-@xmath1 permutation matrices whose entries are @xmath72-th roots of unit , the product of which is a @xmath73-th root of unity . associated to a complex reflection group @xmath12 is its reflection arrangement @xmath74 . this is the arrangement of hyperplanes that are fixed by the reflections generating @xmath12 . the reflection arrangement @xmath75 associated to the group @xmath0 ( @xmath27 ) is defined in @xmath76 by the coordinate hyperplanes @xmath77 ( @xmath78 ) and the hyperplanes @xmath79 we will deal with the case @xmath34 in section [ sec : typea ] . the chevalley shephard todd theorem characterizes complex reflection groups as those subgroups @xmath80 for which the ring of polynomial invariants @xmath81 is itself a polynomial ring . algebraically independent generators of the invariant ring will not be unique , but their degrees @xmath82 will be and are referred to as the _ degrees of @xmath12_. for example , the degrees of @xmath0 are @xmath83 . the coxeter number of @xmath12 is the largest degree of @xmath12 and will be denoted by @xmath19 . for reflection groups @xmath80 generated by @xmath84 reflections ( which are called well - generated ) the integers @xmath85 satisfying @xmath86 are called the _ codegrees of @xmath12_. the codegrees of @xmath0 are thus @xmath87 . the tutte polynomial evaluation @xmath88 is called the characteristic polynomial of the arrangement . the polynomial @xmath89 is hilbert series of the cohomology ring of the complement of the arrangment @xmath90 . when @xmath8 is a reflection arrangement the relationship to the codegrees of @xmath12 is this . [ thm : os ] let @xmath8 be the reflection arrangement of a well - generated complex reflection group @xmath12 with codegrees @xmath91 . then , @xmath92 is the hilbert series of @xmath14 . a coxeter element of @xmath12 is an element @xmath93 where 1 . @xmath94 has an eigenvector @xmath95 that is not contained in any reflecting hyperplane of @xmath12 , 2 . @xmath12 acts freely on the orbit of @xmath95 and , 3 . the multiplicative order of the eigenvalue corresponding to @xmath95 is the coxeter number @xmath19 . it is shown in ( * ? ? ? * theorem 1.3 ) that when @xmath12 is irreducible and well - generated by @xmath1 reflections @xmath50 , and @xmath96 is a coxeter system , being a coxeter element is equivalent to being the product of the elements of @xmath50 in any order . in the group @xmath0 , @xmath1 generating reflections are given by the reflections that exchange the basis vectors @xmath97 and @xmath98 ( @xmath99 ) together with the generalized reflection that scales @xmath100 by @xmath101 . it follows that a coxeter element of @xmath0 is given by the matrix @xmath102 this generates a cyclic subgroup of @xmath0 or order @xmath103 . when our arguments in section [ sec : proofs ] use a particular coxeter element , we will use the one displayed here . the modifications required for a different coxeter element will be evident . any arrangement of hyperplanes has a dual arrangement , determined as follows . if @xmath74 is an arrangement then we can construct @xmath104 , which the vector space whose basis is in bijection with the hyperplanes defining @xmath105 . any choice of linear functionals defining @xmath8 determines a linear map @xmath106 . the kernel of this map is denoted @xmath107 , so we have an exact sequence @xmath108 dualizing gives an exact sequence @xmath109 since there is a natural basis of @xmath110 indexed by the hyperplanes in @xmath8 , the map @xmath111 determines an arrangement in @xmath107 . this is the _ gale dual of @xmath8 _ , which we denote by @xmath112 . assuming that the map to @xmath113 above is surjective , gale duality is an honest duality in the sense that @xmath114 . gale duality can be interpreted at the level of matroids . let @xmath37 be a matroid whose ground set we denote by @xmath49 and whose bases are denoted @xmath115 . the _ dual matroid @xmath116 _ of @xmath37 has ground set @xmath49 and @xmath117 if @xmath37 is the matroid associated to an arrangement @xmath8 then @xmath116 is the matroid associated to @xmath112 . [ prop : dual ] let @xmath37 be a matroid and @xmath116 its dual matroid . then the tutte polynomials of @xmath37 and @xmath116 are related by @xmath118 we employ this result to compute hilbert series as such . let @xmath10 be the reflection arrangement of well - generated complex reflection group @xmath12 . say that the coexponents of @xmath12 are @xmath119 then the hilbert series of @xmath120 is @xmath121 in particular , the dimension of the degree @xmath30 piece of @xmath120 is the product of the non - zero codegrees of @xmath12 . it follows from theorem [ thm : ap ] that the hilbert series in question is the tutte polynomial evaluation @xmath122 . by proposition [ prop : dual ] this is @xmath123 and by the theorem [ thm : os ] we obtain @xmath124 [ cor : hilbg ] let @xmath25 be the reflection arrangement of the group @xmath0 and let @xmath125 be its dual . then the hilbert series of @xmath31 is @xmath126 assume that @xmath13 is a representation of a group @xmath12 and that @xmath12 fixes @xmath8 in the sense that for any @xmath127 , @xmath128 . the arrangement @xmath112 can be made to carry an action of @xmath12 in a natural way , as we now explain . using the natural action of @xmath12 on @xmath113 , the elements of @xmath12 act by permuting and rescaling of the linear functionals defining @xmath8 . this defines an action of @xmath12 on @xmath129 , whereby @xmath12 acts by generalized permutation matrices . this action is cooked up so that the map @xmath106 is equivariant . since @xmath130 is the kernel of a map of @xmath12-modules it is @xmath12 invariant . since the map @xmath111 is a map of @xmath12-modules we see that the arrangement @xmath112 is fixed in @xmath107 by the natural action of @xmath12 . let @xmath131 denote the reflection arrangement associated to the group @xmath0 . we continue to assume that @xmath132 until section [ sec : typea ] . in this section we give an explicit coordinatization of the gale dual @xmath125 and describe how @xmath133 acts on the dual space @xmath107 in which @xmath125 lives . we begin by labeling the hyperplanes that define @xmath25 . to ease the notation ( only slightly ) we will write @xmath134 for @xmath101 let @xmath135 denote the hyperplane defined by the vanishing of the @xmath136-th coordinate function @xmath137 in @xmath3 . let @xmath138 denote the hyperplane defined by @xmath139 . we will use the convention that the superscript in this notation is read modulo @xmath2 and that @xmath140 the @xmath19 s form the basis for the vector space @xmath141 and the linear map @xmath142 sends @xmath143 and @xmath144 . the group @xmath12 is generated by the @xmath1-by-@xmath1 permutation matrices , which we identify with the symmetric group @xmath145 , and the diagonal matrices @xmath146 that have all diagonal entries @xmath147 except the @xmath136-th which is @xmath134 . the action of @xmath12 on @xmath141 is described in the following proposition . given @xmath148 , @xmath149 for all @xmath150 , @xmath151 if @xmath152 then @xmath153 . the kernel @xmath107 of the map @xmath142 is @xmath154 dimensional and has basis given by @xmath155 the arrangement @xmath156 comes from restricting the linear functionals dual to the @xmath19 s to @xmath107 . the action of @xmath12 on @xmath107 is immediate from the previous proposition . [ prop : rules ] given @xmath148 , @xmath157 for all @xmath150 , @xmath158 if @xmath152 then @xmath159 . an elementary computation shows the following result . [ prop : rules2 ] let @xmath93 denote the coxeter element @xmath160 , which has order @xmath103 . then @xmath161 our goal in this section is to prove the following result . [ thm : ind ] let @xmath25 be the reflection arrangement associated to the group @xmath133 . let @xmath93 be a coxeter element . there is an isomorphism of representations , @xmath162 we require a lemma . [ lem : main ] let @xmath107 be the kernel of the surjective linear map @xmath142 . 1 . @xmath107 has a @xmath94 eigenvector @xmath163 with eigenvalue @xmath164 . the eigenvector @xmath163 cyclically generates @xmath107 as a @xmath12-module . let @xmath165 , so that @xmath166 ( recall that @xmath167 ) . define @xmath168 ( notice that if @xmath34 then this sum reduces to @xmath169 using the rules set forth in proposition [ prop : rules ] , in particular proposition [ prop : rules2 ] , we see that @xmath170 acts on @xmath163 by @xmath171 assume that @xmath172 and then consider @xmath173 : @xmath174 let @xmath175 denote the @xmath176-module that @xmath177 generates . consider the @xmath176-invariant subspace of @xmath107 with basis @xmath178 , @xmath179 . this module is isomorphic to @xmath180 and hence there is an injection @xmath181 . the irreducible decomposition of @xmath180 is @xmath182 , where @xmath183 is the subspace of @xmath3 whose coordiates sum to zero ( * ? ? ? * exercise 4.6 ) . it will suffice to see that @xmath177 does not lay entirely within one of these summands , since this implies that injection above is an isomorphism . it will follow that @xmath184 and since @xmath185 generates @xmath107 as a @xmath12-module , we will be done . in the @xmath176-module @xmath3 define @xmath186 , so that @xmath187 forms a basis for @xmath183 and @xmath188 . using the natural isomorphisms @xmath189 we see that @xmath178 maps to @xmath190 . hence @xmath177 maps to @xmath191 the first component of is @xmath192 the coefficient of @xmath193 ( in the usual basis of @xmath194 , @xmath195 ) is @xmath196 if @xmath197 . if @xmath198 then the coefficent of @xmath193 is @xmath199 . in either case , the first component of @xmath200 is not zero . the second component of is @xmath201 and this is not zero . we now consider the case @xmath202 and @xmath2 even , where @xmath203 . consider the elements , @xmath204 for various values of @xmath136 , and the @xmath205 modules they generate . using the rule @xmath206 we can see that modules generated by @xmath207 , @xmath208 have zero intersection since the largest superscript of a term @xmath209 , occuring in @xmath210 is @xmath211 . the @xmath205 module generated by each @xmath212 is isomorphic to @xmath213 , for a @xmath214 dimensional subspace of @xmath107 , which is all of @xmath107 . it follows that the @xmath12 orbit of @xmath163 is all of @xmath107 . when @xmath202 and @xmath2 is odd , we replace @xmath163 with a different eigenvector for the coxeter element : @xmath215 , which we can write as @xmath216 . now apply @xmath217 to this element and get @xmath218 apply @xmath219 to this and divide by a scalar to get @xmath220 the @xmath205 representation generated by @xmath221 and @xmath222 is @xmath223 dimensional , and is thus equal to the representation generated by @xmath224 , since the latter is @xmath223 dimensional and contains the former . now the @xmath12 orbit of this variable spans @xmath107 and we are done . we use the previous result to prove that each graded piece of @xmath31 is a cyclic @xmath12-module . [ cor : spanningpowers ] for each @xmath225 , the @xmath12 orbit of @xmath226 spans the degree @xmath227 component of @xmath31 . it is well known that @xmath228 is spanned by @xmath227th powers of linear forms . since the @xmath12 orbit of @xmath163 is @xmath107 , the @xmath12 orbit of @xmath229 is all of @xmath228 . the corollary follows by taking the quotient . we know that @xmath230 contains a @xmath94 eigenvector with eigenvalue @xmath231 , namely @xmath232 . it follows that there is a surjective map from @xmath233 to the smallest @xmath12-module in @xmath234 containing @xmath232 . by proposition [ cor : spanningpowers ] there is a surjection from @xmath233 to @xmath234 . the dimension of the induced character is @xmath235 and by corollary [ cor : hilbg ] this is the dimension of @xmath234 . it follows that the surjection above is an isomorphism . the group @xmath133 contains @xmath236 as a subgroup , by taking the direct sum of a matrix in @xmath236 with a @xmath147-by-@xmath147 identity matrix . [ cor : reg ] as a representation of @xmath236 , @xmath230 is isomorphic to the regular representation . that is , @xmath237.\ ] ] the fact that the character of @xmath230 is an induced character allows us to see that it is zero at any element of @xmath238 , except the identity where it is @xmath239 . the result follows . in this section we investigate the case when @xmath34 in @xmath0 , so that our reflection group is the symmetric group of @xmath1-by-@xmath1 permutation matrices . to emphasize the dependence on @xmath1 here we will write @xmath176 instead of @xmath12 . the reflection arrangement of @xmath176 is the well - studied braid arrangement @xmath240 whose defining hyperplanes are @xmath241 the macaulay inverse system of the central zonotopal ideal @xmath242 has dimension @xmath243 , and was studied by the author and rhoades in @xcite . there it was shown to be a representation of the symmetric group @xmath244 that restricted to the well - studied _ parking representation _ of @xmath245 . this is the representation with basis give by sequences @xmath246 whose non - decreasing rearrangement @xmath247 satisfies @xmath248 for all @xmath249 . to identify the gale dual of @xmath8 we label the hyperplane @xmath250 by @xmath251 . those @xmath251 with @xmath195 form a basis for @xmath129 , and we stipulate that @xmath252 . the kernel @xmath107 of the natural map @xmath253 is @xmath254 dimensional , and a basis is given by @xmath255 one can check that @xmath256 . this kernel can be naturally identified with the cycle space of the complete graph on @xmath1 vertices , which is spanned by the characteristic vectors of length @xmath257 oriented cycles . a basis is then given by those @xmath257 cycles that visit the vertex labeled @xmath1 . the transposition @xmath258 acts on the variables @xmath259 by the rule , @xmath260 notice that the action of @xmath261 on @xmath107 agrees with action set forth in section [ sec : galedualg ] with @xmath34 and @xmath1 replaced by @xmath30 . we have the following analog of theorem [ thm : ind ] . [ thm : aind ] let @xmath8 denote the reflection arrangement of @xmath176 . let @xmath262 denote a coxeter element , which is an @xmath263-cycle . as a representation of @xmath264 , @xmath265 is isomorphic to @xmath266 . as a representation of @xmath267 , @xmath265 is isomorphic to the regular representation . the induced representation here @xmath266 is called the lie representation , which we write as @xmath268 . it is the representation of @xmath264 afforded by the multilinear component of the free lie algebra on @xmath30 letters . it was shown by stanley [ cite ] that the tensor product of the lie representation with the sign character is isomorphic to the top cohomology of the partition lattice , and this is isomorphic to the top cohomology of the complement of the braid arrangement @xmath269 . the proof is the same as for theorem [ thm : ind ] , except that we need to know that @xmath270 is @xmath271 . this follows as it did before because @xmath8 is a free arrangement with exponents @xmath272 which means that the hilbert series of @xmath265 is @xmath273 . the structure of @xmath265 as a representation of @xmath176 is more subtle , but turns out to have been computed before in work of mathieu @xcite and gaiffi @xcite . to state this result we need to define the whitehouse representation of the symmetric group @xmath176 . this is the ( _ a priori _ virtual ) representation , @xmath274 which was first studied by kontsevich in the cotext of free lie algebras and later by robinson and whitehouse @xcite . [ thm : whitehouse ] there is an isomorphism of representations of @xmath176 , @xmath275 . to prove this we compute a generating set for the defining ideal of @xmath120 this is an ideal in the polynomial ring @xmath276 $ ] . [ thm : type a ideal ] the ideal @xmath277 is equal to @xmath278 we start by proving one containment . the ideal @xmath279 contains @xmath280 . consider the linear functional @xmath281 dual to @xmath282 , restricted to @xmath107 . to show that @xmath283 we must compute the number of @xmath281 , @xmath284 , that do not vanish on @xmath259 . this number is @xmath257 and hence @xmath285 . the ideal @xmath286 is stable under the action of @xmath176 , and hence @xmath287 . subtracting off the quadratic terms proves that @xmath286 contains @xmath288 . the set of polynomials , @xmath289 forms a grbner basis under any term order where the leading terms are underlined above . we use the fact that syzygies of polynomials with relatively prime leading terms need not be computed ( * ? ? ? * exercise 15.20 ) . by symmetry , the computation reduces to the case when @xmath198 . this case is easily checked with a computer ( e.g. , using @xcite ) . the ideal of leading terms of @xmath280 is @xmath290 . it follows that a basis for @xmath291 consists of square - free monomials in the @xmath259 that avoid @xmath292 , for @xmath293 . these monomials are in obvious bijection with the set of decreasing forests on @xmath30 vertices . these are forests with vertex set @xmath294 $ ] where , in every component , each path directed away from the largest vertex in that component decreases . adding a vertex labeled @xmath1 and connecting the largest vertex in each component to @xmath1 , we obtain a bijection between decreasing forests on @xmath30 vertices , and decreasing trees on @xmath1 vertices . by ( * ? ? ? * proposition 1.5.5 ) , there are exactly @xmath295 such trees . it follows that the dimension of @xmath296 as a vector space is @xmath295 . since this is the dimension of @xmath120 ( recall , its hilbert series is @xmath273 and there is a containment between @xmath280 and @xmath277 , the two ideals are equal . to prove theorem [ thm : whitehouse ] we will need a result of sundaram which we recall here . [ prop : sundaram ] let @xmath297 and @xmath298 be ( possibly virtual ) representations of @xmath145 , and let @xmath183 denote the orthogonal complement of the trivial representation in @xmath3 . then the following are equivalent : 1 . 2 . @xmath300 . by theorem [ thm : type a ideal ] , we have identified @xmath120 with the @xmath176 module mathieu studies in ( * ? ? ? * theorem 4.4 ) . we denote the top degree part of this @xmath176 module by @xmath301 . in _ loc . _ it is shown that there is an isomorphism of @xmath176-modules , @xmath302 using theorem [ thm : aind ] and proposition [ prop : sundaram ] , @xmath303 this proves that @xmath304 . in this section we describe in coordinates the defining ideal of @xmath31 . we will use this to prove representation stability in the sense of church , ellenberg and farb and developed for wreath products with symmetric groups by sam and snowden @xcite and gan and li @xcite . recall that @xmath156 , where @xmath107 has coordinates @xmath305 with @xmath195 and @xmath306 . [ thm : ideal ] let @xmath25 be the reflection arrangement of @xmath133 ( @xmath307 . then the ideal defining @xmath31 in @xmath308 is the sum of @xmath309 with the smallest @xmath12-stable ideal @xmath310 containing the single element @xmath311 . we let @xmath286 denote the ideal of @xmath31 . we first show that @xmath312 . consider the element @xmath313 . the hyperplanes in @xmath29 that do not contain this vector are @xmath135 , @xmath314 and @xmath138 , hence @xmath315 . similarly , the number of hyperplanes in @xmath29 that do not contain @xmath316 is @xmath257 , hence @xmath317 . this proves @xmath318 . next we consider @xmath319 , which is contained in all but @xmath257 hyperplanes in @xmath29 , so its square is in @xmath286 . this shows that @xmath320 . we claim that @xmath321 has the same dimension as @xmath322 as a vector space . the non - zero monomials in @xmath323 are in obvious bijection with graphs on vertex set @xmath324 $ ] whose edges are labeled with the numbers @xmath325 . suppose that we have a monomial in @xmath321 whose corresponding graph has a @xmath257-cycle . we can use the @xmath297 invariance of this quoient to assume the three cycle has two edges labeled @xmath326 . rewriting the transformed monomial using the sole given generator of @xmath327 , we will obtain a sum of two monomials both of which reduce to zero modulo @xmath328 . suppose that we have a monomial whose corresponding graph has a cycle of length longer than @xmath257 . similar to the argument above we can use the elements in @xmath327 to rewrite this monomial as a sum of monomials of graphs with shorter cycles . see figure [ fig : cycleshorten ] . the resulting sum then reduces to zero in the quotient by induction on the length of a cycle . in terms of two monomials whose graphs have cycles of smaller length . ] we conclude that @xmath321 is spanned by monomials in bijection with forests on vertex set @xmath324 $ ] whose edges are labeled with the numbers @xmath325 . using the relations in @xmath327 , we see that out of the forests that span @xmath321 , we can use only those whose underlying graph where each connected component is a path , and one of whose endpoints is labeled with the highest label in that component . the number of such paths with exactly one connected component is at once seen to be @xmath329 , which is the dimension of the degree @xmath30 piece of @xmath330 . a calculation with the exponential formula shows that the number of such forests on @xmath1 vertices is @xmath331 , which is the dimension of @xmath31 . we conclude that @xmath332 , and hence @xmath333 , which is what we wanted to show . recall that the group @xmath334 can be interpreted as the wreath product @xmath335 . the category @xmath336 was introduced by sam and snowden in @xcite . its objects are finite sets and a map @xmath337 between two sets is a pair @xmath338 where @xmath339 is an injection and @xmath340 . the composition of @xmath341 with @xmath342 is defined by @xmath343 where @xmath344 . a ( complex ) representation of @xmath336 is a sequence @xmath345 of finite dimensional complex representations @xmath346 of @xmath297 , together with a sequence of compatible transition maps @xmath347 that are @xmath297-equivariant . denote the reflection arrangement associated to the group @xmath297 by @xmath348 , instead of the usual @xmath25 . for each degree @xmath349 , the sequences of representations @xmath350 and @xmath351 are representations of the category @xmath336 . let @xmath352 denote the ambient space of the arrangement @xmath353 . then the obvious inclusion @xmath354 is @xmath0-equivariant . the corresponding injections @xmath355 are @xmath0-equivariant . let @xmath356 denote the internal zonotopal ideal of @xmath353 . by theorem [ thm : ideal ] there is a @xmath0-equivariant inclusion @xmath357 . it follows that the natural maps @xmath358 are @xmath297 equivariant . since all the maps are homogeneous , the result follows . a representation @xmath37 of @xmath336 is said to be finitely generated if there are finitely many elements in the spaces @xmath359 such that the smallest @xmath336 submodule of @xmath37 containing these elements is exactly @xmath37 itself . for each degree @xmath349 , the sequences of representations @xmath350 and @xmath351 are finitely generated . theorem [ thm : ideal ] shows finite generation of the graded pieces of the sequence of ideals directly . this also follows since the sequence @xmath360 is finitely generated and the main theorem of @xcite proves that subrepresentations of finitely generated objects are again finitely generated ( this is certainly overkill ) . the graded pieces of the sequence @xmath361 are finitely generated since they are quotients of a finitely generated object . the importance of being finitely generated here is that it implies representation stability , as described in ( * ? ? ? * definition 1.10 ) . recall that the irreducible ( complex ) representations of the wreath product @xmath335 are indexed by @xmath2-tuples of partitions @xmath362 satisfying @xmath363 . we denote this irreducible by @xmath364 ; we will not need the precise construction here . the entries of this @xmath2-tuple @xmath365 are in bijection with the irreducible characters of @xmath366 and we assume that @xmath367 corresponds to the trivial character . for @xmath1 larger than @xmath368 we let @xmath369 $ ] denote the partition obtained by adding a single part of length @xmath370 to @xmath367 . we say that a sequence @xmath371 of @xmath297 representations is representation stable if for all @xmath372 the following conditions hold : 1 . the map @xmath347 is injective . 2 . the image of @xmath346 in @xmath373 generates @xmath373 as a @xmath374-module . the irreducible decomposition is given by @xmath375)^{\oplus m_{\underline\lambda}}\ ] ] for some integers @xmath376 that do not depend on @xmath1 . gan and li prove ( * ? ? ? * theorem 1.12 ) that @xmath371 is finitely generated if and only if it is representation stable . applying their result we have the following . for each degree @xmath349 , the sequences of representations @xmath350 and @xmath351 are representation stable . one might hope , based on what happens in type a , that @xmath377 could be made to carry an action of @xmath378 . however , there does not appear to be an action generalizing . the goal of this section is to prove an analog of the result of mathieu ( * ? ? ? * theorem 4.4 ) used in the proof of theorem [ thm : whitehouse ] . [ thm : factorization ] let @xmath379 be the graded reprentation of @xmath0 that is trivial in degree @xmath326 and equal to @xmath380 in degree @xmath147 . let @xmath348 be the reflection arrangement of @xmath0 . then there is an isomophism of graded @xmath0-modules , @xmath381 taking the top degree piece of both sides and applying corollary [ cor : reg ] we obtain the following result . maintaining the notation of theorem [ thm : factorization ] , there is an isomorphism of @xmath0-modules , @xmath382 \approx s_{{\mathcal{g}}_n^\perp,-2}(\textup{top } ) \otimes e_1.\ ] ] we can unravel this using the proof of sundaram s propositin [ prop : sundaram ] , which brought us to the whitehouse representation in theorem [ thm : whitehouse ] . we obtain a much more underwhelming result , since we arrive at the simple statement @xmath383 \approx { \mathbf{c}}[g(m,1,n+1)].\ ] ] from the proof of theorem [ thm : ideal ] , a linear basis for @xmath384 consists of monomials indexed by certain forests on vertex set @xmath385 $ ] . specifically , the edges are labeled with @xmath386 , each components is a path , and the largest labeled vertex in each path has degree one . this means that there is a basis indexed by monomials that are either ( a ) not divisble by any variable of the form @xmath387 or ( b ) divible by exactly one variable of the form @xmath387 . this implies that there is an isomorphism of graded vector spaces , @xmath388 where @xmath389 is trivial in degree @xmath326 and in degree @xmath147 has basis @xmath387 , @xmath78 , @xmath306 . the multiplication map @xmath390 is visibly @xmath0-equivariant . the group @xmath0 acts by permuting the variables @xmath387 . the action is transitive on the set of variables . the stabilizer of @xmath391 is precisely @xmath236 and this proves that @xmath392 as representations of @xmath0 . in this section we compute a grobner basis of the ideal @xmath393 when @xmath35 , where @xmath12 is the hyperoctohedral group . we use this to give a non - trivial generalization of the notion of decreasing trees , as discussed in section [ sec : typea ] . we start with a definition . a @xmath394tree on @xmath1 vertices is a rooted tree with vertex set @xmath324 $ ] and root vertex @xmath1 , together with a @xmath395-coloring of its edges . a decreasing @xmath394tree on @xmath1 vertices is a @xmath394tree on @xmath1 vertices such that on any path directed away from the root , 1 . along every edge labeled @xmath396 the vertex labels decrease , 2 . for @xmath397 and arbitrary @xmath16 , there is no path of the form @xmath398^{- } & i\ar[r]^{+ } & k},\ ] ] 3 . for all @xmath399 , there is no subpath of any of the three forms , or their reverse , @xmath400^+ & i_1\ar[r]^{+ } & i_2 \ar[r]^+ & i_3},\\ \xymatrix { i_4 \ar[r]^+ & i_1\ar[r]^{+ } & i_3 \ar[r]^+ & i_2},\\ \xymatrix { i_4 \ar[r]^+ & i_2\ar[r]^{+ } & i_1 \ar[r]^+ & i_3}. \end{aligned}\ ] ] if we ignore the possibility of edges labeled @xmath401 this definition reduces to the usual definition of decreasing trees . here are the @xmath402 decreasing @xmath394trees on @xmath257 vertices . @xmath403^{- } & & 3\ar[d]^+ & & 3\ar[d]^{+ } & & 3\ar[d]^+ \\ 2 \ar[d]^{- } & & 2\ar[d]^{- } & & 2\ar[d]^{+ } & & 1\ar[d]^+ \\ 1 & \quad & 1 & \quad & 1 & \quad & 2 } } \qquad \vcenter { \xymatrix@c1px { & 3 \ar[dr]^{-}\ar[dl]_{- } & \\ 1&&2 } } \quad \vcenter { \xymatrix@c1px { & 3 \ar[dr]^{+}\ar[dl]_{- } & \\ 1&&2 } } \quad\vcenter { \xymatrix@c1px { & 3 \ar[dr]^{-}\ar[dl]_{+ } & \\ 1&&2 } } \quad\vcenter { \xymatrix@c1px { & 3 \ar[dr]^{+}\ar[dl]_{+ } & \\ 1&&2 } } \ ] ] there are @xmath404 decreasing @xmath394trees on @xmath405 vertices . exactly @xmath402 such trees are isomorphic to a path rooted at an endpoint , and these are displayed below . @xmath406^{- } & & 4\ar[d]^{+ } & & 4\ar[d]^{+ } & & 4\ar[d]^+ & & 4 \ar[d]^{+ } & & 4\ar[d]^+ & & 4\ar[d]^{+ } & & 4\ar[d]^+ \\ 3 \ar[d]^{- } & & 3\ar[d]^{- } & & 3\ar[d]^{+ } & & 3\ar[d]^+ & & 3 \ar[d]^{+ } & & 2\ar[d]^+ & & 2\ar[d]^{+ } & & 1\ar[d]^{+ } \\ 2 \ar[d]^{- } & & 2\ar[d]^{- } & & 2\ar[d]^{- } & & 2\ar[d]^+ & & 1 \ar[d]^{+ } & & 3\ar[d]^{+ } & & 3\ar[d]^{- } & & 3\ar[d]^{- } \\ 1 & \quad & 1 & \quad & 1 & \quad & 1 & \quad & 2 & \quad & 1 & \quad & 1 & \quad & 2 } } \ ] ] there are @xmath407 decreasing @xmath394trees on @xmath1 vertices . we know from the proof of theorem [ thm : ideal ] that the graphs corresponding to monomials not in @xmath393 are forests on @xmath324 $ ] whose edges are labeled @xmath326 and @xmath147 . we change every edge labeled @xmath326 to an edge labeled @xmath396 likewise with @xmath147 and @xmath401 . we now show that the monomials of decreasing @xmath394trees form a basis for the quotient @xmath230 . for this , we claim that a grbner basis of @xmath393 is furnished by the following monomials : @xmath408 where @xmath136 , @xmath409 and @xmath16 range of distinct triples of integers , as well as the polynomials @xmath410 where the indices range over distinct tuples of integers , and finally given @xmath411 the polynomials , @xmath412 we use any term order where the underlined monomials are leading terms ( such as graded - reverse lexicographic order with the @xmath413 s coming before the @xmath414 s ) . to see this one applies the buchberger algorithm to the generators of @xmath393 given in theorem [ thm : ideal ] . by ( * exercise 15.20 ) we can reduce the computation to the case @xmath198 , since we need not resolve the syzygies of pair of polynomials with relatively prime leading terms . this case is , once again , handled at once by a computer to produce the described grbner basis . a bijective proof of this result generalizing the one for ordinary decreasing trees is not known . a natural question is whether theorem [ thm : ind ] holds for all complex reflection groups in its present formulation . a number of problems can arise . a necessary condition is that the dimension of an induced character @xmath415 be equal to the dimension of @xmath230 , which is the product of the non - zero codegrees of @xmath12 . for not well - generated reflection groups we take as an analog of a coxeter element an element of @xmath12 with order @xmath416 . the equality of dimensions becomes , @xmath417 for the exceptional groups , we list the shephard todd number of each group where @xmath418 ( we have written the degrees and codegrees in weakly increasing order ) , which signfies that theorem [ thm : ind ] could hold for this group : 5 , 7 , 10 , 11 , 18 , 19 , 26 . this equality also holds for all of the groups @xmath419 , consisting of matrices in @xmath0 the product of whose non - zero entries is an @xmath420th root of unity . in any groups where equality holds to would suffice to prove the following statement : _ let @xmath421 be the reflection arrangement of a given complex reflection group @xmath12 and let @xmath107 be the kernel of the natural map @xmath422 . then @xmath107 has an eigenvector @xmath163 with eigenvalue @xmath423 whose @xmath12 orbit spans @xmath107 . _ in the case of @xmath424 with @xmath425 the reflection arrangement @xmath25 associated to @xmath12 is the same as the arrangement associated to @xmath0 . the proof of lemma [ lem : main ] shows that for @xmath426 , @xmath107 contains an @xmath94-eigenvector @xmath163 whose @xmath12 orbit include the single variable @xmath185 . the @xmath12 orbit of @xmath185 does not span @xmath107 , so more work is required . f. ardila and a. postnikov . correction to `` combinatorics and geometry of power ideals '' : two counterexamples for power ideals of hyperplane arrangements . _ 367 ( 2015 ) , no . 5 , 37593762 . sam and a. snowden . representations of categorgies of @xmath430-maps . _ j. reine angew . . s. sundaram . a homotopy equivalence for partition posets related to liftings of @xmath431-modules to @xmath145 . _ j. combin . theory ser . a _ 94 ( 2001 ) , no . 1 , 156168 .	the group @xmath0 consists of @xmath1-by-@xmath1 monomial matrices whose entries are @xmath2th roots of unity . it is generated by @xmath1 complex reflections acting on @xmath3 . the reflecting hyperplanes give rise to a ( hyperplane ) arrangement @xmath4 . the internal zonotopal algebra of an arrangement is a finite dimensional algebra first studied by holtz and ron . its dimension is the number of bases of the associated matroid with zero internal activity . in this paper we study the structure of the internal zonotopal algebra of the gale dual of the reflection arrangement of @xmath0 , as a representation of this group . our main result is a formula for the top degree component as an induced character from the cyclic group generated by a coxeter element . we also provide results on representation stability , a connection to the whitehouse representation in type @xmath5 , and an analog of decreasing trees in type @xmath6 .
the branch of atomic physics associated with ultracold atoms , ions , and molecules now provides a rich and exciting arena for investigations of strongly interacting , many - body quantum systems . trapping and cooling techniques , coherent laser or microwave interactions , and applied magnetic fields enable exquisite control of both external ( motional ) and internal ( electronic ) degrees of freedom of the particles , allowing one to `` tailor '' particle - particle interactions and thereby implement a broad range of systems that can be described accurately and transparently by idealized ( but nontrivial ) many - body hamiltonians . an important example is the hubbard model , realized with ultracold atoms in periodic optical lattices @xcite , while realizations of other novel and significant lattice - spin models have been proposed , for example , with dipolar molecules in optical lattices @xcite and with chains of trapped atomic ions @xcite . the common , defining feature of these systems is the possibility for quantum critical phenomena , i.e. , transitions between distinct quantum phases , in response to variations of an effective field or particle - particle interaction strength around some critical value . the above - mentioned schemes generally provide many - body quantum systems that are subject to short - range ( e.g. , nearest - neighbor ) interactions . another interesting and commonly studied class of many - body systems are those possessing long - range , or even infinite - range , interactions , for which theoretical models typically allow exact solutions in the thermodynamic limit , or at least enable efficient numerical solution for large numbers of particles . a standard and classic example is the lipkin - meshkov - glick ( lmg ) model @xcite , which was originally introduced in nuclear physics and is described by a hamiltonian of the form @xmath0 where @xmath1 are collective angular momentum operators for @xmath2 spin-1/2 particles , @xmath3 and @xmath4 are parameters giving the effective magnetic field and spin - spin interaction strengths , respectively , and @xmath5 $ ] is an anisotropy parameter . in this model , each spin interacts identically with every other spin and the nature of this interaction may be ferromagnetic ( @xmath6 ) or antiferromagnetic ( @xmath7 ) . significantly , the model exhibits critical behavior at zero temperature ; in particular , either first- or second - order quantum phase transitions may occur ( depending on the choice of @xmath4 and @xmath8 ) as the ratio between @xmath4 and @xmath3 is varied across a critical value . this quantum critical behavior , combined with the relative simplicity of the model , has led to renewed theoretical interest in the lmg model from the point of view of studying entanglement properties of many - particle systems in relation to quantum phase transitions @xcite . bipartite entanglement measures characterizing entanglement between a pair of spins ( e.g. , the concurrence ) or between two blocks of spins ( e.g. , the entanglement entropy ) are relatively straightforward to compute for the lmg model and can display marked critical behavior and scaling at quantum critical points @xcite . given these interesting and very topical features of the lmg model , it follows that the physical realization of a system described accurately by such a model would provide a valuable test bed for studies of quantum critical phenomena and entanglement . however , the question naturally arises as to how realistic such an idealized model could be ; the assumption of `` infinite - range '' interactions is obviously demanding and implies a very specialized system . hamiltonians of the form ( [ lmghamiltonian ] ) ( with @xmath9 ) have appeared recently in reduced two - mode models of atomic bose - einstein condensates undergoing tunnelling in double - well potentials or transitions between two internal atomic states @xcite , and in models of a few trapped ions interacting with laser fields @xcite , but emphasis in these works has been on unitary or adiabatic evolution from some initial atomic state to some final , prescribed ( entangled ) state , while flexibility of these systems with respect to parameters of the lmg model ( i.e. , @xmath4 , @xmath2 , @xmath8 ) appears limited . another possibility , furnished by the field of quantum optics , and for which long - range atom - atom interactions actually occur quite naturally , is cavity quantum electrodynamics ( cavity qed ) @xcite . here , one considers ensembles of atoms interacting , through an electronic transition , with a common electromagnetic field mode supported by an optical resonator . through this common coupling , the field mode can effectively mediate interactions between atoms located at quite arbitrary and separate positions within the mode . so , in particular , the concept of an interaction `` length '' becomes redundant in this setting and a collective description of the atoms is appropriate . in fact , that an ensemble of atoms coupled to a common field mode can be viewed as a many - body system of interacting spins was highlighted many years ago with the prediction of a thermal equilibrium phase transition in the celebrated dicke model of @xmath2 two - level atoms coupled to a single quantized field mode @xcite , @xmath10 where @xmath11 is the annihilation operator for the field mode of frequency @xmath12 , @xmath13 is the atomic transition frequency , and @xmath4 is the atom - field coupling strength ( we set @xmath14 ) . in particular , above a certain critical value of the coupling strength the system enters a so - called `` superradiant '' phase @xcite . this phase transition persists at zero temperature @xcite , with associated critical behavior of both the atom - field and atom - atom quantum entanglement @xcite . the critical coupling strength at zero temperature is given by @xmath15 , which means that @xmath4 must be comparable to the field and/or atomic transition frequencies if the transition regime is to be reached . for atomic dipole transitions , this is typically not the case and , in fact , if it happened to be so , then the model ( [ dickehamiltonian ] ) would be inadequate ; in particular , the @xmath16 term [ omitted from ( [ dickehamiltonian ] ) ] of the minimal coupling hamiltonian should be included and doing so one actually finds that no phase transition exists @xcite . however , a recent proposal for realizing the dicke model quantum phase transition , based on raman transitions between stable atomic ground states in an optical cavity qed setting @xcite , circumvents these issues by ( i ) implementing a system in which the relevant frequency and coupling scales are determined by light - induced frequency shifts and raman transition rates , and ( ii ) utilizing an open - system dynamics ( as opposed to a closed , hamiltonian system ) with input and output fields ( i.e. , external laser fields and cavity mode losses ) , thereby replacing a ( fragile ) thermal equilibrium phase transition with a ( robust ) dynamical , nonequilibrium phase transition . furthermore , as shown in ref . @xcite , the cavity output field offers a unique window on the system s behavior and properties , with , for example , fluorescence and quadrature - variance measurements providing dramatic signatures of criticality in the system , as well as quantitative measures of fluctuations and entanglement . these features of the optical cavity qed system , combined with the observation that , in the dispersive limit @xmath17 , the cavity mode may be adiabatically eliminated and the dicke hamiltonian reduced to the form @xmath18 where @xmath19 , motivate us to explore the possibilities for studying the lmg model in such a setting . in particular , by generalizing the configuration of ref . @xcite to two cavity field modes and operating in a dispersive regime ( amounting to far - off - resonant raman transitions ) , we find that it is possible to implement atomic spin systems that are described by the most general lmg model ( [ lmghamiltonian ] ) , and for which the hamiltonian dynamics may still dominate over losses to the output cavity fields , thus enabling the clear realization of critical phenomena , including both first- and second - order dynamical quantum phase transitions . we find also that the cavity output fields can again be used to provide clear and detailed probes of properties of the atomic collective - spin system , including entanglement , in the critical regime . we note that the present work bears some relation to studies of optical bistability and resonance fluorescence in cooperative atomic systems , which can also exhibit first- and second - order nonequilibrium phase transitions ( see , for example , @xcite ) . there , however , the dynamics explicitly includes ( resonant ) coherent driving of the atomic system by an external laser field ( i.e. , the hamiltonian describing the system contains a driving term linear in @xmath20 or @xmath21 , rather than a direct spin - spin interaction term ) , and relatively little investigation has been made of the quantum entanglement associated with the critical behavior @xcite . a more specific outline of our paper is as follows . in sec . [ sect : general_model ] we describe the microscopic model of atoms and light fields that realizes our effective spin system . in sec . [ sect : spin_models ] we present some background to the lmg collective spin model and show how to engineer it using the general setup presented in sec . [ sect : general_model ] . we conclude sec . [ sect : spin_models ] with a brief overview of the methods of analysis to be used later in the paper . in sec . [ sect : implementation ] we describe a more specific , potential physical implementation of the system we have proposed , based on alkali metal atoms confined within a high - finesse ring cavity . in sec . [ sect : ferro_model ] we focus on the @xmath9 lmg model and focus on the second - order transition ; we first present a linearized analysis of the system in the thermodynamic limit using the holstein - primakoff representation of spin operators . using the input - output theory of quantum optics we relate the internal spin properties to the measurable cavity output field and determine the probe transmission spectrum as an example . the second part of sec . [ sect : ferro_model ] is concerned with the presence and behavior of atom - atom ( or spin - spin ) entanglement in the system , particularly across the quantum phase transition ( qpt ) . we present results for both the steady - state entanglement and the entanglement dynamics , using either exact numerical solutions for finite system size or analytical solutions in the thermodynamic limit . in sec . [ sect : af_model ] we essentially repeat the analysis of the previous section , but focus on a parameter regime where a first - order phase transition occurs in the @xmath9 lmg model as the effective magnetic field parameter , @xmath3 , is varied . finally , in sec . [ sect : conclusion ] we conclude and briefly discuss possible extensions of the current work . we consider a collection of @xmath2 atoms coupled via electric dipole transitions to ( at most ) four laser fields and to a pair of orthogonally polarized optical cavity modes . the atomic level and excitation scheme is shown in fig . [ fig : atomic_level_scheme ] . in particular , the atoms are assumed to possess two stable electronic ground states , labelled @xmath22 and @xmath23 , at energies ( @xmath14 ) @xmath24 and @xmath25 , respectively . the laser and cavity fields combine to drive raman transitions between @xmath22 and @xmath23 , via the excited atomic states @xmath26 and @xmath27 ( energies @xmath28 and @xmath29 , respectively ) . specifically , the laser fields , at frequencies @xmath30 , @xmath31 , @xmath32 , and @xmath33 , couple to the dipole transitions @xmath34 , @xmath35 , @xmath36 , @xmath37 with rabi frequencies @xmath38 , @xmath39 , @xmath40 , and @xmath41 , respectively . cavity field @xmath11 , at frequency @xmath42 , couples to the transitions @xmath34 and @xmath37 with coupling strengths @xmath43 and @xmath44 , respectively , while cavity field @xmath45 , at frequency @xmath46 , couples to the transitions @xmath35 and @xmath36 with coupling strengths @xmath47 and @xmath48 , respectively . all of the fields will be assumed to be far - off resonance with the electric dipole transitions to which they couple , meaning that the atomic states @xmath26 and @xmath27 are only virtually excited and can be eliminated from the dynamics . finally , at the location of the atoms , the cavity and laser fields are taken to be travelling waves copropagating in the @xmath49 direction , with sufficiently broad beam waists so as to ensure a homogeneous atom - field coupling . to facilitate adiabatic elimination of the atomic excited states we move to a rotating frame according to the unitary transformation @xmath50 , with @xmath51 where @xmath52 is a frequency close ( or possibly equal ) to @xmath25 . next , as mentioned above , we assume large detunings of the light fields from the atomic excited states , i.e. , we assume that @xmath53 and @xmath54 are much larger in magnitude than any other rates characterizing the system . this allows the atomic excited states to be adiabatically eliminated and also enables us to neglect the effects of atomic spontaneous emission . additionally , as depicted in fig . [ fig : atomic_level_scheme ] , we assume that only four distinct raman transitions are of significance ( i.e. , resonant or roughly resonant ) ; in particular , in our model we retain only those raman processes that cause a change in the electronic state of the atoms ( @xmath55 or @xmath56 ) and also involve transfer of a photon from a laser field into a cavity mode or vice versa . all other possible raman processes are assumed to be far - off resonant and therefore negligible . quantitatively , this requires , for example , that @xmath57 and @xmath58 are sufficiently large , with , in particular , @xmath59 . retaining only the four dominant raman processes simplifies the model considerably , and with a choice of laser frequencies such that @xmath60 we are able to remove all explicit time dependence from the hamiltonian describing our system . employing the collective spin operators , @xmath61 and omitting constant energy terms , our effective hamiltonian for the collective atomic system and cavity modes can be written in the form @xmath62 where @xmath63 and the effective parameters are given in terms of the microscopic parameters by @xmath64 note that the ( dimensionless ) factors @xmath65 $ ] have been introduced for convenience . note also that for a characteristic level scheme as shown in fig . [ fig : atomic_level_scheme ] , one might typically expect that @xmath66 and @xmath67 , so assuming @xmath68 we would therefore also expect that @xmath69 . in summary , the master equation for the reduced density operator , @xmath70 ( i.e. , with the atomic excited states eliminated and spontaneous emission neglected ) , is given by @xmath71 + \kappa_a d[a]\rho_{\rm g } + \kappa_b d[b]\rho_{\rm g } , \label{eq : master_eqn_atom_cavity}\end{aligned}\ ] ] where @xmath72\rho = 2 a \rho a^\dagger - a^\dagger a \rho - \rho a^\dagger a$ ] and @xmath73 is the cavity field decay rate . we now consider the limit @xmath74 . in this limit , the cavity modes are only ever weakly or virtually excited and may also be adiabatically eliminated from the dynamics . following the standard adiabatic elimination procedure @xcite , we derive the following master equation for the reduced density operator , @xmath75 , of the collective atomic system alone : @xmath76 + \gamma_a d[x_a^\dagger]\rho + \gamma_b d[x_b^\dagger]\rho , \label{eq : spin_master_eqn_general}\end{aligned}\ ] ] with @xmath77 where the effective spin - spin interaction strengths and collective atomic dissipative rates are ( @xmath78 ) @xmath79 note that both dispersive nonlinear terms [ terms proportional to @xmath80 and @xmath81 in eq . ( [ eq : heff ] ) ] do not contribute in the adiabatic approximation since in this limit we assume a vacuum state for both cavity modes . taking a brief step backwards now to the atom - cavity hamiltonian ( [ eq : heff ] ) , and using the input - output theory of open quantum optical systems @xcite , we can derive quantum langevin equations for the cavity mode operators ; in particular , for the mode @xmath45 , we have ( neglecting the term proportional to @xmath81 ) @xmath82 where @xmath83 describes the quantum noise input to the cavity mode ( see fig . [ fig : implementation ] ) and satisfies the commutation relation @xmath84=\delta(t - t')$ ] . equation ( [ eq : qle ] ) illustrates the linear relationship between the cavity operator and atomic operator @xmath85 . the adiabatic limit of the preceding subsection amounts , in the present context , to the assumption that @xmath86 varies on a much slower time scale than @xmath87 [ and @xmath88 , so that we can write @xmath89 the cavity output field is given by @xmath90 , so we in turn obtain a direct relationship between the dynamics of the ( internal ) collective atomic spin and the ( external ) cavity output field . hence , spin - spin correlations of the form @xmath91 and @xmath92 could be deduced from correlations of the cavity output field , which may be measured , for example , by performing broadband homodyne detection on the emitted light @xcite . the lmg model , originally introduced in nuclear physics to model collective motion in nuclei @xcite , describes @xmath2 interacting fermions distributed on two @xmath2-fold degenerate levels ( denoted by @xmath93 ) separated by an energy @xmath94 . denoting the fermion annihilation operator by @xmath95 , where @xmath96 and @xmath97 , the hamiltonian for this system may be written as @xmath98 introducing the collective spin operators , @xmath99 and @xmath100 allows us to reexpress the hamiltonian as @xmath101 this hamiltonian commutes with @xmath102 , thus conserving the total angular momentum , and with @xmath103 , corresponding to a parity ( spin - flip ) symmetry @xcite . it is straightforward to rewrite this hamiltonian in terms of @xmath20 and @xmath21 , defined via @xmath104 , giving the generalized lmg model , @xmath105 where @xmath106 , @xmath107 ( we will only consider @xmath108 $ ] ) , and @xmath109 . this model is well known for its second - order symmetry breaking phase transition in the ferromagnetic regime ( @xmath6 ) @xcite . for small interaction strength the system is in the normal phase , where the ground state is unique and polarized in the direction of the magnetic field . as the interaction is increased above a critical value , @xmath110 , the system enters the broken phase , where the ground state becomes doubly degenerate and macroscopically displaced from its original configuration , thus breaking the parity symmetry . for the special case @xmath111 the hamiltonian also commutes with @xmath112 , thus enabling a direct analytic solution . all other cases @xmath113 lie in a separate universality class . in the antiferromagnetic regime , @xmath7 , the model exhibits a first - order phase transition as the effective magnetic field @xmath3 crosses @xmath114 ( provided @xmath115 ) . using the setup described in the previous section we can implement the generalized lmg model for any @xmath8 by making appropriate choices of @xmath116 in the hamiltonian ( [ eq : spin_hamiltonian_general ] ) . we now consider three specific cases of general interest . the @xmath118 lmg hamiltonian may be implemented by choosing @xmath119 and @xmath120 ( corresponding to @xmath121 and @xmath122 ) , and setting @xmath123 ( note that the signs of @xmath124 are determined by the signs of the detunings @xmath125 ) , so that @xmath126 with @xmath127 and @xmath128 . this instance of the lmg model has been most widely studied for its phase transition properties . for the dissipative terms we assume , for simplicity , that @xmath129 , so that the full master equation reduces to the form @xmath130 + \frac{\gamma_+}{n } d[j_+]\rho + \frac{\gamma_-}{n } d[j_-]\rho , \label{eq : convlmg(b)}\end{aligned}\ ] ] where @xmath131 and @xmath132 . the hamiltonian dynamics can be expected to play a dominant role if @xmath133 ( which corresponds to @xmath134 ) . the isotropic @xmath135 lmg hamiltonian may be obtained , for example , by choosing @xmath136 and @xmath137 ( corresponding to @xmath138 and @xmath139 ) , and setting @xmath140 , which gives @xmath141 where @xmath127 . the full master equation is @xmath130 + \frac{\gamma_a}{n } d[j_-]\rho + \frac{\gamma_b}{n } d[j_+]\rho \label{eq : isolmg(b)}.\end{aligned}\ ] ] the @xmath142 lmg hamiltonian , which will be focus of our attention in this paper , may be obtained by choosing @xmath143 ( corresponding to @xmath144 ) , and setting @xmath145 ( so that @xmath146 ) . this gives @xmath147 where @xmath148 . while the raman channels involving the cavity mode @xmath45 could be omitted completely , here we retain one of them ( for reasons to be discussed below ) , with the choice @xmath149 , and @xmath150 , corresponding to @xmath151 . hence , the full master equation we consider is @xmath130 + \frac{\gamma_a}{n } d[2j_x]\rho + \frac{\gamma_b}{n } d[j_+]\rho , \label{eq : master_equation_gamma0_lmg_model}\end{aligned}\ ] ] where the factor @xmath152 has been absorbed into @xmath153 . if we now consider the case where @xmath154 and @xmath155 , then the role played by each cavity mode in relation to the atomic system is quite distinct . specifically , cavity mode @xmath11 mediates the collective spin - spin interaction required for the hamiltonian dynamics ( with coupling strength @xmath156 ) , while cavity mode @xmath45 effectively mediates the collective atomic decay ( with rate @xmath157 ) . importantly , we note that @xmath158 implies a quite direct relationship between moments of the cavity mode operators @xmath45 and @xmath159 and moments of the collective atomic spin operators @xmath160 ; in particular , measurements of the output light field from cavity mode @xmath45 will provide , rather directly and transparently , characteristic properties of the collective atomic spin . in contrast , for the @xmath118 model the two cavity modes mediate the collective spin - spin interaction on an equal footing , i.e. , @xmath161 , while the operators @xmath162 and @xmath163 are linear combinations of @xmath164 and @xmath165 , which leads to a somewhat less transparent ( i.e , arguably less convenient ) relationship between correlations of the cavity output fields and atomic spin - spin correlations . partially for this reason , we focus in this paper on the @xmath142 model , with a clear distinction between the effective roles of the two cavity modes and a potentially better suitability for measurements of the collective atomic spin properties . to analyze the atomic - spin master equations presented in the preceding sections , we make use of both numerical and analytical techniques . for finite spin @xmath166 , the master equations can be solved numerically for quite large @xmath2 @xcite , owing to the linear scaling of the hilbert space dimension , @xmath167 , with the system size , i.e. , @xmath168 . in what follows , we will typically present results of numerical simulations for @xmath169 . for very large system sizes , @xmath170 , it is possible to linearize the quantum fluctuations around the mean spin state ( i.e. , around the `` bloch vector '' ) . first we find this mean spin state by calculating the steady - state solutions of the semiclassical equations of motion for the components of the bloch vector . after a suitable rotation ( determined by the mean state ) of the spin coordinate system , we use the holstein - primakoff ( hp ) representation of angular momentum operators @xcite , which enables a systematic large-@xmath2 expansion of the master equation , to which we then apply the limit @xmath171 . while all of the results obtained in the linearized regime are exact analytical results , in many cases the expressions obtained are too lengthy to give any useful information ; in these cases we simply plot the relevant quantities . for a possible experimental implementation of our scheme , we consider , as mentioned previously , an ensemble of atoms confined inside a high finesse ring cavity that supports two travelling - wave modes , @xmath11 and @xmath45 . the required laser fields , which are assumed to be at frequencies that are not supported by the resonator , are injected through one of the resonator mirrors so as to be copropagating with the cavity fields through the ensemble . if we take @xmath172 as the atomic species , then the atomic level scheme of fig . [ fig : atomic_level_scheme ] can be implemented directly with the two ground magnetic substates @xmath173 as @xmath174 and @xmath175 , and with a magnetic field applied perpendicular to the cavity axis to provide a frequency splitting @xmath176 between these two states . the modes @xmath11 and @xmath45 would be orthogonal , linearly polarized cavity modes , with , in particular , mode @xmath11 polarized along the direction of the magnetic field . ( note that if the two modes happen to be very different in frequency due , for example , to birefringence in the cavity mirrors , then the magnetic field may not be necessary . ) another possibility , illustrated in fig . [ fig : implementation ] , might be a configuration based on the @xmath177 transition of @xmath178 , in which the states @xmath174 and @xmath175 are the ground magnetic substates @xmath179 , with frequency splitting @xmath176 due to a magnetic field applied along the cavity axis . the modes @xmath11 and @xmath45 would be orthogonal , linearly polarized cavity modes , polarized perpendicular to the magnetic field . note , however , that the modes would need to be sufficiently different in frequency ( which could be imposed , for example , by cavity birefringence ) in order that the raman channels involving different modes are distinct . and frequency @xmath180 . ( b ) possible atomic level scheme as described in the text.,width=325 ] alternatively , the modes @xmath11 and @xmath45 could be two entirely different ( linearly polarized ) longitudinal modes of the resonator , one quasiresonant with the @xmath177 transition of the d2 line and the other quasiresonant with the @xmath181 transition of the d1 line . for specific parameter values , we consider experimental systems such as those realized recently in ref . @xcite , where cold atoms are held inside a high - finesse optical ring cavity . in particular , let us assume a single - atom single - photon dipole coupling strength of @xmath182 and a cavity field decay rate of @xmath183 . for @xmath184 atoms and a characteristic laser - rabi - frequency - to - detuning ratio of @xmath185 , we have @xmath186 ( @xmath187 ) . if we assume a raman detuning @xmath188 , we then have , for example , @xmath189 and @xmath190 . this illustrates that it should be possible to achieve a regime where the ( coherent ) hamiltonian dynamics is dominant over the effective dissipation . note also that , for these parameters , readily achievable ground state magnetic level shifts ( @xmath176 ) of tens of mhz would suffice to ensure distinct raman channels . the same parameter regime could obviously be chosen for cavity mode @xmath45 , but if we consider the @xmath9 model as discussed in sec . [ sub_sect : spin_model_3 ] , then we might , for example , assume mode @xmath45 to be more strongly damped ( i.e. , the two cavity polarizations have different finesses ) , e.g. , @xmath191 , and , with smaller raman transition rate @xmath192 and detuning @xmath193 , we would then have @xmath194 . given these considerations , in the next section , where we examine the second - order transition of the @xmath9 model , we will typically employ the set of normalized parameters @xmath195 , which give a critical coupling strength @xmath196 . finally , we note that the rate for single - atom spontaneous emission ( neglected in our model ) is estimated by @xmath197 , where an atomic exited state linewidth of @xmath198 has been assumed . we focus first on the positive field case ( @xmath199 ) of the @xmath9 lmg model with ferromagnetic interactions ( @xmath6 ) , for which a second - order phase transition occurs as the magnitude of the interaction strength is varied @xcite . this transition will turn out to be similar to the one recently studied in the dissipative dicke model with resonant atom - cavity interactions ( as considered in ref . however , it should be noted that in the dicke model the cavity field plays an intrinsic role in the dynamics and associated critical behavior , unlike in our present model where it has been adiabatically eliminated . consequently atom - field entanglement is effectively negligible in the present context , while atom - atom entanglement is significant and will be the focus of our study . in sec . [ sect : lin_master_eqn ] we consider the spin master equation in a linearized regime , appropriate for @xmath200 , and determine the transmission spectrum of a weak probe laser . spin - spin entanglement is studied in sec . [ sect : entanglement ] both in the thermodynamic limit and for finite @xmath2 ; specifically the behavior of the steady - state entanglement , as well as entanglement dynamics , is examined in the vicinity of the quantum phase transition . in this section we study the master equation model ( [ eq : master_equation_gamma0_lmg_model ] ) in the thermodynamic limit by linearizing the quantum fluctuations around the mean - field state . note that the atom - cavity coupling strengths appearing in the effective coupling constants ( [ eq : def(d ) ] ) and ( [ eq : def(e ) ] ) scale as @xmath201 , where @xmath202 is the cavity mode volume . the thermodynamic limit corresponds to @xmath203 and @xmath204 with @xmath205 , the atomic density in the cavity , constant . since the thermodynamic limit does not alter the effective coupling strengths , which scale as @xmath206 , we will henceforth refer to the thermodynamic limit as @xmath207 @xcite . firstly , we present the semiclassical analysis which determines the mean - field state relevant for @xmath200 . we then expand the angular momentum operators around the semiclassical steady state using the holstein - primakoff representation , thus obtaining a linearized version of the master equation , the eigenvalues of which are subsequently analyzed . finally , we calculate , for the linearized model , the transmitted amplitude of a weak probe laser through the atom - cavity system as a function of the probe frequency , i.e. , the probe transmission spectrum . this physically measurable quantity probes the energy , or eigenvalue , structure of the system and , as we will see , provides clear signatures of the dynamical quantum phase transition . the equations of motion for the expectation values of the spin components of the bloch vector , @xmath208 , @xmath209 , and @xmath210 , are readily derived from the master equation ( [ eq : master_equation_gamma0_lmg_model ] ) , but do not form a closed set of equations . however , by factorizing all terms @xmath211 with @xmath212 , which corresponds to neglecting quantum fluctuations , we obtain a closed set of equations , which we call the semiclassical equations of motion from hereon . introducing the notation @xmath213 , @xmath214 , @xmath215 , where @xmath166 , the semiclassical equations of motion are found to be @xmath216 with the constraint @xmath217 corresponding to conservation of angular momentum . the steady - state solutions of these equations of motion exhibit a bifurcation at a critical coupling strength @xmath218 ( note @xmath219 for @xmath220 ) . for @xmath221 the stable steady - state solutions are @xmath222 while for @xmath223 they become @xmath224 where @xmath225 the bifurcation at @xmath110 is illustrated in fig . [ fig : semicl ] , where , to facilitate a comparison between semiclassical and finite-@xmath2 solutions ( computed from numerical solution of the master equation ) , we plot the second - order moments @xmath226 , @xmath227 , and @xmath228 ( since the finite-@xmath2 master equation gives @xmath229 for all @xmath4 ) . we note that the two approaches are already in reasonable agreement for @xmath230 . [ h!t ] steady - state second - order moments for @xmath231 , @xmath232 , @xmath233 , and @xmath234 ( dotted ) , @xmath235 ( short dashed line ) , @xmath236 ( long dashed line).,width=325 ] the quantum fluctuations that are neglected in the semiclassical analysis can be included in the limit @xmath237 as a first - order correction . this is achieved by using the holstein - primakoff ( hp ) representation of the angular momentum operators @xcite , which in the present context takes the form @xmath238 where @xmath239 and @xmath240 are bosonic annihilation and creation operators , respectively , satisfying @xmath241=1 $ ] . in particular , if @xmath200 and @xmath242 , i.e. , @xmath243 ( so that the bloch vector points essentially along the @xmath244 axis ) , then the hp representation of @xmath164 and @xmath165 can be reduced to @xmath245 and @xmath246 , effectively linearizing the dynamics . in the normal phase ( @xmath221 ) , this approach can be applied immediately since the steady - state solutions @xmath247 . however , in the broken phase ( @xmath223 ) , the steady - state solutions @xmath248 , i.e. , the bloch vector is rotated away from the @xmath244 axis , and the hp representation is most conveniently applied with respect to the new orientation of the bloch vector . we do this by first rewriting the semiclassical steady - state solutions in terms of spherical coordinates @xmath249 and @xmath250 as @xmath251 , @xmath252 , and @xmath253 , and then applying a unitary rotation @xmath254 around an axis @xmath255 , so that the transformed operators @xmath256 describe quantum fluctuations around the semiclassical steady state . the hp representation ( [ eq : hp_j_z])-([eq : hp_j_- ] ) and subsequent large-@xmath2 expansion is then applied to the operators @xmath257 . the master equation obtained in this way may be written , for both phases , in the general form ( omitting constant energy terms in the hamiltonian ) @xmath258 + \gamma_{+,k } d[c_k^\dagger]\rho + \gamma_{-,k } d[c_k]\rho \nonumber \\ & & + \gamma_{+,k}^s \left [ 2c_k\rho c_k + 2 c_k^\dagger \rho c_k^\dagger -\{c_k^2+(c_k^\dagger)^2,\rho \ } \right ] \nonumber\\ & & -i \gamma_{-,k}^s \left [ -2c_k\rho c_k + 2 c_k^\dagger \rho c_k^\dagger -\{-c_k^2+(c_k^\dagger)^2,\rho \ } \right]\nonumber , \\ \label{eq : damping_spin_louvillian}\end{aligned}\ ] ] with @xmath259 \nonumber \\ & & ~~ + ia_{3,k } \left[(c_k^\dagger)^2-c_k^2 \right ] , \label{eq : hp_spin_hamiltonian}\end{aligned}\ ] ] where @xmath260 and @xmath261 ( @xmath262 ) denotes the bosonic operator for the normal ( broken ) phase . the coefficients in the normal phase are given by @xmath263 while in the broken phase they are given by @xmath264 \nonumber \\ & & + \frac{\gamma_b}{4\lambda \lambda^2}\sqrt{\lambda^2-\gamma_b^2}(4h\lambda_{\rm c } -2\lambda\lambda ) , ~~ \\ \gamma_{-,>}^s & = & \frac{\gamma_a \gamma_b}{2\lambda^2 \lambda } ( \gamma_b^2 -4h^2 + 4h \sqrt{\lambda^2 - \gamma_b^2 } ) \nonumber \\ & & + \frac{\gamma_b^2}{4\lambda\lambda^2 } ( \lambda^2 -4h^2 ) .\end{aligned}\ ] ] note that the hamiltonian ( [ eq : hp_spin_hamiltonian ] ) does not contain any terms linear in @xmath265 and @xmath266 , which is a consequence of the applied rotation , and also means that @xmath267 . while the coefficients for the broken phase are rather complicated , they do simplify considerably in the limit of very large @xmath4 ; in particular , for @xmath268 one finds @xmath269 , @xmath270 , and @xmath271 , while @xmath272 , @xmath273 , and @xmath274 . the master equation then corresponds to that of a simple quantized harmonic oscillator coupled to a somewhat unconventional ( squeezed - type ) reservoir @xcite . it is interesting to examine the eigenvalues associated with the linear set of equations of motion for the first - order moments @xmath275 , which may be expressed as @xmath276 , where @xmath277 and @xmath278 is a @xmath279 matrix . the real and imaginary parts of these eigenvalues are plotted in fig . [ fig : eig_lin ] for our characteristic set of numerical parameters . we note that except for the region near zero coupling strength the eigenvalues exhibit very similar behavior to that found in the dissipative dicke model @xcite . in the normal phase ( @xmath280 ) the eigenvalues of @xmath278 are given by @xmath281 the imaginary parts of which go to zero at the point @xmath282 , with a characteristic scaling of @xmath283 . for @xmath284 the eigenvalues are real and distinct , with one going to zero at @xmath110 ( i.e. , critical slowing down ) and the other to @xmath285 . in the broken phase ( @xmath223 ) the eigenvalues of @xmath278 are given by @xmath286 in the region @xmath287 , where @xmath288 , the eigenvalues are complex conjugate pairs with a real part that diminishes for @xmath289 like @xmath290 . provided @xmath291 , then @xmath292 and the imaginary parts vanish as @xmath4 approaches @xmath293 from above with the scaling @xmath294 ( which can be shown using a first - order taylor series expansion about @xmath295 ) . the imaginary parts are zero in the interval @xmath296 , while the real parts again approach @xmath297 and @xmath285 , respectively , as @xmath298 . if @xmath299 then @xmath300 , and the eigenvalues are complex conjugate pairs immediately above the critical point . in this situation , the dissipation is stronger than the hamiltonian dynamics ; this is also an interesting regime , but not one that we will consider in the present paper . , as given by eqs . ( [ eq : eig_normal_phase ] ) and ( [ eq : eig_broken_phase ] ) , for @xmath231 and @xmath233 . the right - hand column gives a magnified view of the region around @xmath301.,width=325 ] a standard way to examine the structure and dynamics of an atomic system is to measure the transmission of a ( weak ) probe laser field through the medium as a function of the probe frequency . this amounts simply to detecting the frequency response of the system to an applied field or `` force '' . a schematic diagram illustrating the setup for such a measurement in the present context is shown in fig . [ fig : implementation ] ( a ) . for our theoretical investigation of the transmission spectrum we retain the two cavity modes in our model and make use of the input - output theory of open quantum systems @xcite . in particular , our starting point is the atom - cavity hamiltonian ( [ eq : heff ] ) and we again consider the limit @xmath200 , so that we can perform a linearization . to do this , we follow our previous working and determine the stable semiclassical steady - state amplitudes of the atom - cavity system from the semiclassical ( i.e. , factorized ) equations of motion for the moments @xmath302 . note that the steady - state cavity mode amplitudes in this approach can be expressed in terms of the atomic amplitudes as ( neglecting terms proportional to @xmath303 and setting @xmath145 ) @xmath304 using the hp representation of the atomic spin operators and linearizing about the semiclassical steady states as before leads to the following hamiltonian for the normal and broken phases , @xmath305 where @xmath306 , @xmath307 and @xmath308 denote the annihilation operators for the intracavity modes in the normal and broken phases , and the coefficients @xmath309 are given in appendix [ sect : appone ] . employing the quantum langevin equations of the input - output theory of open quantum systems we can analytically solve for any cavity output correlations and spectra of interest @xcite . here , however , we focus simply on the amplitude of a probe laser field transmitted through the system and into the output field , as depicted in fig . [ fig : implementation ] ( a ) . we consider only the case in which a probe laser of frequency @xmath180 ( in the rotating frame ) and amplitude @xmath310 drives cavity field mode @xmath45 . the analytical expression for the amplitude of the transmitted probe , @xmath311 , is rather complicated , but if we restrict ourselves to a frequency range where @xmath312 ( also with @xmath313 ) , then for @xmath221 the transmitted probe intensity is well approximated by @xmath314 + i\gamma_b } \right . \nonumber \\ & & \left . \left . - \frac{\left ( \sqrt{h}-\sqrt{h-\lambda } \right)^2}{\left [ \nu + 2\sqrt{h(h-\lambda ) } \right ] + i\gamma_b } \right\ } \right\|^2 , \label{eq : tp_nu}\end{aligned}\ ] ] where we have normalized the intensity such that it takes a maximum value of unity for an empty cavity . this form for @xmath315 highlights the presence of atomic resonances at the frequencies @xmath316 , superimposed on a broad background corresponding to the bare cavity mode resonance . note that this is in contrast to the findings in the dissipative dicke model @xcite where the probe laser transmission spectrum exhibits strongly coupled atom - cavity resonances . in fig . [ fig : trans_spect ] we plot the transmission spectrum [ computed from the full theory note that the approximate expression ( [ eq : tp_nu ] ) is in good agreement for the parameters chosen ] for a series of values of @xmath4 on either side or the transition . note that we have chosen @xmath317 here in order to highlight the main features of the spectrum more clearly . for @xmath318 , we observe , at @xmath319 , a single sharp dip of width @xmath320 in the envelope of cavity mode resonance , corresponding to a cavity - mediated collective atomic emission resonance ( @xmath321 ) ; at this @xmath4 , spin - spin interactions mediated by cavity mode @xmath11 are small and have little effect on the spectrum . as the interaction strength @xmath4 is increased , this dip moves to smaller frequencies and reduces in depth ( eventually inverting ) , while a peak emerges at the corresponding negative frequency . the positions and widths of these features reflect the real and imaginary parts , respectively , of the eigenvalue structure of the system , while their `` intensities '' also relate to the populations of the energy levels . at @xmath322 the two peaks merge into a single peak centered at @xmath323 , with a height @xmath324 . then , as @xmath298 , this peak diverges ( corresponding to eigenvalue @xmath325 ) in a pronounced signature of the phase transition . a similar divergence in the probe laser transmission spectrum is found in the dissipative dicke model @xcite . just above the critical point , two peaks reappear in the spectrum and move apart with increasing @xmath4 , as shown in fig . [ fig : trans_spect ] . the negative frequency peak diminishes in strength , while the peak at positive frequency inverts to a dip , which narrows and moves to increasingly larger frequencies . in fact , for @xmath326 , its position is approximated by @xmath327 and its width by @xmath328 . , with microscopic parameters @xmath329 , @xmath330 , and @xmath331 , @xmath332 , giving @xmath317 . we set @xmath231 as usual . note that @xmath333 is chosen to give the indicated @xmath4 for the given choice of @xmath334 and @xmath335 , _ viz_. eq . ( [ eq : lambda_i ] ) and recalling that @xmath336 , while @xmath337 varies according to eq . ( [ eq : gamma_i ] ) , with @xmath338 when @xmath339 . , width=325 ] recently a criterion for bipartite entanglement in collective spin systems was derived @xcite , and the connection to spin squeezing established . for the case of symmetric states the criterion is both necessary and sufficient , and reads @xmath340 where @xmath341 . in this work , we will use the magnitude of @xmath342 as a quantitative measure of the entanglement in the system . note that for finite @xmath2 , and also in the linearized analysis , we have @xmath343 [ since there are no linear driving terms in the effective hamiltonians ( [ eq : spin_gamma_0_hamiltonian ] ) and ( [ eq : hp_spin_hamiltonian ] ) ] , and thus @xmath344 . note also that @xmath345 , which was shown to be equivalent to the concurrence , @xmath346 @xcite , in nondissipative lmg models @xcite . for finite @xmath2 we numerically solve the master equation for the steady - state density matrix and then compute the operator averages required to determine @xmath342 and @xmath348 . in fig . [ fig : fncssrot ] we plot @xmath352 as a function of @xmath4 and @xmath353 for @xmath354 . we see that below the critical point , @xmath355 , entanglement is present for a broad range of angles @xmath353 . however , as the critical point is approached the range of angles @xmath353 which gives nonzero entanglement , @xmath356 , becomes increasingly narrow . once above the transition , @xmath223 , the region of finite @xmath342 continues to narrow until it eventually disappears altogether . for @xmath354 , @xmath231 , @xmath232 , and @xmath233.,width=325 ] to help interpret the behavior of @xmath342 , we make use of a phase space representation of the atomic state that employs the spin coherent states , which are defined by @xcite @xmath357 where @xmath358 , with @xmath249 and @xmath250 corresponding to spherical coordinates , and @xmath359 are the dicke states with @xmath360 $ ] ( for our system , @xmath166 ) . using these states we can define the spin @xmath361-function , @xmath362 fig . [ fig : spin_qfunc ] displays @xmath363 on the surface of the bloch sphere for @xmath364 and for a series of interaction strengths @xmath4 . below the critical point , @xmath363 is single - peaked and centered around the top of the bloch sphere ( @xmath365 ) , with little obvious angular dependence . correspondingly , the entanglement measure @xmath342 is finite over a rather broad range of @xmath353 , with a maximum close to @xmath366 ( i.e. , near @xmath367 ) . note that this slight shift of the optimum away from @xmath366 is a consequence of the dissipation ( @xmath153 ) in the system . -function , @xmath368 , on the bloch sphere for ( a ) @xmath369 , ( b ) @xmath370 , ( c ) @xmath371 , and ( d ) @xmath372 , with @xmath364 , @xmath231 , @xmath232 , and @xmath233 . note that dark blue corresponds to the minimum value of zero of @xmath373 while dark red indicates the maximum value of @xmath373.,width=325 ] as @xmath4 increases towards the critical point , @xmath363 becomes increasingly elongated along a direction close to the @xmath49 axis , until , at the transition , it splits into two peaks located approximately at the two semiclassical steady - state amplitudes ( [ eq : semicl_ss_sols_(b ) ] ) and ( [ eq : semicl_ss_sols_(c ) ] ) . these peaks continue to move apart in phase space as the interaction strength is increased further ; eventually both peaks will lie in the equatorial plane corresponding to @xmath374 and @xmath375 . correspondingly , the range of @xmath353 over which @xmath342 remains finite becomes increasingly narrow and is focussed around an axis perpendicular to that along which the two peaks lie . this narrowing of the `` width '' of @xmath342 can be explained by noting that , since @xmath343 , we have @xmath376 ^ 2 \rangle}$ ] . for increasing interaction strength @xmath223 , @xmath226 becomes of order @xmath377 ( see fig . [ fig : semicl ] ) , and so the optimal choice of @xmath353 becomes more critical . in fact , one can show for @xmath223 that the range of @xmath353 over which @xmath378 scales as @xmath379 . next , we consider the rescaled concurrence , @xmath348 , as a function of the interaction strength @xmath4 . in fact , one finds that @xmath380 i.e. , @xmath381 is simply the optimal value of the quantity @xmath342 just considered . in fig . [ fig : c_r_second_fn_and_hp_combo ] we plot @xmath348 versus @xmath4 and observe that the entanglement reaches a maximum for @xmath4 close to @xmath110 [ at finite @xmath2 the critical point is slightly shifted from @xmath110 as given in eq . ( [ eq : lambda_c ] ) ] . this peaking of the entanglement at the quantum phase transition has been conjectured and demonstrated theoretically for the equivalent closed ( nondissipative ) spin models @xcite . our results confirm that this behavior can persist in steady state in our nonequilibrium , open - system version of these models , and can in principle be measured within our proposed setup . for @xmath354 ( dashed line ) and in the thermodynamic limit ( solid line ) with @xmath231 , @xmath232 , and @xmath233.,width=325 ] in the linearized treatment ( @xmath237 ) of the hp representation , we can write @xmath382 , where @xmath383 . noting that @xmath384 , the entanglement measure @xmath342 can be expressed as @xmath385 while the rescaled concurrence is @xmath381 is given by @xmath386 where @xmath387 and @xmath388 using the linearized master equation ( [ eq : linearised_master_equation ] ) , we can derive a closed set of equations for the second - order moments @xmath389 , @xmath390 , and @xmath391 , from which we may determine the steady - state solutions analytically . note that the fourth - order moment appearing in @xmath392 can be expressed in terms of second - order moments , since the states we are dealing with in this linearized approximation are necessarily gaussian . in the thermodynamic limit for @xmath231 , @xmath232 , and @xmath233 . above the critical point , the system is linearized about only one of the two semiclassical steady - state amplitudes ; hence the less - sensitive dependence of @xmath352 on @xmath353 for @xmath223 as compared with the finite-@xmath2 results.,width=325 ] in fig . [ fig : hpcssrot ] we plot @xmath342 as a function of @xmath353 and @xmath4 , as determined from the linearized hp representation . the behavior below the critical point ( @xmath221 ) is very similar to the finite-@xmath2 case . however , the behavior above the critical point ( @xmath223 ) is very different . here , the sensitivity of @xmath342 to @xmath353 , for @xmath393 , is much less critical because the linearized model describes only the fluctuations around one of the two semiclassical steady - state amplitudes ( i.e. , around one of the two lobes appearing in the spin @xmath361-function for @xmath223 ) . note that we can obtain plots of @xmath352 similar to fig . [ fig : fncssrot ] for the region @xmath223 , but determined from the linearized hp model ( with a finite value of @xmath2 ) , by making a rotation back to the original coordinate system and then setting , by hand , @xmath394 , to mimic an equal , incoherent mixture of the states associated with the two semiclassical amplitudes . finally , returning to fig . [ fig : c_r_second_fn_and_hp_combo ] , we have plotted @xmath348 as a function of @xmath4 , computed from the hp model in the thermodynamic limit . again , @xmath348 corresponds to the value of @xmath342 optimized over @xmath353 , and , since the optimal @xmath353 corresponds to an axis perpendicular to the ( above - transition ) splitting of the semiclassical amplitudes , we expect , and indeed find , good agreement with the finite-@xmath2 results over the full range of @xmath4 . if we make the simplifying assumption that @xmath395 , then , for @xmath280 one can show that @xmath396 and @xmath397 this shows reasonable agreement with the plot , but reaches a maximum value of 0.5 at the critical point . we now consider the dynamics of the entanglement ; starting from an initially unentangled state , we examine the time evolution of the state and its entanglement as quantified by @xmath398 . the initial state is taken as the @xmath399 ground state , i.e. , the state with all atomic spins pointing up ( which is a convenient state to prepare in an experiment ) . as in the previous section , we compute the entanglement both numerically for finite @xmath2 and analytically for @xmath200 in the linearized approximation . in fig . [ fig : finitenconcurrtimedep ] we plot @xmath398 versus @xmath4 and time @xmath400 for @xmath354 . at long times we recover the results of the previous section , but at short times the behavior as a function of @xmath4 is quite different ; the entanglement rises to a high value and remains at that value for increasing interaction strength @xmath4 . this behavior can be attributed to the hamiltonian dynamics , which dominate the dissipation at short times and can create highly entangled states . the potential of such hamiltonian dynamics for generating such highly entangled states has been proposed previously , for example , in refs . note , however , that the presence of the term @xmath401 in our system hamiltonian tends to make the generated states more complicated and less straightforward to interpret . although the focus of this paper is on the quantum phase transition , it is clear that with a slight modification the scheme also has interesting potential for the controlled generation of specific , highly entangled multiatom states ( e.g. , greenberger - horne - zeilinger states ) . in connection with this , an important aspect of our implementation should be highlighted here : because both the effective interaction and dissipation of the spins is controlled by the optical laser fields , we can in principle `` freeze '' the state of the atomic system at any time by simply turning these fields off . as a function of @xmath4 and @xmath400 for @xmath354 , @xmath231 , @xmath232 , and @xmath233.,width=325 ] as a function of @xmath4 and @xmath400 in the thermodynamic limit , with @xmath231 , @xmath232 , and @xmath233 . above the critical point the dynamics is linearized around one of the two possible semiclassical steady - state amplitudes.,width=325 ] in the linearized regime , @xmath237 , we solve the equations of motion for the second - order moments , @xmath389 , @xmath390 , and @xmath391 , with the initial conditions @xmath402 , the results for @xmath398 are shown in fig . [ fig : hp_c_r_t ] . below the critical point the behavior is similar to that observed for finite @xmath2 . however , above the critical point , where the dynamics is linearized about only one of the two allowed semiclassical steady - state amplitudes , the rescaled concurrence is , as expected , quite different , owing to the more limited range of entangled states that the linearized ( gaussian ) theory can accommodate . we now turn to the case of a fixed , positive interaction strength ( @xmath6 ) of the @xmath9 model ( sec . [ sub_sect : spin_model_3 ] ) with variable effective field @xmath3 . in the absence of dissipation this model exhibits two second - order transitions as @xmath3 is varied , one occurring at positive @xmath3 ( the transition discussed in the previous section ) and the other , equivalent transition occurring at negative @xmath3 . however , in this section we show that with the addition of dissipation this model actually exhibits a first - order phase transition near @xmath404 ( note that in the absence of dissipation no such transition exists ) . as in the previous section , we begin with a study of the linearized spin master equation , including an eigenvalue analysis and calculation of the probe transmission spectrum , after which we focus again on the entanglement properties of the system . for numerical calculations we will typically employ the set of normalized parameters @xmath405 , which correspond to a critical effective field strength @xmath406 . as before , we consider the thermodynamic limit and linearize the master equation ( [ eq : master_equation_gamma0_lmg_model ] ) about the mean - field state . to do so , we first find the semiclassical steady - state solutions and then expand the angular momentum operators around these mean - field solutions using the holstein - primakoff representation . from the ( factorized ) semiclassical equations of motion for @xmath407 , @xmath408 , and @xmath409 , eqs . ( [ eq : semicl(a)])-([eq : semicl(c ) ] ) we again obtain the stable steady - state solutions . these exhibit discontinuities at the critical field strengths @xmath410 and @xmath411 . for @xmath412 the stable steady - state solutions are given by eq . ( [ eq : semicl_ss_sols ] ) , while for @xmath413 @xcite the stable steady states are given by eqs . ( [ eq : semicl_ss_sols_(a)])-([eq : semicl_ss_sols_(c ) ] ) . while outside the region @xmath414 the stable steady states are unique , inside the region @xmath414 both steady - state solutions ( [ eq : semicl_ss_sols ] ) and ( [ eq : semicl_ss_sols_(a)])-([eq : semicl_ss_sols_(c ) ] ) are in fact stable . however , for the characteristic parameters we consider here this region is very small ( @xmath415 ) . moreover , we have verified ( from a linearized analysis ) that the steady - state solution ( [ eq : semicl_ss_sols ] ) is more stable in the region @xmath414 and thus we will only consider this solution in that region . note that for larger values of the dissipation , @xmath153 , this region becomes more pronounced ( in this case all stable steady states should be considered @xcite ) , but this is beyond the regime we wish to consider here . the relevant stable steady - state solutions @xmath416 , @xmath417 and @xmath418 are plotted in fig . [ fig : semicl_af ] , together with results from numerical solutions of the master equation for a range of values of @xmath2 up to 100 ( at which agreement between the two approaches is already quite good ) . the discontinuous jump of @xmath416 at @xmath406 signifies the first - order phase transition . note that for the case @xmath7 the same first order transition occurs except that it is shifted to @xmath419 ( i.e. , in fig . [ fig : semicl_af ] all curves are flipped about @xmath411 ) . steady - state solutions for @xmath420 , @xmath232 , @xmath233 , and @xmath234 ( dotted ) , @xmath235 ( short dashed line ) , @xmath236 ( long dashed line ) . note the inset in the bottom panel is a magnified plot of the semiclassical solution of @xmath421.,width=325 ] here , we again include the quantum fluctuations for @xmath237 as a first - order correction by linearizing the spin operators around the semiclassical steady state via the hp representation . for @xmath422 ( normal phase ) the linearized master equation is identical to eq . ( [ eq : linearised_master_equation ] ) with @xmath423 , while for @xmath424 ( broken phase ) the linearized master equation is also identical to eq . ( [ eq : linearised_master_equation ] ) but with @xmath425 . the eigenvalues of the linearized system , i.e. , of the matrix @xmath278 , where @xmath276 and @xmath426 , for @xmath422 are given by eq . ( [ eq : eig_normal_phase ] ) while for @xmath424 they are given by eq . ( [ eq : eig_broken_phase ] ) . in fig . [ fig : eig_lin_first ] the real and imaginary parts of the eigenvalues are plotted for our characteristic set of parameters . in the normal phase ( @xmath422 ) we see that the imaginary parts go to zero at the point @xmath427 . in the region @xmath428 both eigenvalues are real and distinct , with one going to zero at @xmath429 and the other going to @xmath285 . this behavior is the same as that found for the second - order transition of the earlier section . however , immediately above the transition , @xmath424 , the eigenvalues become complex conjugate pairs with a nonzero real part that diminishes for @xmath430 like @xmath290 . this discontinuous jump of the eigenvalues is an additional signature of the first - order phase transition . , as given by eqs . ( [ eq : eig_normal_phase ] ) and ( [ eq : eig_broken_phase ] ) , for @xmath420 , and @xmath233 . the right - hand column gives a magnified view of the region around @xmath431.,width=325 ] we determine the probe transmission spectrum in the linearized regime following exactly the same calculations as outlined in sec . [ sect : transmission_spectrum ] . the linearized hamiltonian describing the full atom - cavity system is easily obtained ; for @xmath422 it is given by eq . ( [ eq : linearised_atom_cavity_hamiltonian ] ) with @xmath365 and @xmath432 , while for @xmath424 it is also given by eq . ( [ eq : linearised_atom_cavity_hamiltonian ] ) , but with @xmath249 and @xmath250 given according to the semiclassical solutions eqs . ( [ eq : semicl_ss_sols_(a)])-([eq : semicl_ss_sols_(c ) ] ) as explained in sec . [ sect : hp_rep ] . , with microscopic parameters @xmath329 , @xmath330 , @xmath433 , @xmath331 , and @xmath332 , giving @xmath420 , @xmath338 and @xmath317.,width=325 ] restricting ourselves again to a frequency range where @xmath434 ( also with @xmath435 ) , then as previously an approximate expression for the transmitted probe intensity can be derived in the normal phase for @xmath422 and takes exactly the same form as eq . ( [ eq : tp_nu ] ) . in fig . [ fig : trans_spect_af ] we plot the transmission spectrum for a series of values of @xmath3 across the critical point @xmath429 . in the normal phase ( @xmath436 ) , we observe the same behavior as in the normal phase of the system in the previous section ( @xmath355 ) , except that the orientations of the peaks and dips have inverted in accordance with the change of sign of the field ( @xmath412 ) . the central peak diverges as the critical point is approached from below in the normal phase , again signifying the phenomenon of critical slowing down in the vicinity of the phase transition . however , immediately above the critical point , the spectrum splits discontinuously into two sharp peaks of width @xmath437 , located at frequencies @xmath438 . this jump from a single divergent peak at @xmath323 to a two - peaked spectrum offers a pronounced , observable signature of the first - order transition . we compute , as before , the entanglement measures @xmath342 and @xmath348 , both numerically for finite @xmath2 and analytically for @xmath237 in the linearized regime . [ fig : fn_c_varphi_af ] shows a plot of @xmath342 as a function of @xmath3 and @xmath353 for @xmath354 . we see that , well away from the critical point , substantial entanglement is present over a broad range of angles @xmath353 . as the critical point is approached from below , significant entanglement persists , but for a somewhat narrower range of angles @xmath353 . however , immediately above the critical point the entanglement drops suddenly to zero for all values of @xmath353 . as a function of @xmath3 and @xmath353 for @xmath354 , @xmath420 , @xmath232 , and @xmath233.,width=325 ] to help understand these results we again utilize the atomic coherent state representation and study the spin @xmath361-function @xmath373 . in fig . [ fig : spin_qfunc_first ] we plot @xmath373 on the bloch sphere for a series of values of @xmath3 in the vicinity of the first - order transition . well below the critical point , in the normal phase , @xmath373 is a single peaked function with little angular dependence . correspondingly , @xmath342 is nonzero over a broad range of @xmath353 , with a maximum close to @xmath439 ( i.e. , near @xmath440 ) . again , note that this slight shift of the optimum away from @xmath439 is a consequence of the dissipation ( @xmath153 ) in the system . as @xmath3 increases towards the critical point , @xmath373 becomes increasingly stretched along the @xmath441 axis . as the critical point is traversed @xmath373 rapidly rotates around from the @xmath441 axis towards the @xmath49 axis , and splits into the familiar two - lobed structure associated with the two semiclassical steady - state amplitudes of the broken phase . at the same time as the critical point is approached , the range of @xmath353 over which @xmath342 remains finite narrows and immediately above the critical point it drops abruptly to zero for all choices of @xmath353 . this behavior is akin to the behavior we observed for large interaction strength in the regime of the previous section , where @xmath226 becomes of order @xmath377 ( see fig . [ fig : semicl_af ] ) , which severely restricts the range of @xmath353 for which @xmath356 . note that at larger values of @xmath3 than displayed in fig . [ fig : fn_c_varphi_af ] , the entanglement , @xmath342 , once again becomes nonzero ( centered around @xmath442 ) coinciding with the broken phase behavior of the second - order transition discussed in the previous section . in fig . [ fig : c_r_first_fn_and_hp_combo ] we plot the rescaled concurrence @xmath348 as a function of the effective field strength @xmath3 and again find that close to the critical point , @xmath429 , the entanglement reaches its peak value . although the equivalent closed system would not feature a maximum in the entanglement near @xmath429 ( due to the complete absence of a phase transition ) , this result is in agreement with a conjecture concerning entanglement in open systems at quantum critical points @xcite . versus @xmath3 for @xmath354 ( dashed line ) and in the thermodynamic limit ( solid line ) with @xmath420 , @xmath232 , and @xmath443.,width=325 ] in the linearized treatment ( @xmath200 ) we obtain very similar plots of @xmath342 to those of finite @xmath2 ( fig . [ fig : fn_c_varphi_af ] ) and for @xmath348 the result is shown in fig . [ fig : c_r_first_fn_and_hp_combo ] . in the limit where we consider @xmath395 we can again obtain an approximate expression for the rescaled concurrence ( for @xmath444 ) given , in this instance , by @xmath445 this again has a maximum value of @xmath446 at the critical point , and , for large @xmath447 , drops off like @xmath448 , in reasonable agreement with the plots . finally , in fig . [ fig : fn_c_r_t_af ] for @xmath354 we illustrate the time - dependent behavior of the rescaled concurrence , @xmath398 , for varying @xmath3 , given an initial ( unentangled ) state with all spins up . once again , we observe an interesting oscillatory behavior of @xmath398 , with , in particular , highly entangled states generated by the hamiltonian dynamics at short times ( for almost all values of @xmath3 ) , before dissipation has had time to play a significant role . for the linearized regime ( @xmath237 ) a similar plot of @xmath398 can be obtained which agrees well with the finite @xmath2 result for @xmath422 but shows zero entanglement for almost all values of @xmath449 because of the restricted linearization around only one of the two permitted semiclassical steady - state amplitudes . for @xmath354 , with @xmath420 , @xmath232 , and @xmath233.,width=325 ] we have proposed in this paper a feasible cavity qed setup , consisting of a collective atomic pseudospin and two quantized cavity modes , which realizes a dissipative version of the lmg model in which the interacting spin system displays both first- and second - order nonequilibrium quantum phase transitions . the lossy cavity s output light fields can be utilized to monitor the system as the model parameters are varied ; specifically , we showed that the transmission spectra vary dramatically in the vicinity of the transition , with features that are characteristic of the criticality . a further important result is the steady - state entanglement criticality at the qpt and the possibility of directly observing this via homodyne detection of the cavity output fields . in particular , the entanglement can be quantified rather directly in terms of measurable atomic quadrature variances . we also observed an important sensitivity of the entanglement measure to the quadrature phase angle in the critical regimes , which we were able to interpret by employing an atomic phase space distribution . finally , we have considered how entanglement evolves in this system , observing not only the criticality at the qpt at long times ( corresponding to the steady state ) , but also a rich transient behavior at shorter times . for future studies , it is clear that the system we have proposed offers a variety of opportunities , such as ( i ) investigating phase transitions in response to variation of the strength of dissipation ( i.e. , @xmath153 ) , ( ii ) examining a system of multiple ( separately addressable ) atomic pseudospins all coupled to the same quantized cavity modes , which would permit the study of entanglement between different spin blocks @xcite , ( iii ) controlled preparation of robust ( insensitive to noise / environment ) , highly entangled states by evolution from an initial product state @xcite , ( iv ) measurement of more general atomic spin correlations and their evolution with time , which can also provide signatures of criticality in qpt s @xcite , ( v ) extending our system to accommodate more complex spin models , e.g. , by adding additional lasers to the setup explained in sec . [ sect : general_model ] to realize the so - called `` two - field model '' @xcite , and ( vi ) imposing some spatial variation on the cavity mode to provide , for example , short ranged interactions , which could be uniform or quasirandom . the authors thank a. daley and h. carmichael for discussions and acknowledge support from the austrian science foundation and from the marsden fund of the royal society of new zealand . in sec . [ sect : transmission_spectrum ] we gave the general form of the linearized hamiltonian of the joint atom - cavity system , eq . ( [ eq : linearised_atom_cavity_hamiltonian ] ) . the coefficients of this hamiltonian in terms of the system parameters , @xmath450 and the angles @xmath451 from sec . [ sect : hp_rep ] are @xmath452 , \\ a & = & \frac{\lambda_a}{2 } \left [ ( 1+\cos{\theta})+(1-\cos{\theta})(\sin{\phi}+i\cos{\phi})^2\right ] , \nonumber \\ \\ b_1 & = & \frac{\lambda_b}{2}\left [ ( 1-\cos{\theta})(\sin{\phi}+i\cos{\phi})^2\right ] , \\ b_2 & = & \frac{\lambda_b}{2}(1+\cos{\theta}).\end{aligned}\ ] ] note that for @xmath280 one has @xmath453 and @xmath432 , giving the simplified expressions @xmath454 , @xmath455 , @xmath456 , and @xmath457 . similar to sec . [ sect : hp_rep ] , we can also derive simplified expressions in the limit @xmath289 , i.e. , for @xmath458 , one has @xmath459 , @xmath460 , @xmath461 , and @xmath462 h. j. lipkin , n. meshkov , and a. j. glick , nucl . phys . * 62 * , 188 , 199 , 211 ( 1965 ) ; n. meshkov , a. j. glick , and h. j. lipkin , _ ibid_. * 62 * , 199 ( 1965 ) ; a. j. glick , h. j. lipkin , and n. meshkov , _ ibid_. * 62 * , 211 ( 1965 ) . for numerical solutions of the atomic collective - spin master equation we make use of : s. m. tan , _ quantum optics and computation toolbox for matlab _ , available at http://www.qo.phy.auckland.ac.nz/qotoolbox.html .	we propose a cavity qed setup which implements a dissipative lipkin - meshkov - glick model an interacting collective spin system . by varying the external model parameters the system can be made to undergo both first - and second - order quantum phase transitions , which are signified by dramatic changes in cavity output field properties , such as the probe laser transmission spectrum . the steady - state entanglement between pairs of atoms is shown to peak at the critical points and can be experimentally determined by suitable measurements on the cavity output field . the entanglement dynamics also exhibits pronounced variations in the vicinities of the phase transitions .
the observation of the paleomagnetic data @xcite have shown that , unlike the solar magnetic field , where the polarity reversals are strictly periodic , geomagnetic measurements of the last @xmath0 million years present rather sudden and occasional polarity reversals . the reversal process is normally very rapid with respect to the typical time interval between successive reversals , which may range from @xmath1 up to @xmath2 years @xcite . recent works on data analysis , experimental dynamo and theoretical modeling have inproved the knowledge of the earth dynamo . however , the main fundamental questions concerning the polarity reversals still remain unanswered @xcite . the nature of the triggers ( external or internal to earth ) and the physical mechanisms giving rise to the reversals , the reason for the long time variations in the average reversal rate ( cf . e.g. @xcite ) , are still open problems . the sequence of geomagnetic reversals ( see the example from the ck95 database @xcite shown in fig . [ fig1 ] ) seems to result from a of a stochastic process . the same behaviour is observed for experimental dynamo @xcite and from numerical simulations @xcite . while experimental dynamo is a recent excellent achievement , the numerical approach , namely the direct solution of the maghetohydrodynamics ( mhd ) equations ( see @xcite ) is still far from being satisfactory for a statistical analysis . however , reversals are also observed in field resulting from simplified models , such as few modes models @xcite , models of noise - induced switchings between two metastable states @xcite , or mean - field dynamo models with a noise - perturbed @xmath3 profile @xcite . of persistence times @xmath4 for ck95 database ( statistical errors are shown as vertical bars).,width=10 ] recently , it has been shown through a simple statistical analysis , that the reversals of the paleomagnetic field are not random @xcite , namely the statistics of interevent times ( @xmath5 , where @xmath6 is the time of the @xmath7-th event of the record ) departs from a poissonian distribution ( namely an exponential law @xmath8 , where @xmath9 represents the reversal occurrence rate @xcite ) , including a non - stationary poisson process , in which case a power - law distribution could arise from the superposition of poisson distributions with time variable rates @xmath10 , see @xcite . this result shows that geomagnetic reversals are clustered in time , probably because of presence of memory in the process generating polarity reversals . here we show that experimental dynamo reversals also are characterized by correlations and clustering , suggesting that the reversal process is a universal property of dynamo , which does not need any external triggering . in this section we will describe the statistical tool used in this work to test , as a zero - th order hypothesis @xmath11 , whether the observed sequence is consistent with a _ local poisson process_. the reversals rate profile @xmath10 being in principle unknown , the test should be independent on it . a method introduced in cosmology @xcite and more recently used for solar flares @xcite geomagnetic activity @xcite , random lasers in liquid crystals @xcite , and stock market analysis @xcite will be used here . consider the time sequence of reversals as a point - like process , and suppose that each reversal occurs at a discrete time @xmath6 . the suitably normalized local time interval @xmath12 between reversals can be defined by introducing @xmath13 as @xmath14 and @xmath15 by @xmath16 @xmath13 and @xmath15 are then the two persistence times following or preceeding a given reversal at @xmath6 . if the local poisson hypothesis @xmath11 holds , both @xmath13 and @xmath15 are independently distributed according to an exponential probability density : @xmath17 and @xmath18 with local rate @xmath19 . the distribution of the variable @xmath12 defined by @xmath20 will not depend on @xmath19 . for the surviving function of the probability density @xmath21 where @xmath22 is the probability density function of @xmath12 and @xmath23 \ ; , \ ] ] it can be easily shown that , under the hypothesis @xmath11 , @xmath24 that is , @xmath12 is a stochastic variable uniformly distributed in @xmath25 $ ] . in a process where @xmath15 s are systematically smaller than @xmath26 s , clusters are present and the average value of @xmath12 is greater than @xmath27 . on the contrary , when the process is characterized by voids , the average value of @xmath12 is less than @xmath27 . from time series , it is easy to calculate the surviving function @xmath28 and the probability density function @xmath22 . the test described above has been recently applied to four different datasets of geomagnetic polarity reversals , including the already mentioned ck95 citeprl , pepi , gafd . the probability density function @xmath22 is reported in fig . reffig2 for the ck95 datasets . a significant deviation from the uniform distribution was observed in all the datasets , due the presence of clusters . of the stochastic variable @xmath12 and corresponding surviving functions @xmath29 for all the empirical datasets . the theoretical probability expected under a poisson statistics is also shown.,width=10 ] so as to generate a toroidal component @xmath30 . ( b ) postive feed - back : the amplitude of @xmath30 is used to drive a power source which generates the current in the external loop . two helmoltz coils are set on either end of the cylindrical flow vessel ; @xmath30 is measured in the mid - plane by a hall probe connected to a bell gaussmeter . the measured value is fed into a linear amplifier whose output drives a kepco current source . in order to explore the role of the turbulent fluctuations , the amplifier has separate channels for the dc and flcutuating parts of the induction . , width=321 ] the dynamo laboratory model @xcite mimics an alpha - omega cycle where part of the dynamo cycle is generated by an external feed - back but the flow turbulence is still included and has a leading role . in order to achieve this in a simple laboratory dynamo , we relax the requirement that the current path be fully homogeneous , and we effectively prescribe an alpha mechanism by which a toroidal magnetic field generates an induced poloidal one . however , the omega poloidal to toroidal conversion still results from a fully turbulent process . our experimental fluid turbulent dynamo is very much inspired by a variation of the solid rotor dynamo proposed by sir edward bullard in the early 20th century , and described in figure 1 . two coaxial disks counter rotate at a rate @xmath31 . when an axial magnetic field @xmath32 is externally applied , the flow differential rotation induces a toroidal field @xmath33 , this is the omega effect . the value of this field is then used to drive a linear current amplifier in the loop that generates @xmath32 . the poloidal to toroidal conversion is entirely due to the fluid motion , and incorporates all turbulence effects . it has been extensively studied in previous `` open loop '' experiments . when @xmath32 is externally fixed , one has @xmath34 where @xmath35 is the magnetic reynolds number ( with @xmath9 the magnetic diffusivity of liquid gallium ) and @xmath36 is a `` geometric '' constant which in our experiment has been measured of the order of 0.1 . the toroidal to poloidal conversion is then obtained by feeding the axial coils with an electrical current linearly driven by a signal proportional @xmath37 , so that @xmath38 which reinforces @xmath39 with @xmath40 an adjustable gain . in such a closed loop setup , one then has @xmath41 , and self - sustained dynamo is reached as @xmath42 . clearly , the adjustable gain of the linear amplifier allows to adjust the value of @xmath43 to an experimentally accessible range . at this point it should be emphasized that although the feed - back scheme is very similar for the bullard rotor dynamo and for our fluid experiment , the expected dynamics is much richer because of the strong fluctuations in the turbulent flow , where reynolds numbers in excess of @xmath44 are reached . indeed , the von krmn flow is known for its complex dynamics , presenting not only small scale turbulent fluctuations but also large scale ones for instance fluctuation up to 114% for the differenteial rotation effect has been reported ) . compared to the 1963 pioneering experiment of lowes and wilkison with solid rotor motions , the study here fully incorporates fluid turbulence and the associated fluctuations of magnetic induction . the role of these fluctuations , inherent to large reynolds number flows , remains one of the mysterties of natural dynamos , and of noisy instabilities in a braoder framework . in this experiment the value of the magnetic field at saturation @xmath45 is fixed by the maximum current that can be drawn from the power amplifier driving the coils . we measure @xmath46 g , a value such that the lorentz forces can not modify the hydrodynamic flow since it yields an interaction parameter of the order of @xmath47 . the saturation of the instability is therefore driven by the amplifier non - linearities rather than by the back - reaction of lorentz forces on the dynamical velocity field . as a consequence , the @xmath48 component of the generated magnetic field saturates at the same mean amplitude @xmath45 for all rotation rates ( @xmath45 corresponds to the magnetic field generated by the coils when the current source is saturated ) , the saturation amplitude of the toroidal field @xmath49 linearly increases with @xmath31 . another noteworthy observation is that the presence of turbulent fluctuations plays a crucial role in the triggering of the magnetic field reversals . in the experimental results reported here , the current source is driven by an amplifier whose input is @xmath50 , with @xmath51 the low pass dc component of @xmath30 and @xmath52 its ac fluctuating part . this arrangement allows to study separately the role of slow variations and turbulent fluctuations in the feed - back loop . in the results reported in this article , we have set @xmath53 . a homopolar dynamo , i.e. without reversals , was obtained for smaller values of @xmath54 or when the @xmath52 input in the amplifier was replaced by a synthetic gaussian white noise ( even with a high amplitude ) . we show here the results of the @xmath12-test obtained in a realization ( serie27 ) with @xmath55 hz , and cutoff frequency @xmath56 mhz . similar resutls were observed with different parameters , and this study is left for more extended work . figure reffig3 shows the reversals surviving function in the case described here . the behaviour is very similar to the paleomagnetic case , indicating again presence of clustering and correlations , rather than a random behaviour . this indicates that the mechanism responsible for the clustering is present in both dynamoes , suggesting some sort of universality of the process . of the stochastic variable @xmath12 and corresponding surviving functions @xmath29 for the experimental dataset described in the text . the theoretical probability expected under a poisson statistics is also shown.,width=377 ] in this short paper , the statistical properties of persistence times between geomagnetic reversals have been investigated . we performed a statistical test which showed that geomagnetic reversals are produced by an underlying process that is far from being locally poissonian , as recently conjectured by @xcite . thus , the sequence of geomagnetic reversals is characterized by time correlations . as spontaneous reversals of the geodynamo field have been observed in high resolution numerical simulations @xcite , the main results contained in this paper seem to indicate that such reversals could be related to the non - linear nature of the turbulent dynamo . in order to confirm this conjecture , we performed the statistical test mentioned above on recent results from laboratory dynamo . our analysis has shown that the departure from poisson statistics found in the paleomagnetic data , related with the long range correlations introduced by the chaotic dynamic of the system citepepi , gafd , are also present in the laboratory dynamo . such correlations can be associated with the presence of some degree of memory in the underlying dynamo process @xcite which gives rise to clustering of reversals . a. kageyama , m. kameyama , s. fujihara , m. yoshida , m. hyodo , y. tsuda , a 15.2 tflops simulation of geodynamo on the earth simulator . in proc . acm / ieee supercomputing conference , sc2004 , pitssburgh ( 2004 ) .	statistical properties of the temporal distribution of polarity reversals of the geomagnetic field are commonly assumed to be a realization of a renewal poisson process with a variable rate . however , it has been recently shown that the polarity reversals strongly depart from a local poisson statistics , because of temporal clustering . such clustering arises from the presence of long - range correlations in the underlying dynamo process . recently achieved laboratory dynamo also shows reversals . it is shown here that laboratory and paleomagnetic data are both characterized by the presence of long - range correlations .
on many crystal surfaces step edge barriers are observed which prevent interlayer ( downward ) hopping of diffusing adatoms @xcite . in homoepitaxy from a molecular beam this leads to a growth instability which can be understood on a basic level : adatoms form islands on the initial substrate and matter deposited on top of them is caught there by the step edge barrier . thus a pyramid structure of islands on top of islands develops . at late stages of growth pyramids coalesce and form large `` mounds '' . their lateral size @xmath3 is found experimentally to increase according to a power law in time , @xmath0 with @xmath4 6 depending on the material and , possibly , deposition conditions used . a second characteristic is the slope of the mounds hillsides @xmath5 , which is observed to either approach a constant ( often referred to as a `` magic slope '' since it does not necessarily coincide with a high symmetry plane ) or to increase with time as @xmath6 @xcite . the surface width ( or the height of the mounds ) then grows as @xmath7 with @xmath8 , where @xmath9 for the case of magic slopes . on a macroscopic level these instabilities can be understood in terms of a growth - induced , slope - dependent surface current @xcite . since diffusing adatoms preferably attach to steps from the terrace _ below _ , rather than from _ above _ , the current is uphill and destabilizing . the concentration of diffusing adatoms is maintained by the incoming particle flux ; thus , the surface current is a nonequilibrium effect . the macroscopic view is quantified in a continuum growth equation , which has been proposed and studied by several groups @xcite . the goal of the present contribution is to obtain analytic estimates for the scaling exponents and scaling functions of this continuum theory . to give an outline of the article : in the next section we briefly introduce the continuum equations of interest . a simple scaling ansatz , presented in section 3 , leads to scaling relations and inequalities for the exponents @xmath10 and @xmath11 . in section 4 we present a solvable mean field model for the dynamics of the height height correlation function . up to logarithmic corrections , the relations of section 3 are corroborated . finally , in the concluding section 5 the mean field correlation functions are compared to numerical simulations of the full growth equation , and the special character of the mean field approximation is pointed out . under conditions typical of molecular beam epitaxy ( mbe ) , evaporation and the formation of bulk defects can be neglected . the height @xmath12 of the surface above the substrate plane then satisfies a continuity equation , @xmath13 where @xmath14 is the incident mass flux out of the molecular beam . since we are interested in large scale features we neglect fluctuations in @xmath14 ( `` shot noise '' ) and in the surface current ( `` diffusion noise '' ) . in general , the systematic current @xmath15 depends on the whole surface configuration . keeping only the most important terms in a gradient expansion , such as @xmath16 , though they may well be relevant for the coarsening behavior of the surface @xcite . ] , subtracting the mean height @xmath17 , and using appropriately rescaled units of height , distance and time @xcite , eq . ( [ cont1 ] ) attains the dimensionless form @xmath18 the linear term describes relaxation through adatom diffusion driven by the surface free energy @xcite , while the second nonlinear term models the nonequilibrium current @xcite . assuming in - plane symmetry , it follows that the nonequilibrium current is ( anti)parallel to the local tilt @xmath19 , with a magnitude @xmath20 depending only on the magnitude of the tilt . we consider two different forms for the function @xmath20 : \(i ) within a burton - cabrera - frank - type theory @xcite , for small tilts the current is proportional to @xmath21 , and in the opposite limit it is proportional to @xmath22 . this suggests the interpolation formula @xcite @xmath23 . since we are interested in probing the dependence on the asymptotic decay of the current for large slopes , we consider the generalization @xmath24}.\ ] ] since @xmath25 also in the extreme case of complete suppression of interlayer transport @xcite , physically reasonable values of @xmath26 are restricted to @xmath27 . \(ii ) magic slopes can be incorporated into the continuum description by letting the nonequilibrium current change sign at some nonzero tilt @xcite . a simple choice , which places the magic slope at @xmath28 , is @xmath29};\ ] ] a microscopic calculation of the surface current for a model exhibiting magic slopes has been reported by amar and family @xcite . the stability properties of a surface with uniform slope @xmath30 are obtained by inserting the ansatz @xmath31 into ( [ cont2 ] ) and expanding to linear order in @xmath32 . one obtains @xmath33 \epsilon,\ ] ] where @xmath34 denotes the partial derivative parallel ( perpendicular ) to the tilt * m. the coefficients are @xmath35 and @xmath36 . if one of them is negative , the surface is unstable to fluctuations varying in the corresponding direction : variations perpendicular to m * will grow when the current is uphill ( when @xmath37 ) , while variations in the direction of * m * grow when the current is an increasing function of the tilt . both models have a change in the sign of @xmath38 , model ( i ) at @xmath39 , model ( ii ) at @xmath40 . for model ( i ) @xmath41 always , corresponding to the step meandering instability of bales and zangwill @xcite . in contrast , for model ( ii ) the current is downhill for slopes @xmath42 , and these surfaces are absolutely stable . in this work we focus on singular surfaces , @xmath43 , which are unstable in both models ; coarsening behavior of vicinal surfaces has been studied elsewhere @xcite . the situation envisioned in the rest of this article is the following : for solutions of the pde ( [ cont2 ] ) we choose a flat surface with small random fluctuations @xmath44 as initial condition . mostly the initial fluctuations will be uncorrelated in space , though the effect of long range initial correlations is briefly addressed in section [ meanfield ] . the fluctuations are amplified by the linear instability , and eventually the surface enters the late time coarsening regime that we wish to investigate . in this section we assume that in the late time regime the solution of ( [ cont2 ] ) is described by a scaling form , namely that the surface @xmath45 at time @xmath46 has the same ( statistical ) properties as the rescaled surface @xmath47 at time @xmath48 . the equal time height height correlation function @xmath49 then has a scaling form @xmath50 where the relevant lengthscales are the surface width @xmath51 , i.e. the typical height of the mounds , and their lateral size @xmath52 , given by the first zero of @xmath53 . these choices correspond to @xmath54 and @xmath55 . moreover they lead to a definition of the typical slope of mounds as @xmath56 with @xmath57 . we start our reasoning with the time dependence of the width @xmath58 clearly @xmath59 . since we expect the width to increase with time , we obtain the inequalities @xmath60 and @xmath61 the first conclusion can be drawn even without the scaling assumption : for model ( ii ) and model ( i ) with @xmath27 , @xmath62 has an upper bound , and so has @xmath63 . therefore @xmath64 . we conclude that the increase of the width @xmath65 can not be faster than @xmath66 if it is caused by a destabilizing nonequilibrium current on a surface with step edge barriers . assuming scaling we estimate @xmath67 and @xmath68 . for model ( i ) we further have @xmath69 . in terms of the scaling exponents @xmath70 and @xmath71 inequality ( [ ineqa ] ) yields @xmath72 , while the second inequality ( [ ineqb ] ) leads to @xmath73 . combining both inequalities we have @xmath74 to proceed we note that an upper bound on the lateral mound size @xmath3 can be obtained from the requirement that the mounds should be stable against the bales - zangwill step meandering instability @xcite : otherwise they would break up into smaller mounds . > from ( [ linstab ] ) it is easy to see that , for the large slopes of interest here , fluctuations of a wavelength exceeding @xmath75 are unstable . since @xmath76 , we impose the condition @xmath77 or , in terms of scaling exponents , @xmath78 hence the first relation in ( [ ineq1 ] ) becomes an equality ( which was previously derived for the one - dimensional case @xcite ) , and the second relation yields @xmath79\ ] ] for model ( ii ) we assume that the slope @xmath5 approaches its stable value @xmath80 as @xmath81 with @xmath82 . the estimate of the last term in ( [ widthdyn ] ) then becomes @xmath83 . thus inequality ( [ ineqa ] ) yields @xmath84 , and from ( [ ineqb ] ) it follows follows that @xmath85 . as for model ( i ) the next estimation uses @xmath86 with @xmath87 . again we obtain the inverse of the second of the above inequalities , _ viz . _ @xmath88 . altogether this yields @xmath89\ ] ] to summarize the general results obtained in this section : in addition to the bound on the temporal increase of the surface width , @xmath90 , the scaling ansatz yields an upper bound on the increase of the lateral length scale , @xmath91 , valid for both models . a more elaborate approximation , to be presented in the next section , predicts the above inequalitites ( [ scalineq],[scalineq2 ] ) to hold as equalities ( up to logarithmic corrections ) . we consider the time dependence of the equal time height - height correlation function defined above : @xmath92 where @xmath93 is the laplace operator . in order to obtain a closed equation for @xmath94 we replace @xmath20 by @xmath95 in the second term on the right hand side . this approach is inspired by the spherical ` large @xmath96 ' limit of phase ordering kinetics @xcite , and will be referred to as the spherical approximation . the argument of @xmath97 is then easily expressed in terms of @xmath53 : @xmath98 and the closure of ( [ corrdyn ] ) reads @xmath99 since we consider dynamics which are isotropic in substrate space , and also isotropic distributions of initial conditions , @xmath94 will only depend on @xmath100 and @xmath46 . consequently we consider the structure factor @xmath101 as a function of @xmath102 and @xmath46 , which satisfies @xmath103 s(k , t).\ ] ] here we have defined the function @xmath104 through @xmath105 and @xmath106 denotes the surface dimensionality ( @xmath107 for real surfaces ) . the formal solution of ( [ seq ] ) then reads @xmath108.\ ] ] the initial condition @xmath109 reflects the disorder in the initial configuration of eq . ( [ cont1 ] ) . it consists of fluctuations at early times , i.e. the first nucleated islands , from which mounds will later develop . simulations of microscopic models for mbe on singular surfaces at submonolayer coverages @xcite indicate the following shape of @xmath109 : from a hump at some finite wavenumber , corresponding to the typical distance @xmath110 between island nuclei , it falls off to zero for @xmath111 . for @xmath112 it goes down to a _ finite _ value @xmath113 . at late times the hump in @xmath101 persists , situated at some @xmath114 near the maximum of the exponential in ( [ formal ] ) . it belongs to a lateral lengthscale @xmath3 , denoting the typical distance of neighboring mounds . for late times @xmath114 will go to zero , so we need only consider @xmath109 near @xmath115 . in fact for the leading contribution to @xmath114 ( and the leading power in @xmath3 ) we only need @xmath116 . more detailed remarks on the case @xmath117 and on the presence of long range correlations in the initial stage can be found at the end of this section . the particular value of @xmath118 has no influence on the coarsening exponent , so we take @xmath119 which corresponds to @xmath120 . to follow the analysis , note that @xmath104 is a functional ( [ a(t ) ] ) of @xmath101 , and on the other hand it is used for the calculation of @xmath101 . this imposes a condition of self consistency on the solution , which we write as follows @xmath121 we used the initial conditions motivated above , and the shorthand @xmath122 . the integral can be evaluated , yielding for @xmath123 the differential equation @xmath124 where @xmath125 denotes a parabolic cylinder function @xcite . equation ( [ dgl ] ) can not be solved explicitly , but for the late time behavior we can use an asymptotic approximation for @xmath125 , since its argument @xmath126 for @xmath127 . to see this , note that ( [ dgl ] ) is of the form @xmath128 therefore , if @xmath129 remained bounded , for large @xmath46 the argument of @xmath97 in eq . ( [ dgl ] ) would be close to 0 , and ( [ dgl ] ) would approximately be @xmath130 . this is in contradiction to the assumption @xmath131 const.@xmath132 which therefore can not be true . for large @xmath46 ( and large @xmath129 ) we then approximate ( [ dgl ] ) by @xmath133 this shows that @xmath129 must grow more slowly than any power of @xmath46 : if @xmath134 then @xmath135 , whereas the argument in @xmath97 would increase exponentially , dominated by a term @xmath136 . for both choices ( [ model(i ) ] ) and ( [ model(ii ) ] ) of @xmath97 the right hand side of ( [ dgl ] ) would decrease much faster than the left or even become negative . depending on the choice of the current function @xmath97 we get different asymptotic behaviors in the leading logarithmic increase of @xmath137 . we first consider the case ( ii ) , where @xmath138 . here ( [ dgl ] ) reads @xmath139 where @xmath140 is a constant without any interest . none of the terms must increase with time as a power of @xmath46 . hence asymptotically the term in brackets must vanish , which requires that @xmath141 . the leading behavior of @xmath142 is therefore @xmath143 similarly we treat case ( i ) , using the asymptotic behavior @xmath144 for large @xmath145 . equation ( [ dgl ] ) then becomes @xmath146 again the powers of @xmath46 in the last term must cancel , yielding @xmath147 there is a noteworthy correspondence between models ( i ) and ( ii ) : the solution of ( ii ) is the limit @xmath148 of the solution of ( i ) . in this sense , a current which vanishes at a finite slope is equivalent to a positive shape function @xmath149 decreasing faster than any power of @xmath5 . the same correspondence applies also on the level of the inequalities derived in section [ scalingansatz ] , as can be seen by letting @xmath148 in ( [ scalineq ] ) and comparing to ( [ scalineq2 ] ) . the asymptotic form of @xmath123 gives us the following time dependence of the coarsening surface structure : inserting @xmath123 into the expression for the structure factor @xmath101 ( [ formal ] ) we obtain for each time @xmath46 a wavenumber @xmath150 which has the maximal contribution to @xmath101 . it can be interpreted as the inverse of a typical lateral lengthscale @xmath151 . up to a logarithmic factor , we obtain lateral coarsening with a power @xmath152 for both choices of @xmath149 . this corresponds to @xmath2 , which saturates the bound derived in section [ scalingansatz ] . it is however important to note that the resulting structure factor _ can not _ be written in a simple scaling form @xmath153 , as would be expected if @xmath154 were the only scale in the problem @xcite . rather , one obtains the _ multiscaling form _ @xcite @xmath155 where @xmath156 , and @xmath157 is a second lengthscale in the system . in contrast to @xmath154 , the exponent describing the temporal increase of @xmath158 _ does _ depend on the shape of the current function @xmath97 . next we discuss the behavior of the typical slope of the coarsening mounds , given by @xmath159 . this is obtained directly from ( [ selfc ] ) . for model ( ii ) @xmath104 approaches the stable value ( `` magic slope '' ) @xmath160 , with a leading correction @xmath161 note that the approach to the magic slope is very slow , a possible explanation for the common difficulty of deciding whether @xmath104 attains a final value or grows indefinitely in numerical simulations @xcite . we further remark that , up to a logarithmic factor , the inequality @xmath162 derived in ( [ scalineq2 ] ) for the exponent describing the approach to the magic slope becomes an equality within the spherical approximation . for model ( i ) the typical slopes diverge as @xmath163 consistent with the value @xmath164 derived as a bound in ( [ scalineq ] ) . in the limit @xmath148 the slope does not increase at all , which again is comparable to the presence of a stable slope . to close this section we briefly comment on the the shape of the structure factor @xmath101 and the correlation function @xmath94 obtained within the spherical approximation . assuming initial correlations as used above , @xmath116 , the structure factor is analytical at any time @xmath46 , as can be seen in equation ( [ formal ] ) . the corresponding correlation function therefore decays faster than any power of @xmath100 , modulated with oscillations of wavenumber @xmath165 . we can also predict the further evolution of long range correlations , assuming that they were initially present . a power law decay of @xmath166 corresponds to a singularity in @xmath109 . suppose the singularity is located at some point @xmath167 ( the power decay of @xmath53 is then modulated by oscillations ) . then the singularity will remain present in @xmath101 , but it will be suppressed as @xmath168 for late times . this implies that @xmath94 has a very weak power law tail for very large @xmath100 , but up to some @xmath169 ( which increases with time ) it decays faster than any power . however a singularity in @xmath109 will not be suppressed if it lies at the origin @xmath170 , since then in eq.([formal ] ) it is multiplied by unity . in real space , this implies that a power law decay of correlations without oscillations will remain present . even if @xmath109 is singular at @xmath171 , the scaling laws derived above remain valid . suppose for example that @xmath172 for @xmath112 . in transforming ( [ formal ] ) back to real space , such a power law singularity can be absorbed into the phase space factor @xmath173 involved in the @xmath174-integration . the result is simply a shift in the dimensionality , @xmath175 , which affects the prefactors of the scaling laws ( [ km ] ) , ( [ magic ] ) and ( [ steep ] ) for @xmath176 and @xmath104 but not the powers of @xmath177 . we have presented two approximate ways to predict the late stage of mound coarsening in homoepitaxial growth . to our knowledge this is the first theoretical calculation of coarsening exponents for this problem . although we have made heavy use of concepts developed in phase ordering kinetics @xcite , our results can not be directly inferred from the existing theories in that field . as was explained in detail by siegert @xcite , equation ( [ cont2 ] ) rewritten for the slope @xmath178 has the form of a relaxation dynamics driven by a generalized free energy , @xmath179 . phase ordering with a conserved vector order parameter @xmath30 is described by a similar form , @xmath180 , however it appears that the interchange of the order of the differential operators , from @xmath181 to @xmath182 , may lead to a qualitatively different behavior @xcite . nevertheless , the results obtained so far must be refined . ideally , one would like to derive _ equalities _ for the exponents using the scaling ansatz of section [ scalingansatz ] . more modestly , it would be desirable to extend the approach so that the effects of current functions without in - plane isotropy @xcite and of contributions proportional to @xmath16 @xcite on the scaling behavior can be assessed . the main drawback of the spherical approximation in section [ meanfield ] is that it does not predict conventional scaling . the experience from phase ordering kinetics in the @xmath183 model suggests that the multiscaling behavior obtained above may be an artefact of the spherical approximation @xcite . to address this issue , we have carried out a numerical integration of equation ( [ cont2 ] ) , with weak uncorrelated noise as initial condition . the results indicate conventional scaling behavior in the late stage of growth , with exponents @xmath2 , @xmath164 , which saturate the bounds of section [ scalingansatz ] . figure [ fig : scalplot ] shows a scaling plot of @xmath184 of model ( i ) with @xmath25 for times @xmath185 obtained from the numerical integration of ( [ cont2 ] ) . it is compared at times @xmath186 to the spherical approximation . the first zero and the width at @xmath187 are rescaled to @xmath188 . initial conditions of the approximation were chosen to coincide with the full dynamics at @xmath189 . the spherical approximation of @xmath53 takes a slightly different shape its oscillations are more pronounced for larger @xmath100 . we obtained a similar scaling plot for @xmath190 and for model ( ii ) . the inset shows the evolution of the first zero of @xmath53 : in the full dynamics it is best approximated by a power law @xmath191 with @xmath192 for the full dynamics , where @xmath193 . the spherical approximation deviates from a power law . here for late times the best fit is @xmath194 with @xmath195 . the beginning of the time integration @xmath196 does not coincide with the exptrapolated zero of the power laws @xmath197 , because the mounds take a finite time to develop out of the initial growth instability . this introduces the additional fitting parameter @xmath197 . the steepening of the mounds ( not shown in the graph ) develops with the power @xmath198 . for @xmath199 in model ( i ) we obtain @xmath200 and @xmath201 . for model ( ii ) we refer to the intergrations of siegert @xcite , which indicate @xmath202 . note that the multiscaling behavior of @xmath53 in the spherical approximation is very weak , in the sense that the curves at different times do not differ in shape too much . a more sensitive test of the scaling behavior of equation ( [ cont1 ] ) , in order to pin down the difference to the spherical approximation , would be desirable and can be achieved by extracting the function @xmath203 ( see eq.([multi ] ) ) from the data of the numerical integration . conventional scaling yields @xmath204 . work in this direction is currently in progress @xcite . * we thank p. milauer , f. rojas inguez , a. j. bray and c. castellano for many helpful hints and discussions . this work was supported by dfg within sfb 237 _ unordnung und grosse fluktuationen_. 99 g. ehrlich and f.g . hudda , j. chem . phys . * 44 * , 1039 ( 1966 ) . schwoebel and e.j . shipsey , j. appl . , 3682 ( 1966 ) ; r.l . schwoebel , j. appl . phys . * 40 * , 614 ( 1969 ) . some selected experiments are h .- j . ernst , f. fabre , r. folkerts and j. lapujoulade , _ phys . _ 72:112 ( 1994 ) ; c. orme , m.d . johnson , k .- leung , b.g . orr , p. milauer and d. vvedensky , _ j. cryst . growth _ 150:128 ( 1995 ) ; j.e . van nostrand , s.j . chey , m .- a . hasan , d.g . cahill and j.e . greene , _ phys . _ 74:1127 ( 1995 ) ; k. thrmer , r. koch , m. weber and k.h . rieder , _ phys . _ 75:1767 ( 1995 ) ; as well as refs.@xcite . a review is given in j. krug , origins of scale invariance in growth processes , _ advances in physics _ ( in press ) . j. villain , j. de physique i * 1,1 ( 1991 ) . j. krug , m. plischke and m. siegert , phys . lett . 70 * , 3271 ( 1993 ) . johnson , c. orme , a.w . hunt , d. graff , j. sudijono , l.m . sander and b.g . orr , phys . lett . * 72 * , 116 ( 1994 ) . stroscio , d.t . pierce , m. stiles , a. zangwill and l.m . sander , phys . * 75 * , 4246 ( 1995 ) . m. siegert and m. plischke , phys . lett . * 73 * , 1517 ( 1994 ) . m. siegert , _ in : _ scale invariance , interfaces and non - equilibrium dynamics " , a.j . mckane , m. droz , j. vannimenus and d.e . wolf , eds . , plenum press , new york ( 1995 ) , p.165 . a.w . hunt , c. orme , d.r.m . williams , b.g . orr and l.m . sander , europhys . * 27 * , 611 ( 1994 ) . j. krug and m. schimschak , j. phys . i france * 5 * , 1065 ( 1995 ) . m. rost , j. krug and p. milauer , surface science , in press . p. politi and j. villain , phys . rev . * 54 * , 5114 ( 1996 ) . mullins , j. appl . phys . * 30 * , 77 ( 1959 ) . j. krug , on the shape of wedding cakes ( to appear in journal of statistical physics ) . amar and f. family , phys . b ( in press ) . g.s . bales and a. zangwill , phys . b * 41 * , 5500 ( 1990 ) . bray , theory of phase - ordering kinetics , advances in physics , * 43 * , 357 ( 1994 ) . h. kallabis and p. milauer , private communications . i.s . gradshteyn and i.m . ryzhik , tables of integrals , series and products , section * 3.462 * , academic press ( 1980 ) . a. coniglio and m. zannetti , europhys . lett . , 575 ( 1989 ) . p. milauer and d.d . vvedensky , phys . b * 52 * , 14263 ( 1995 ) . c. castellano , private communication	we study unstable epitaxy on singular surfaces using continuum equations with a prescribed slope - dependent surface current . we derive scaling relations for the late stage of growth , where power law coarsening of the mound morphology is observed . for the lateral size of mounds we obtain @xmath0 with @xmath1 . an analytic treatment within a self consistent mean field approximation predicts multiscaling of the height height correlation function , while the direct numerical solution of the continuum equation shows conventional scaling with @xmath2 , independent of the shape of the surface current .
au mic is a nearby ( @xmath4 pc , * ? ? ? * ) m1ve flare star @xcite with the young age of @xmath5 myr @xcite , a key epoch in the formation of planetary systems . the star is surrounded by a nearly edge - on circumstellar disk discovered in coronographic images of scattered starlight that extends to a radius of at least 210 au @xcite . since the small grains responsible for scattering should be removed rapidly by stellar pressure forces , the disk is thought to consist of the collisional debris of unseen planetesimals experiencing ongoing impacts ( for recent reviews of the debris disk phenomenon , see * ? ? ? * ; * ? ? ? because of its proximity , the au mic debris disk has become one of the best studied examples at optical and infrared wavelengths , including detailed imagery of both intensity and polarization from the _ hubble space telescope _ @xcite . many observational properties of the au mic debris disk bear striking similarities to the archetype debris disk surrounding @xmath2 pic , also viewed nearly edge - on and located in the same young moving group . in particular , the midplane optical surface brightness profiles of these disks are remarkably alike , with shallow inner slopes ( @xmath6 ) that steepen substantially ( @xmath7 ) in the outer regions , near 35 au and 100 au for au mic and @xmath2 pic , respectively @xcite . these similarities inspired the development of a unified framework for debris disks based on a localized belt of planetesimals , or `` birth ring '' , that produces dust in a collisional cascade @xcite . the smallest grains are blown out from the belt by radiation and winds from the star , larger grains are launched into eccentric orbits with the same periastron as the belt , and the largest grains , which are minimally affected by non - gravitational forces , do not travel far from the belt before being ground down . the grains are therefore segregated according to their size , which gives rise to the characteristic scattered light profile . for these dynamics to prevail , the disks must be substantially free of gas @xcite , a property confirmed by measurements at ultraviolet wavelengths @xcite . a `` birth ring '' , if present , would remain hidden in optical and infrared images dominated by small grains that populate an extended halo ( e.g. , * ? ? ? by contrast , observations at ( sub)millimeter wavelengths highlight thermal emission from the largest grains and hence the location of the dust - producing planetesimals @xcite . to date , the best case for a `` birth ring '' comes from millimeter observations of @xmath2 pic @xcite . while the optical disk of @xmath2 pic extends more than @xmath8 au from the star , the millimeter imaging reveals a much more compact belt of emission at @xmath9 au radius . this millimeter emission belt coincides with the change in the slope of the optical surface brightness , which in the models marks the outer extent of the colliding planetesimals . previous ( sub)millimeter - wave observations of au mic did not have sufficient angular resolution to reveal much about the emission morphology . a detection at 850 @xmath10 m using jcmt / scuba ( @xmath11 beam ) indicated a reservoir of cold dust with mass @xmath12 m@xmath13 , but did not resolve any structure @xcite . subsequent observations at 350 @xmath10 m using cso / sharc ii ( @xmath14 beam ) marginally resolved an orientation compatible with the scattered light , but were otherwise limited @xcite . in this _ letter _ , we present imaging observations of au mic at 1.3 millimeters using the submillimeter array ( sma ) that obtain @xmath0 resolution and provide evidence for a planetesimal belt . we observed au mic with the sma @xcite on mauna kea , hawaii at 1.3 millimeters wavelength using the compact and extended configurations of the array . table [ tab : obs ] provides basic information about these observations , including the observing dates , baseline lengths , and atmospheric opacities . the weather conditions were good for all of these observations and best for the two compact configuration tracks when only 6 of the 8 array antennas were available . the phase center was @xmath15 , @xmath16 ( j2000 ) , offset by @xmath17 from the location of the star . the @xmath18 ( fwhm ) field of view is set by the primary beam size of the 6 meter diameter array antennas . the total bandwidth available was 8 ghz derived from two sidebands spanning @xmath19 to @xmath20 ghz from the lo frequency . time dependent complex gains were calibrated using observations of two quasars , j2101 - 295 ( 3.9 degrees away ) and j1924 - 292 ( 17.4 degrees away ) , interleaved with observations of au mic in a 15 minute cycle . the passband shape for each track was calibrated using available bright sources , mainly j1924 - 292 . the absolute flux scale was set with an accuracy of @xmath21 using observations of callisto or ganymede in each track . all of the calibration steps were performed using the idl based mir software , and imaging and deconvolution were done with standard routines in the miriad package . we made a series of images with a wide range of visibility weighting schemes to explore compromises between higher angular resolution and better surface brightness sensitivity . lcccc array configuration & extended & compact & compact & compact + number of antennas & 8 & 6 & 6 & 7 + baseline lengths ( m ) & 10189 & 868 & 868 & 868 + lo frequency ( ghz ) & 235.6 & 225.4 & 225.4 & 235.6 + 225 ghz atm . opacity & 0.10 & 0.100.06 & 0.06 & 0.08 + figure [ fig : image ] shows a contour image of the 1.3 millimeter emission overlaid on a _ hubble space telescope_/acs coronographic image of optical scattered light ( f606w filter ) from @xcite . the synthesized beam size for this 1.3 millimeter image is @xmath22 ( @xmath23 au ) , position angle @xmath24 , obtained with natural weighting and an elliptical gaussian taper @xmath25 ( fwhm ) oriented east - west to make the beam shape more circular . the rms noise in this image is 0.40 mjy beam@xmath26 , and the individual peaks have a signal - to - noise @xmath27 . the star symbol marks the stellar position corrected for proper motion , offset by @xmath28 from the phase center . the image shows a resolved band of 1.3 millimeter emission that extends approximately symmetrically from the stellar position to the southeast and northwest . by fitting a line to the positions of the two peaks in the image , we estimate the position angle of the emission structure to be @xmath29 , in excellent agreement with the disk orientation inferred from scattered light observations @xcite . from its position angle and double peaked morphology , we identify the emission structure as a limb - brightened dust belt . note that thermal emission from the stellar photosphere contributes only @xmath30 mjy at this wavelength , and even the strongest synchrotron radio flares from stellar activity should remain well below the noise @xcite . the total flux density obtained by integrating over a box that surrounds the emission is @xmath31 mjy , where the uncertainty is estimated conservatively by evaluating boxes of the same size in emission - free regions of the image . no other significant features are detected in the field of view . an extrapolation of the measurements at shorter wavelengths indicates that missing flux due to the spatial filtering by the sma is not a significant issue for these observations . the spectral index between 350 @xmath10 m @xcite and 850 @xmath10 m @xcite is @xmath32 , which predicts @xmath33 mjy at 1.3 millimeters , in good agreement with the sma measurement . this shallow index is consistent , within the uncertainties , with values near 2.0 determined for a sample of debris disks by @xcite , notably for stars later than f - type where dust temperatures are so low that the rayleigh - jeans approximation in this part of the spectrum is invalid . we conclude that the compact structure detected by the sma accounts for all of the 1.3 millimeter emission in the au mic system . to characterize the millimeter emission structure , we use a simple parametric disk model , following closely the method used by @xcite for analyzing similar observations of the @xmath2 pic disk . we assume the emission arises from a geometrically thin , axisymmetric belt with a radial surface brightness profile given by @xmath34 . this functional form is intended to capture the essence of low optical depth emission from a disk with a spatially invariant dust emissivity , surface density profile @xmath35 , and temperature profile that falls off with radius as a power - law . we fix the power - law index @xmath36 to approximate radiative heating from the central star , and we consider two shapes for @xmath35 : ( 1 ) an annulus with power law slope , @xmath37 for @xmath38 , @xmath39 , and ( 2 ) a gaussian , @xmath40}$ ] . in each case , the model is described by three parameters : a center radius @xmath41 , width @xmath42 , and total flux density @xmath43 . we fix the disk inclination at @xmath44 and position angle at @xmath45 , as determined from scattered light data @xcite , and place the origin within @xmath46 of the nominal stellar position ( within the sma astrometric accuracy ) . note that the vertical thickness of the millimeter grain population is expected to be @xmath47 @xcite , negligible compared to the resolution of the data . the utility of this simple modeling approach is to provide basic constraints while avoiding the multitudinous assumptions about grain properties , dynamics , and radiative transfer required in more sophisticated treatments . to estimate the model parameters and their uncertainties , we calculate three - dimensional grids spanning appropriate ranges of the parameter values . for the gaussian model , the grid covers @xmath48 au and @xmath49 au in 1 au steps , and @xmath50 mjy in 0.2 mjy steps . for each position in a model grid , we compute a set of synthetic sma visibilities and compare directly to the data with a @xmath51 value ( the sum of real and imaginary components over all spatial frequencies ) that uses natural weights modified by the gaussian taper used for the image in figure [ fig : image ] . the rightmost panel of figure [ fig : model ] shows the @xmath51 surface in the @xmath52 plane for the gaussian model after marginalizing over @xmath53 . the best - fit model is marked by a cross , and the contours delineate ( marginalized ) @xmath54 intervals in @xmath55 . the best - fit center radius is @xmath56 au , with corresponding width @xmath57 au ( or fwhm = @xmath58 au ) and flux density @xmath59 mjy . the three lefthand panels of figure [ fig : model ] show the original 1.3 millimeter image from figure [ fig : image ] , the gaussian model image obtained using the best - fit parameter values and the same visibility weighting scheme , and the imaged residuals ( where data@xmath60model subtraction is conducted on the visibilities ) . this comparison shows clearly that the best - fit model reproduces the main features of the image and leaves no systematic residuals . the power - law models for @xmath35 give similar best - fit parameter values : for @xmath61 , we find @xmath62 au and @xmath63 au ( where the upper limit is set by the computational grid boundary ) . for @xmath64 , the best - fit locus is shifted to roughly @xmath65 larger values of @xmath41 , a manifestation of the degeneracy between the radial gradient and emission extent . all of these best - fit models reproduce the observed sma visibilities equally well ( i.e. , have effectively the same minimum @xmath51 values ) . the modest signal - to - noise of the observations preclude placing tighter constraints on the model parameters . nonetheless , it is reassuring that consistent results are obtained from the various assumptions about the emission morphology . each gives a similar center radius , as well as an effectively cleared central cavity with size compatible with that inferred from scattered light ( @xmath66 au * ? ? ? * ) and the lack of mid - infrared emission ( @xmath67 au * ? ? ? the best - fit gaussian model reproduces better the contrast of the maxima to the extended emission of the disk in the image plane , but there is no clear preference for any of the functional forms . of course , none of these _ ad hoc _ models are perfect , and more sensitive data will be needed to constrain further the details of the emission distribution . we have spatially resolved the thermal emission at 1.3 millimeters from the au mic debris disk , thereby revealing the distribution of its millimeter grain population . these large grains are expected to have dynamics similar to the unseen dust - producing planetesimals , and the emission at this long wavelength provides a direct link to their spatial distribution . the overall structure of the au mic millimeter emission is reproduced well by a belt surrounding the star centered near a radius of @xmath68 au . the inferred location of this emission belt coincides with the region over which the scattered light brightness profile of the disk steepens markedly , the feature that provided impetus for developing the `` birth ring '' model . like au mic s more massive sister star @xmath2 pic @xcite , the resolved multi - wavelength data are consistent with a scenario where destructive collisions of planetesimals within the belt create grains with a spectrum of sizes , and the effects of size - dependent dust dynamics generate the millimeter emission and spatially extended scattered light nebula . for @xmath2 pic , the intense stellar radiation pressure can account for the large halo of small grains in the outer disk . but for au mic , which is about two orders of magnitude less luminous , the much weaker stellar radiation field is apparently augmented by a stellar wind to expel small grains @xcite . though the fractional radial extent of the au mic millimeter emission is not strongly constrained , the best - fit models suggest that it could be broad , @xmath69 . for @xmath2 pic , the millimeter emission is characterized by @xmath70 , with the upper end favored for the width of the underlying dust - producing planetesimal belt from detailed models of the knee of the scattered light profile @xcite . such broad belts are not unusual features of debris disks , in particular among the few resolved at millimeter wavelengths @xcite . an inner cavity devoid of dust is plausibly maintained by the combination of grinding collisions in the belt and clearing by stellar radiation forces , though a tenable alternative is that the inner regions are swept clear by a planet . planets seem to be the most viable explanation for the millimeter cavities common to the gas - rich transition disks around younger , pre - main - sequence stars @xcite . while a planet has been imaged to orbit within the cavity in the disk around @xmath2 pic @xcite , direct evidence for planets in the au mic system remains elusive , both from high contrast imaging observations @xcite and from transit searches @xcite . the outer boundary of the emission belts may reflect the outer extent of successful planetesimal formation , or it may simply mark the edge of the region where the planetesimal stirring mechanism has been successful at initiating dust - producing collisions , as in the models of @xcite . these new millimeter observations of au mic show the global structure of large grains in its debris disk , but they do not have sufficient resolution and sensitivity to reveal any significant departures from axisymmetry or substructure that might result from gravitational interactions with unseen planets . the higher resolution images of scattered light clearly show several radial and vertical inhomogeneities at subarcsecond scales ( features a - e , see * ? ? ? * ; * ? ? ? * ; * ? ? ? if these substructures have counterparts at millimeter wavelengths , then future observations with the atacama large millimeter / submillimeter array ( alma ) may be able to detect them , pinpoint the location of parent bodies , and constrain their origins .	we present imaging observations at 1.3 millimeters of the debris disk surrounding the nearby m - type flare star au mic with beam size @xmath0 ( 30 au ) from the submillimeter array . these data reveal a belt of thermal dust emission surrounding the star with the same edge - on geometry as the more extended scattered light disk detected at optical wavelengths . simple modeling indicates a central radius of @xmath1 au for the emission belt . this location is consistent with the reservoir of planetesimals previously invoked to explain the shape of the scattered light surface brightness profile through size - dependent dust dynamics . the identification of this belt further strengthens the kinship between the debris disks around au mic and its more massive sister star @xmath2 pic , members of the same @xmath3 myr - old moving group .
the cooling of compact stars has been discussed mainly in the context of neutron stars for decades @xcite . it has been believed that some stars require exotic cooling to explain the observed effective temperature and others can be explained by the modified urca and bremsstrahlung processes , where the central density of the star determines which cooling process works : an exotic cooling phase appears at higher density above a threshold density ( e.g. , * ? ? ? * ; * ? ? ? as a consequence , the heavier star which has higher central density cools faster than lighter one @xcite . however , as described below this scenario becomes inconsistent when we consider the recent observation of the effective temperature of cas a whose mass has been found to be unexpectedly large . cas a is the youngest - known supernova remnant in the milky way and it is located @xmath1 from the solar system @xcite . the supernova explosion occurred about 330 years ago , but due to absorption by the interstellar medium , there are no exact historical records except for an unclear detection by j. flamsteed in 1680 @xcite . recently , @xcite and @xcite have analyzed the x - ray spectra of cas a. they give the effective temperature and possible regions occupied by mass and radius relations . since cas a is the isolated remnant , the uncertainty of mass radius relation could be large . the lowest mass obtained from the @xmath2 fitting is about @xmath3 . considering the age of @xmath4 yr , @xmath5 of cas a must occupy a point of a cooling curve due to the scenario with modified urca and bremsstrahlung processes included on the @xmath6 plane . this gives strong constraint on the equation of state ( eos ) and cooling processes . furthermore , @xcite reported the observation of @xmath7 for cas a in the past 10 years . @xcite , @xcite , and @xcite insist that the rapid decrease in @xmath7 over time shows that the transition to nucleon superfluidity occurs . on the other hand , there are some cooled stars whose effective temperature can not be explained by the neutrino emission processes without nucleon superfluidity , including the modified urca and bremsstrahlung . it needs stronger cooling process as in the case of j0205 + 6449 in 3c58 ( hereafter `` 3c58 '' ) or vela pulsar ( b083345 ) . also an accreting neutron star sax j1808 requires strong cooling . 3c58 and vela may be explained by the minimal cooling model which includes nucleon superfluidity @xcite . however , sax j1808 needs stronger cooling than the minimal cooling @xcite . if we consider the strong cooling process according to the conventional scenario , their masses should become larger than that of cas a , which may be inconsistent with the mass observations of double neutron stars ; the mass of each neutron star is nearly @xmath8 ( e.g. , * ? ? ? * ) . isolated stars should have smaller ( similar ) masses compared with the case of ns - wd ( ns - ns ) binaries , respectively . the long - standing accretion from companions make the primaries heavier in the case of ns - wd binary systems @xcite . although a single eos must be applied to all the compact stars , the existing phase of matter depends on the density . therefore , the location of the cas a observation on the @xmath6 plane becomes very difficult to interpret if we believe the models with strong cooling mechanisms explain all other the observations of @xmath5 . in this paper , we present models that satisfy both cases of cas a and other cooled stars such as 3c58 and/or vela , by considering hybrid stars composed of quark matter , hadron matter , and their mixed phase ( mp ) , where a characteristic property of color superconducting ( csc ) phase is utilized . in addition , we also show cooling curves of cas a for observations over the past 10 years and indicate that the phase transition to the superfluidity is consistent with the observations . we construct a model that includes both quark hadron mp and its csc phase . considering the first - order phase transition between hadron and quark phases , it would be plausible that both phases coexist and form some kind of mp . similar to the `` nuclear pasta '' phase in the crust of a neutron star @xcite , it has been shown that mp could form geometrical structures @xcite ; @xcite have made eos of an mp under a wigner - seitz ( hereafter `` ws '' ) approximation using an mit bag model for a quark phase in finite temperature . in the present study , we employ an eos with the same framework using the bag constant @xmath9 , the coupling constant @xmath10 , and the surface tension parameter @xmath11 . for hadron phase , we adopt the results of the bruekner hartree fock ( bhf ) theory including hyperons , @xmath12 , and @xmath13 @xcite . however , the hyperons do not appear for the eos calculation including geometrically structured mp @xcite ; therefore , we do not include the effects of hyperons . although this does not occur in our model , if hyperons appear in other models , the hyperon mixed matter has large neutrino emissivity called hyperon direct urca process ( e.g. , @xmath14 ) , and causes the rapid cooling of compact stars @xcite . since the bhf results are inappropriate for low - density matter in the crust , we apply eos of bps @xcite for the crust . the eos gives a maximum mass of @xmath15 with a radius of @xmath16 , and the mass lies within the limits of the observation of cas a. although our eos is inconsistent with the recent observation of the mass @xmath17 of pulsar j16142230 @xcite , we could overcome this issue by adopting other eos models ( e.g. , * ? ? ? * ) . using the ws approximation , we obtain a cell radius of each phase and calculate the volume fraction of quark matter in mp as seen in figure [ fig : mp ] . it is difficult to calculate the neutrino emissivity in mp . therefore , the volume fraction @xmath18 is multiplied by the original quark neutrino emissivity @xmath19 @xcite ; the total emissivity by quarks is set to be @xmath20 . we adopt well - known neutrino emission processes without nucleon superfluidity for hadronic matter @xcite : modified urca process for the higher density region and bremsstrahlung process for the crust . we note that the special case of cas a during the past 10 years is discussed in section 3 . and a coupling constant @xmath21 @xcite . ] the color superconductivity is the key of the present study . there are different kinds of quark pairings such as cfl ( color flavor locking ) or 2sc ( two - flavor color superconducting ) according to the degrees of freedom of quark flavor and color . it is considered that the energy gap of @xmath22 is very large compared with the temperature of the center of compact stars , @xmath23 ( e.g. , * ? ? ? once matter becomes superconducting , neutrino emissivity must be suppressed due to the large energy gap and it could be proportional to @xmath24 , where @xmath25 is the temperature at the relevant layer and @xmath26 is the boltzmann constant @xcite . therefore , in the csc phase with a large energy gap ( @xmath27 ) , neutrino emissivity by quarks is almost negligible @xcite . we note that if we adopt a large energy gap we do not need to consider which kinds of csc pairing appear . in particular , the difference between cfl and 2sc is unimportant for the cooling . in fact , the pseudo ng bosons may exist and decay in the cfl phase and make changes to neutrino emissivity @xcite . however , it is known that it is not quantitatively affected in low temperature region @xcite . we assume that the energy gap in the csc phase is very large and the phase appears in the density region above the threshold density @xmath28 . for simplicity , we use the critical volume fraction @xmath29 instead of @xmath28 . if the matter has a volume fraction @xmath30 at the density in the layer of mp , quarks change into a csc state . as a consequence , the neutrino emission process due to the quark @xmath31 decay works only in the lower density region of the quark hadron mp . , @xmath32 , and @xmath33 , respectively . thick gray line on the middle panel denotes a @xmath34 model with nucleon superfluidity and carbon envelope . dot - dashed lines with marks on the middle panel indicate the model of the mass @xmath35 except for the neutrino emissivity in normal quark phase multiplied by one - tenth and one - one - hundredth for the lines with marks of triangle and circle , respectively . ] we select stellar masses of compact stars to be @xmath36 , @xmath37 , and @xmath38 whose central densities are 1.48 , 1.82 , and 2.67 @xmath39 , respectively . we take the critical fraction @xmath29 to be 0.1 , 0.125 , and 0.2 which are appropriate to explain the observations with our scenario . the results are shown in figure [ fig : cc ] , with available observational values . most of the observational data are taken from the table 7.2 in @xcite : @xcite for 3c58 ; @xcite for crab ; @xcite for vela ; @xcite for 0656 + 14 ; @xcite for 170644 ; @xcite for 18111925 ; @xcite for 182313 ; @xcite for 2334 + 61 . other data are taken from @xcite . the data of cas a are attached @xcite as the youngest compact star . we find that the cooling curves transit from hotter regions to cooler regions for the parameter between @xmath40 . as indicated in the middle panel of figure [ fig : cc ] , the cooling curves split into two regions for @xmath41 and stars with larger masses cool more slowly than those with lighter masses . since lighter mass stars cool faster , they are suitable for the 3c58 case which does not have lower limit of the effective temperature observation . however , the calculated cooling curves are inconsistent with the vela case which has lower limit . also , the quark cooling is still too strong to explain this case . since the neutrino emissivity of a quark phase involves large uncertainty , we have calculated the additional cooling curves for the mass @xmath35 in the case of @xmath42 with the neutrino emissivity reduced by a factor of @xmath43 and @xmath44 . there are some possible factors of this reduction for neutrino emissivity accompanying quark @xmath31-decay , such as an increase of the abundance of strange quarks ; a decrease in electron numbers inside mp leads to a reduction of neutrino emissivity @xcite . the presence of 2sc at low density also suppresses the emissivity ; @xcite discussed that the abundance of quarks in mp changes and may cause csc phase . we suppose that the reduction of emissivity originates from the above physical processes . if the emissivity of quarks is reduced by these factors , the observation of vela can be explained as shown in the middle panel of figure [ fig : cc ] . we demonstrate the effect of color superconductivity in quark hadron mp on the cooling curve : the larger the masses of the compact stars , the slower the speed of the cooling . this situation is caused by the layer that emits a large number of neutrinos through quark @xmath31 decay processes , which encircles the center of the star . the thickness of this layer decreases as stellar mass increases . although the maximum mass of our model is at most @xmath45 that would be near the lower limit for the recent observation of cas a , our cooling scenario can be applied for the cases of @xmath46 if more concrete eos are devised . some problems also remain concerning the fundamental physics : uncertain physical properties of quark hadron mp , indefinite tuning of the threshold density of csc , unknown values of the energy gap @xmath47 . nonetheless , to explain the observed effective temperature of cas a , our model incorporating the color superconductivity with a large energy gap is compatible with available observational data . the cooling mechanism associated with csc quarks could be plausible because , as shown by the existence of csc , it may be quite natural from the recent study of phase diagram between quarks and hadrons @xcite . the fundamental physics associated with compact stars is still largely uncertain . in the quark phase , the abundance of each quark is still unknown , and therefore the neutrino emissivity is not determined from the fractions of @xmath48 , @xmath49 , and @xmath50 quarks and/or the chemical potential of electrons @xcite . although the existence of the csc phase has been well studied , it is still open to debate which type of csc appears , and the quantitative values of the critical density and the energy gap should be clarified . since the physical properties of quarks in the mp are different from those in the uniform phase , there would be many factors to change the emissivity . further theoretical study in collaboration with the observations is required to constrain the physics of compact stars . considering the above uncertainties , the observations of the rapid cooling for cas a @xcite would give insight for constraining some properties of high density matter . we have tried to fit the observational data using a model with nucleon superfluidity . we adopt the neutrino emissivity accompanied by the phase transition from the normal state to that of the nucleon superfluidity ( @xcite ) , where we tune the critical temperature of the neutron @xmath51 superfluidity and the associated neutrino emissivity . for simplicity , we adopted only the superfluidity effect of the neutron @xmath51 , not of the neutron / proton @xmath52 . the singlets affect the cooling of a compact star , but the neutron singlet works in a lower density than the triplet , and the proton singlet is still speculative . the most effective nucleon superfluidity is caused by the neutron triplet state @xcite . since @xcite concluded that in order to reproduce the observations in the x - ray spectrum the surface composition of cas a must be carbon and/or helium , we set the surface composition to be of carbon and a small amount of helium . this is because the existence of carbon results in a rather high effective temperature at the beginning of cooling phase . to calculate the cooling curves , we assume the functional form of the critical temperature which is a phenomenological extension of that for the @xmath51 neutron superfluidity as seen in the left panel of figure [ fig : nsccc ] . this approach is similar to the method used by @xcite except for the profile of the critical temperature . considering that the cooling curve sensitively depends on the critical temperature ( left panel in figure [ fig : nsccc ] ) and the neutrino emissivity , we have fine tuned the two quantities to fit the observational data(@xmath7 ) of cas a over the past 10 years ( right panel in figure [ fig : nsccc ] ) . superfluidity used by this study ( solid and thick line ) and previous works . dot - dashed ( labeled `` 1 '' ) , long dashed ( labeled `` 2 '' ) , and short dashed ( labeled `` 3 '' ) denote the critical temperature adopted by @xcite , @xcite , and @xcite , respectively . right panel : cooling curve with the nucleon superfluidity which crosses the observational data of @xmath7 in cas a @xcite.,title="fig : " ] superfluidity used by this study ( solid and thick line ) and previous works . dot - dashed ( labeled `` 1 '' ) , long dashed ( labeled `` 2 '' ) , and short dashed ( labeled `` 3 '' ) denote the critical temperature adopted by @xcite , @xcite , and @xcite , respectively . right panel : cooling curve with the nucleon superfluidity which crosses the observational data of @xmath7 in cas a @xcite.,title="fig : " ] there are some experimental projects of hadron colliders with intermediate energy , such as j - parc or gsi , that may help us to understand the state of ultrahigh density . they are very useful for examining high density character composed of hyperons , mesons , transition to normal quark matter and csc phase @xcite . however , it is still difficult to reproduce the same phase as in the core of compact stars by the colliders , where they produce high temperature ( @xmath53 mev ) in the high density region of @xmath54 compared with the core of compact stars @xcite . therefore , it is worthwhile to check the theories by both observing as many compact stars as possible and comparing theoretical predictions with observations such as the effective temperature . from this viewpoint , further observations of cas a and central sources of other supernova remnants such as sn1987a are necessary to understand the fundamental physics in these extreme conditions . there are some recent studies of hybrid star cooling , such as @xcite , @xcite or @xcite . @xcite employed a similar model to our model , but the assumed energy gap @xmath47 is in the range of @xmath43@xmath55 , and the pair breaking and formation ( pbf ) process is not included ; thus resulting rapid decrease in the effective temperature of cas a is not compatible with their model . @xcite considered the quark hadron mixture and direct urca process in hadronic phase , but did not included quark csc , and resulting the cooling with the direct urca is too strong to explain the observational data of older compact stars . @xcite adopted the rotation of compact stars which delays the isothermal relaxation , and this effect would help us to understand the temperature drop of cas a. these effects should be included in our further study . even if the quark hadron mp does not exist , our scenario of color superconducting core surrounded by exotic phase could be applied to a cooled object such as vela . considering a meson condensed phase , a nucleon superfluidity , which reduces the strong neutrino emission by mesons , is expected to explain the vela data consistent with cooling curves with some mass ranges . the cooling scenario of compact stars not only affects the cooling of isolated stars , but also the stars in binary systems . there are some observational data of x - ray transients ( e.g. , * ? ? ? * ) that have gravitational energy supply on the surface due to accretion from companions , and x - ray bursts could result from thermonuclear reactions on the surface . these systems are worthwhile to re - examine from the point of view of exotic cooling as was partly done in @xcite for a quiescence period . since binary systems have larger luminosity and/or orbital information such as inclination angle , rotational period , and a signal of the gravitational wave , more accurate mass detection than isolated compact stars would be possible ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? this work has been supported in part by the grant - in - aid for scientific research ( 19104006 , 21105512 , 21540272 , and 24540278 ) and 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 . alford , m. g. , braby , m. , paris , m. & reddy , s. 2005 , , 629 , 969 alford , m. g. , schmitt , a. , rajagopal , k. & schfer , t. 2008 , rev . phys . , 80 , 1455 andronic , a. , blaschke , d. , braun - munzinger , p. , et al . 2010 , , 837 , 65 andronic , a. , braun - munzinger , p. & stachel , j. 2009 , phys . b , 673 , 142 ashworth , w. b. 1980 , j. hist . , 11 , 1 baldo , m. `` nuclear methods and the nuclear equation of state '' , 1999 , world scientific baldo , m. , burgio , g. f. & schulze , h. -j . 1998 , , 58 , 3688 baym , g. , pethick , c. j. & sutherland , p. 1971 , , 170 , 299 becker , w. ed . , `` neutron stars and pulsars '' , 2009 , springer becker , w. , brazier , k. t. s. & truemper , j. 1996 , , 306 , 464 bogomazov , a. i. , abubekerov , m. k. , lipunov , v. m. & cherepashchuk , a. m. 2005 , astron . rep . , 49 , 295 brisken , w. f. , thorsett , s. e. , golden , a. & goss , w. m. 2003 , , 593 , l89 demorest , p. b. , pennucci , t. , ransom , s. m. , roberts , m. s. e. & hessels , j. w. t. 2010 , , 467 , 1081 friman , b. l. & maxwell , o. v. 1979 , , 232 , 541 gaensler , b. m. , schulz , n. s. , kaspi , v. m. , pivovaroff , m. j. & becker , w. e. 2003 , , 588 , 441 gotthelf , e. v. , halpern , j. p. & dodson , r. 2002 , , 567 , l125 gusakov , m. e. , kaminker , a. d. , yakovlev , d. g. & gnedin , o. y. 2005 , , 363 , 555 hashimoto , m. , seki , h. & yamada , m. 1984 , prog . , 71 , 320 heinke , c. o. & ho , w. c. g. 2010 , , 719 , l167 heinke , c. o. , jonker , p. g. , wijnands , r. , deloye , c. j. & taam , r. e. 2009 , , 691,1035 ho , w. c. g. & heinke , c. o. 2009 , , 462 , 71 iwamoto , n. 1980 , , 44 , 1637 jaikumar , p. , prakash , m. & schfer , t. 2002 , , 66 , 063003 kaminker , a. d. , gusakov , m. e. , yakovlev , d. g. & gnedin , o. y. 2006 , , 365 , 1300 kaspi , v. m. , roberts , m. s. e. & harding , a. k. 2006 , in compact stellar x - ray sources , eds . w. h. g. lewin & m. van der klis ( cambridge university press ) , 279 - 340 kramer , m. , stairs , i. h. , manchester , r. n. , et al . 2006 , science , 314 , 97 lattimer , j. m. 2010 , prog . phys . suppl . , 186 , 1 lattimer , j. m. , pethick , c. j. , prakash , m. & haensel , p. 1991 , , 66 , 2701 maruyama , t. , chiba , s. , schulze , h. -j . & tatsumi , t. 2007 , , 76 , 123015 maruyama , t. , chiba , s. , schulze , h. -j . & tatsumi , t. 2008 , phys . b , 659 , 192 negreiros , r. , dexheimer , v. a. & schramm , s. 2012 , , 85 , 035805 page , d. 2002 , http://www.astroscu.unam.mx/neutrones/ns-data/data_good.dat page , d. , lattimer , j. m. , prakash , m. & steiner , a. w. 2009 , , 707 , 1131 page , d. , prakash , m. , lattimer , j. m. & steiner , a. w. 2011a , , 106 , 081101 page , d. , prakash , m. , lattimer , j. m. & steiner , a. w. 2011b , pos(xxxiv bwnp)005 , ( arxiv:1110.5116 ) pavlov , g. g. , zavlin , v. e. , sanwal , d. , burwitz , v. & garmire , g. p. 2001 , , 552 , l129 ravenhall , d. g. , pethick , c. j. & wilson , j. r. 1983 , , 50 , 2066 reed , j. e. , hester , j. j. , fabian , a. c. & winkler , p. f. 1995 , , 440 , 706 rster , s. b. , werth , v. , buballa , m. , shvkovy , i. a. & rischke , d. h. 2005 , , 72 , 034004 rutledge , r. e. , bildsten , l. , brown , e. f. , pavlov , g. g. & zavlin , v. e. 2002 , , 577 , 346 schmitt , a. `` dense matter in compact stars '' , 2010 , springer schramm , s. , dexheimer , v. , negreiros , r. , schrhoff , t. & steinheimer , j. 2012 , arxiv:1202.5113 schulze , h. -j . , lejeune , a. , cugnon , j. & lombardo , u. 1995 , phys . b , 355 , 21 shternin , p. s. , yakovlev , d. g. , heinke , c. o. , ho , w. c. g. & patnaude , d. j. 2011 , , 412 , l108 slane , p. o. , helfand , d. j. & murray , s. s. 2002 , , 571 , l45 takatsuka , t. & tamagaki , r. 1997 , prog . phys . , 97 , 345 tsuruta , s. 1998 , , 292 , 1 weisberg , j. m. , nice , d. j. & taylor , j. h. 2010 , , 722 , 1030 weisskopf , m. c. , odell , s. l. , paerels , f. , et al . 2004 , , 601 , 1050 yakovlev , d. g. , gnedin , o. y. , gusakov , m. e. , kaminker , a. d. , levenfish , k. p. & potekhin , a. y. 2005 , , 752 , 590 yakovlev , d. g. , gnedin , o. y. , kaminker , a. d. , levenfish , k. p. & potekhin , a. y. 2004 , adv . space res . , 33 , 523 yakovlev , d. g. , ho , w. c. g. , shternin , p. s. , heinke , c. o. & potekhin , a. y. 2011 , , 411 , 1977 yasutake , n. , maruyama , t. & tatsumi , t. 2009 , , 80 , 123009 yin , s. , van heugten , j. j. r. m. , diederix , j. , et al . 2011 , arxiv:1112.1880	we present a new scenario for the cooling of compact stars considering the central source of cassiopeia a ( cas a ) . the cas a observation shows that the central source is a compact star that has high effective temperature , and it is consistent with the cooling without exotic phases . the observation also gives the mass range of @xmath0 , which may conflict with the current plausible cooling scenario of compact stars . there are some cooled compact stars such as vela or 3c58 , which can be barely explained by the minimal cooling scenario , which includes the neutrino emission by nucleon superfluidity ( pbf ) . therefore , we invoke the exotic cooling processes , where a heavier star cools faster than lighter one . however , the scenario seems to be inconsistent with the observation of cas a. therefore , we present a new cooling scenario to explain the observation of cas a by constructing models that include a quark color superconducting ( csc ) phase with a large energy gap ; this phase appears at ultrahigh density region and reduces neutrino emissivity . in our model , a compact star has csc quark core with a low neutrino emissivity surrounded by high emissivity region made by normal quarks . we present cooling curves obtained from the evolutionary calculations of compact stars : while heavier stars cool slowly , and lighter ones indicate the opposite tendency without considering nucleon superfluidity . furthermore , we show that our scenario is consistent with the recent observations of the effective temperature of cas a during the last 10 years , including nucleon superfluidity .
structural multi - factor _ economic capital _ ( ec ) models derived from the creditmetrics framework @xcite have become the most widely adopted tools for risk quantification in credit portfolios . an outcome of these models , a portfolio ec and its allocation down to individual facilities , is used by financial institutions for any or all of the following : internal capital adequacy assessment , external reporting , risk - based pricing , performance management , acquisition / divestiture analyses , stress - testing and scenario analysis , etc . while in most cases monte carlo simulations are used due to limited analytical tractability of the multi - factor models , the recently reported advanced analytical techniques @xcite may be viewed as an alternative . unfortunately , neither the industry standard simulation - based approach nor the existing analytical techniques can fully address the needs of the financial institutions . in particular , the risk - based real - time pricing remains the ultimate challenge : none of the existing models is capable of providing sufficiently accurate , stable and time - efficient input . yet another practical aspect which has not received enough attention in the literature is the sometimes overly complex structure of the models as perceived by end users . very often the complexity of the models makes them hard to be understood and , hence , affects their acceptance within an organization . the approach presented here aims to overcome the above mentioned difficulties and is in its spirit similar to the one reported by @xcite . the content , however , is quite different since the presented model has more solid theoretical background , is easier to implement and use and is capable of covering fully featured multi - factor setup . the model described here was developed with kiss principle in mind . while based on the previous author s research on the analytical tractability of multi - factor models @xcite , the proposed model has very simple and intuitive structure . despite the simple structure , the model produces meaningful and reasonably accurate results and can be used by financial institutions for any of the purposes described above . in particular , the problem of capital allocation has a simple and time - efficient solution allowing real - time risk - based pricing . from conceptual point of view , one of the most attractive features of the model is its ability to quantify risk concentrations on both sector and obligor levels in a similar fashion . this article is organized as follows . a short description of structural multi - factor model and the necessary theoretical background are given in section [ sec : background ] . mathematical foundations of the proposed model are presented in section [ sec : kiss ] and are substantiated by benchmarking with monte carlo simulations in section [ sec : benchmarking ] . section [ sec : summary ] contains some concluding remarks and summarizes the presentation . let us consider a portfolio of credit risky facilities with loss functions @xmath0 at horizon ( one year ) being a function of random variables ( normalized asset returns ) @xmath1 . dependencies within the portfolio are modeled by means of a set of common factors @xmath2 : @xmath3 the random variables @xmath4 are independent and standard normally distributed . the instrument specific @xmath5 and @xmath6 define the systematic sensitivities of the instruments . the latter are subject to normalization condition @xmath7 . the idiosyncratic risk components are represented by @xmath8 . the economic capital of the portfolio is defined as na @xmath9 - quantile ( usually set to 99.9% or higher ) of the portfolio loss distribution @xmath10 relative to the expected loss of the portfolio : @xmath11 - \text{e}[l]\end{aligned}\ ] ] the above quantifies the overall portfolio risk which can be consistently distributed between the underlying facilities using the euler principle as ( * ? ? ? * see e.g. ) : @xmath12 where @xmath13 is a weight of the @xmath14th asset in the portfolio . to simplify the notations , these weights will not be written explicitly in what follows . no closed form analytical solution exists for either portfolio ec or its allocation @xmath15 in general case . however , in a single - factor case , i.e. one common factor @xmath16 and @xmath17 for any @xmath14 , the portfolio loss distribution @xmath18 quantile can be trivially found to be is an invertible function . ] @xmath19=l_{1f}(\eta_{1f}=n^{-1}(\alpha))$ ] , where @xmath20 is an inverse cumulative standard distribution function . this provides motivation to look for a solution of in a form of a sum of a single factor approximation and some corrections ( for detailed explanation of this approach see e.g. @xcite and references therein ) : @xmath21 = q_\alpha [ l_{1f } ] + \delta q_{\alpha}[\delta l_{mf } ] , \qquad l_{1f } = \text{e}[l\|\eta_{1f } ] , \quad \text{e}[l_{mf } ] = 0\end{aligned}\ ] ] here @xmath18 is conditional on the single factor value loss distribution of the portfolio . the single factor is constructed as a linear combination of the systematic factors @xmath22 with the normalization condition @xmath23 . obviously , the choice of @xmath16 significantly affects the quality of the approximation and will be given a particular attention in what follows . the conditional expectation series expansion technique @xcite can be applied to facilitate calculations of @xmath24 . translated to the notations introduced here this technique allows writing the conditional portfolio loss distribution @xmath18 as a sum of the conditional expectations of the constituents @xmath25 $ ] which can be expressed as : @xmath26 where the inner vector product @xmath27 stands for @xmath28 and @xmath29 are hermite polynomials . the series converge very well provided values of @xmath30 are not too close to 1 ( which is the case in practice ) and allow for very fast ( re)calculations of the conditional expectations once the constants @xmath31 have been computed . the above technique is particularly useful when considering arbitrary loss functions @xmath32 . as long as credit portfolio modeling is concerned , it became a common practice to distinguish the unsystematic @xmath2 and the idiosyncratic @xmath8 risk components . the usual assumption is that the former drive the portfolio risk dynamics while the latter only give minor contributions . however , the two sets of random variables do not differ from mathematical point of view . in fact , one can not draw a clear line between the systematic and idiosyncratic components using practical considerations either . indeed , imagine that the portfolio contains a single relatively big exposure or a set of exposures corresponding to the same borrower and , hence , sharing the same idiosyncratic random variable @xmath33 . depending on the size of this exposure(s ) , the risk brought to the portfolio by @xmath33 may be higher than the one originating from some or even all of the systematic factors @xmath2 . big enough exposure will eventually dominate the portfolio dynamics even if its sensitivity @xmath34 to the systematic factors is zero . introducing credit contagion effects by assigning more than one overlapping idiosyncratic factors to a group of dependent borrowers makes it even harder to make a distinction between the systematic and idiosyncratic factors . treating the systematic and the idiosyncratic risk equally not only simplifies the model structure , but also allows straightforward incorporation of the borrower concentration effects into the portfolio risk metrics . the notations used so far can be generalized as follows . for @xmath35 common factors @xmath2 and the portfolio consisting of @xmath36 borrowers let us introduce @xmath37 the single factor approximation @xmath18 of the portfolio loss can be written as @xmath38 unification of the systematic and idiosyncratic risk factors results in idiosyncratic factors being incorporated in @xmath39 which , as will be shown , defines the portfolio risk dynamics . thus , the idiosyncratic risk is accounted for in the same fashion as the systematic one . as was mentioned before , analytical tractability of the single factor case is the starting point for approaching the more general multi - factor setup . starting with a single factor approximation and calculating multi - factor ( including idiosyncratic ) adjustments as in , one can in principle calculate the portfolio economic capital . this approach , however , suffers from some difficulties . first , the choice of @xmath16 is not obvious , yet a very important first step . next , calculations of the multi - factor corrections may be quite laborious and hardware demanding . finally , as will be demonstrated later , the multi - factor corrections are not guaranteed to be convergent . instead of trying to overcome the difficulties associated with the multi - factor adjustments calculations , let us put all the efforts into constructing the single - factor approximation . some factor @xmath40 should exist which maximizes the relative contribution of the single - factor approximation in and , thus , diminishing the relative importance of the multi - factor corrections . assuming that the multi - factor corrections give positive contribution to the @xmath9-quantile of the portfolio loss distribution @xmath41 , the optimization problem reduces to maximization of the contribution from the single - factor approximation @xmath18 : @xmath42 \approx \max_{1f } q_\alpha [ l_{1f}]\end{aligned}\ ] ] the validity of this crucial assumption will be substantiated later in section [ sec : benchmarking ] by benchmarking with monte carlo simulations . from now on let us define the economic capital ec of the credit portfolio as an @xmath9-quantile of the optimal single factor distribution @xmath18 . using the notations introduced in this section the economic capital can be written as @xmath43 , \qquad r_i=\vec{\alpha}\cdot\vec{\rho_i}\end{aligned}\ ] ] the optimal single factor is defined by @xmath39 which maximizes the above expression @xmath44 and has the following solution @xmath45 this equation , however , contains @xmath39 on both sides ( @xmath46 on the right contains @xmath9 ) and does not allow a straightforward analytical solution . the problem can still be solved numerically by applying , for example , the method of steepest descent . based on , the following starting point can be suggested @xmath47 the calculations can be significantly facilitated by the series expansion . in practice , only few iterations are needed to have an accurate solution to the optimization problem . the calculations are not hardware demanding and very fast . the optimal single factor defined by leads to another simplification for the portfolio capital allocation problem . the individual capital contributions @xmath48\right)\end{aligned}\ ] ] can be written as @xmath49 + \sum_j\frac{\partial\overline{l}_i}{\partial r_i}\vec{\rho}_j\cdot w_i\frac{\partial}{\partial w_i}\frac{\vec{p}}{\\|\vec{p}\\| } = \overline{l}_i(\vec{\alpha}\vec{\rho}_i ) - \text{e}[l_i]\end{aligned}\ ] ] where the third term can be shown to be zero : @xmath50 in other words , the choice of @xmath39 according to leads to particularly simple expressions for capital contributions . the overall portfolio ec is a sum of conditional expectations of the excess losses @xmath51 $ ] which happen to coincide with the individual capital contributions . let us emphasize that once the optimal single factor has been computed , the set of parameters @xmath39 is sufficient to perform the capital allocation calculations . this allows real time calculations necessary for risk - based pricing . in fact , the proposed method for the economic capital calculations and its allocation is so efficient that the calculations can be performed on huge portfolios containing millions of facilities in minutes using entry level desktop computer . the accuracy of the method is demonstrated in the next section through benchmarking against monte carlo simulations . to highlight the points made and substantiate the assumptions used in the previous section , let us compare the performance ( accuracy ) of the proposed analytical approximation with monte carlo simulations . to demonstrate the advantages of the proposed technique , the comparison analysis will also cover the previously reported analytical technique @xcite which , in contrast with the one presented here , aims at precise calculations of the multi - factor corrections in . @xmath52{pd1conc } \includegraphics[width=0.4\textwidth , viewport=0 595 265 785,clip]{pd1highconc } \\ \includegraphics[width=0.4\textwidth , viewport=0 600 265 790,clip]{pd01conc } \includegraphics[width=0.4\textwidth , viewport=0 595 265 785,clip]{pd01highconc } \\ \includegraphics[width=0.8\textwidth , viewport=0 600 530 790,clip]{cap_vs_conf2 } \end{array}$ ] @xmath53{pd1conc } \includegraphics[width=0.4\textwidth , viewport=0 595 265 785,clip]{pd1highconc } \\ \includegraphics[width=0.4\textwidth , viewport=0 600 265 790,clip]{pd01conc } \includegraphics[width=0.4\textwidth , viewport=0 595 265 785,clip]{pd01highconc } \\ \includegraphics[width=0.8\textwidth , viewport=0 0 530 190,clip]{cap_vs_conf2 } \end{array}$ ] let us start the analysis by considering a simple portfolio of 1000 bullet loans maturing at the horizon . each loan has sensitivity @xmath54 to the single systematic factor @xmath55 and a unique idiosyncratic component @xmath33 . the loss at horizon functions are @xmath56 where @xmath57 are _ probabilities of default _ and are equal for all the loans . two sets of experiments were conducted with pd=1% and pd=0.1% . the loss severities @xmath58 are initially set equal to all loans in the portfolio . the concentration effects are studied by gradually increasing the loss severity of one of the loans and examining the impact on the portfolio ec . the confidence level for ec is set to 99.9% except for the last experiment where the dependency of the ec on the confidence level was investigated for fixed 10% concentration . both ec and concentration are measured as a fraction of the total portfolio exposure ( i.e. the sum of all the loss severities ) . the results of the monte carlo simulations are compared with the numbers produced using the kiss model ( section [ sec : kiss ] ) and with the output of the approach based on the calculations of the multi - factor corrections in . in the latter case the single systematic factor @xmath55 was used for the single - factor approximation . based on the results of the comparison presented in fig . [ fig : artificial ] , one can conclude the following . multi - factor corrections may lead to very accurate results in case of moderate concentrations ( pd=1% , concentration @xmath59 10% ) . in case of high concentrations , however , the results suggest that the higher order ( 3rd order in this case ) multi - factor corrections are divergent , while limiting the multi - factor contributions to the 2nd order corrections leads to significant underestimation of ec . the kiss model , on the other hand , despite being not extremely accurate , produces robust and reliable ec estimates for a wide range of parameters . this is especially obvious when studying the dependency of the portfolio ec on the confidence level . here the analysis of the previous section is complemented by the one conducted on more realistic portfolio . the portfolio consisted of 2,000 loans to distinct customers randomly selected from a loan portfolio of a large european bank . the set of common systematic factors covering 45 geographic regions and 61 regions , as well as the valuation function at horizon @xmath60 used in the experiment were similar to those of the portfoliomanager @xcite model . both the portfolio ec and its allocation @xmath15 were estimated using unbiased monte carlo simulations . the confidence level was set to 99.9% and the ec contributions were estimated as average realized values in the interval 99.85%-99.95% . a total of @xmath61 ( ten billion ) scenarios were used . the simulation - based estimates were compared to both the kiss model outcome as well as with the output of the approach based on the calculations of the multi - factor corrections in . in the latter case the single - factor @xmath40 used as a starting point was obtained using restricted to systematic components . second ( @xmath62 ) and third order ( @xmath63 ) corrections were calculated . the results on the portfolio level are summarized in table [ tab : portfolio ] and the facility - level results are presented as scatter plots in fig.[fig : realistic ] . c c c c 1f & kiss & 1f+mf2 & 1f+mf2+mf3 + + -4.3% & -2.0% & -0.5% & -0.1% + on the portfolio level the accuracy of the proposed approximation is more than sufficient for any practical purposes . the performance may seem to be less impressive on the facility level . as expected , the most significant mismatch is observed for the facilities with high concentrations of capital . these discrepancies , however , do not jeopardize the practical validity of the kiss model . indeed , the uncertainty in the input parameters observed in practice ( i.e. default rates , recoveries , correlation parameters , etc ) as well as dependencies on the modeling assumptions diminish to large extent the approximation errors of the proposed technique . this , combined with the robustness demonstrated in the previous section , makes the kiss approximation a valid alternative for quantification of risks in credit portfolios . @xmath64{single } & \includegraphics[width=0.45\textwidth , viewport=5 590 385 790,clip]{single_kiss } \\ \includegraphics[width=0.45\textwidth , viewport=5 590 385 790,clip]{second } & \includegraphics[width=0.45\textwidth , viewport=5 590 385 790,clip]{third } \end{array}$ ] @xmath65{single } & \includegraphics[width=0.45\textwidth , viewport=0 5 380 205,clip]{single_kiss } \\ \includegraphics[width=0.45\textwidth , viewport=0 5 380 205,clip]{second } & \includegraphics[width=0.45\textwidth , viewport=0 5 380 205,clip]{third } \end{array}$ ] despite its simplicity , the analytical approximation presented here is capable of quantifying credit portfolio risks in a general multi - factor setup . the _ var _ risk measure used here can easily be replaced with the _ expected shortfall_. the arbitrary loss functions @xmath66 used allow for covering not only default - only regime , but also mtm valuation or even the dependency of in - default loss severities on the systematic factors . the default - only case has a particularly simple solution mimicing the well - known irb capital rules : @xmath67 the less than perfect accuracy of the approximation is not crucial for day - to - day practical needs of credit portfolio managers . the advantages of the proposed technique are significant . the model allows very fast and straightforward calculations including real - time risk - based pricing . the simple , robust and transparent structure can facilitate user acceptance and integration on all levels of financial institutions . 99 j. cespedes , j. herrero , a. kreinin and d. rosen ( 2006 ) a simple multifactor `` factor adjustment '' for the treatment of credit capital diversification . _ journal of credit risk _ , vol.2 , no.3 , fall 2006 .	a simple , yet reasonably accurate , analytical technique is proposed for multi - factor structural credit portfolio models . the accuracy of the technique is demonstrated by benchmarking against monte carlo simulations . the approach presented here may be of high interest to practitioners looking for transparent , intuitive , easy to implement and high performance credit portfolio model .
for any missing notation or reference let us refer to @xcite . for a graph @xmath7 , let @xmath17 ( @xmath18 , respectively ) denote its vertex set ( edge set , respectively ) . for a subset @xmath19 , let @xmath20 @xmath21 is adjacent to some @xmath22 be the _ neighborhood of u in g _ , and @xmath23 be the _ anti - neighborhood of u in g_. if @xmath24 , then let us simply write @xmath25 instead of @xmath26 , and @xmath27 instead of @xmath28 . for @xmath19 let @xmath29 $ ] denote the subgraph of @xmath7 induced by @xmath30 . for a vertex @xmath31 and for a subset @xmath32 ( with @xmath33 ) , let us say that _ @xmath21 contacts @xmath30 _ if @xmath21 is adjacent to some vertex of @xmath30 , and _ @xmath21 dominates @xmath30 _ if @xmath21 is adjacent to each vertex of @xmath30 . a _ component of @xmath7 _ is the vertex set of a maximal connected subgraph of @xmath7 . independent set _ ( or a _ stable set _ ) of a graph @xmath7 is a subset of pairwise nonadjacent vertices of @xmath7 . an independent set of @xmath7 is _ maximal _ if it is not properly contained in any other independent set of @xmath7 . for a given graph @xmath34 , a graph @xmath7 is _ @xmath34-free _ if none of its induced subgraphs is isomorphic to @xmath34 ; in particular , @xmath34 is called a _ forbidden induced subgraph of @xmath7_. given two graphs @xmath7 and @xmath35 , @xmath36 denotes the disjoint union of @xmath7 and @xmath35 ; in particular , @xmath37 and in general , for @xmath4 , @xmath5 denotes the disjoint union of @xmath6 copies of @xmath7 . the following specific graphs are mentioned later . a _ chordless path _ @xmath38 has vertices @xmath39 and edges @xmath40 for @xmath41 . a _ chordless cycle _ @xmath42 , @xmath43 , has vertices @xmath39 and edges @xmath40 for @xmath41 and @xmath44 . a @xmath45 is a complete graph of @xmath46 vertices . a @xmath47 is a complete bipartite graph whose sides respectively have one vertex , called the _ center _ of @xmath47 , and @xmath46 vertices , called the _ leaves _ of @xmath47 ( if @xmath48 then there are two trivial centers ) . @xmath49 is also called _ claw_. a _ fork _ ( sometimes called _ chair _ ) has vertices @xmath50 , and edges @xmath51 ( thus , a fork contains a claw as an induced subgraph ) . an _ apple _ is formed by a @xmath42 , @xmath52 , plus one vertex adjacent to exactly one vertex of the @xmath42 . for indices @xmath53 , let @xmath54 denote the graph with vertices @xmath55 , @xmath56 , @xmath57 such that the subgraph induced by @xmath55 forms a @xmath58 @xmath59 , the subgraph induced by @xmath60 forms a @xmath61 @xmath62 , and the subgraph induced by @xmath63 forms a @xmath64 @xmath65 , and there are no other edges in @xmath54 . thus , _ claw _ is @xmath66 , and @xmath38 is isomorphic to e.g. @xmath67 . let @xmath7 be a given graph and let @xmath68 be a weight function on @xmath17 . for an independent set @xmath69 , its weight is @xmath70 . let @xmath71 independent in @xmath72 denote the maximum weight of any independent set of @xmath7 . the _ maximum weight independent set _ ( _ mwis _ ) problem asks for an independent set of @xmath7 of maximum weight . if all vertices @xmath21 have the same weight @xmath73 , @xmath74 and mwis is called the _ mis _ problem . mwis is np - hard @xcite and remains np - hard under various restrictions , such as for triangle - free graphs @xcite and more generally for graphs without chordless cycle of given length @xcite , for cubic graphs @xcite and more generally for @xmath75-regular graphs @xcite , and for planar graphs @xcite . it can be solved in polynomial time for various graph classes , such as for @xmath76-free graphs @xcite and more generally perfect graphs @xcite , for claw - free graphs @xcite and more generally fork - free graphs @xcite and apple - free graphs @xcite , for @xmath11-free graphs @xcite and more generally @xmath13-free graphs for any constant @xmath6 ( by combining an algorithm generating all maximal independent sets of a graph @xcite and a polynomial upper bound on the number of maximal independent sets in @xmath13-free graphs @xcite ) , @xmath77claw - free graphs @xcite , and @xmath14-free graphs @xcite . furthermore mwis can be solved in polynomial time for @xmath78-free graphs as recently proved in @xcite . the first two polynomial time algorithms for mwis on claw - free graphs were introduced in 1980 by minty @xcite and independently by sbihi @xcite , then revisited by nakamura and tamura @xcite , and recently improved by faenza , oriolo , and stauffer @xcite , and by nobili and sassano @xcite with the best known time bound in @xcite . [ theo : claw]@xcite for claw - free graphs , the mwis problem can be solved in time @xmath79 . @xmath80 obviously , for every graph @xmath7 the following holds : @xmath81 ) : v \in v\}\ ] ] thus , for any graph @xmath7 , mwis can be reduced to the same problem for the anti - neighborhoods of all vertices of @xmath7 . then we have : [ k1 ] for any graph @xmath35 , if m(w)is can be solved for @xmath35-free graphs in polynomial time then m(w)is can be solved for @xmath82-free graphs in polynomial time . @xmath80 let us report the following result due to alekseev @xcite . let us say that a graph is of _ type @xmath83 _ if it is graph @xmath54 for some indices @xmath84 . [ theo : alekseev]@xcite let @xmath85 be a class of graphs defined by a finite set @xmath86 of forbidden induced subgraphs . if @xmath86 does not contain any graph every connected component of which is of type @xmath83 , then the m(w)is problem is _ np_-hard for the class @xmath85 . alekseev s result implies that m(w)is is np - hard for @xmath87-free graphs @xmath88 the fact that m(w)is is np - hard for @xmath87-free graphs is already mentioned in @xcite . unless p = np , alekseev s result implies that for any graph @xmath35 , if m(w)is is polynomial time solvable for @xmath35-free graphs , then each connected component of @xmath35 is of type @xmath83 . by proposition [ k1 ] , for any graph @xmath35 , if m(w)is can be solved in polynomial time for @xmath35-free graphs then for any constant @xmath6 , m(w)is can be solved in polynomial time for @xmath89-free graphs . it follows that , since for any constant @xmath6 , m(w)is can be solved in polynomial time for @xmath13-free graphs @xcite , for fork - free graphs @xcite , for @xmath77claw - free graphs @xcite , for @xmath14-free graphs @xcite , and for @xmath78-free graphs @xcite , the minimal graphs @xmath35 of type @xmath83 for which the complexity of m(w)is for @xmath35-free graphs was open are : @xmath90 , @xmath91 , @xmath92 , @xmath93 , @xmath15 , @xmath16+claw , and thus , the minimal graph classes , defined by forbidding one induced subgraph , for which the complexity of m(w)is was open are : * @xmath90-free graphs , @xmath91-free graphs , @xmath92-free graphs , @xmath93-free graphs , @xmath15-free graphs , @xmath16+claw - free graphs . in this manuscript , we show that for any constant @xmath6 , mwis can be solved for @xmath6claw - free graphs in polynomial time . this extends the known results for mwis on claw - free graphs , @xmath13-free graphs for any constant @xmath6 , @xmath10+claw - free graphs , @xmath14-free graphs , and solves the open question for mwis on @xmath15-free graphs and on @xmath16+claw - free graphs . our approach is based on farber s approach showing that every @xmath11-free graph has @xmath12 maximal independent sets @xcite ( reported in section [ 2k2fr ] ) , which directly leads to a polynomial time algorithm to solve mwis for @xmath11-free graphs by dynamic programming . in this section let us refer to algorithm @xmath94 ( subsequently called algorithm alpha ) from @xcite which formalizes the aforementioned approach by farber @xcite ; our subsequent approach for mwis on @xmath6claw - free graphs is based on this algorithm . for a @xmath11-free input graph @xmath7 , algorithm alpha produces a family @xmath95 of independent sets of @xmath7 , which can be computed in time @xmath96 and which contains @xmath12 members such that each maximal independent set of @xmath7 is contained in some member of @xmath95 . for a graph @xmath97 with @xmath98 , a _ vertex ordering _ @xmath99 of @xmath7 is a total ordering of the vertex set @xmath100 of @xmath7 . for such a vertex ordering @xmath99 of @xmath7 , let @xmath101 $ ] denote the subgraph of @xmath7 induced by the first @xmath102 vertices , @xmath103 . given a vertex ordering @xmath99 , at each loop @xmath102 , @xmath104 , algorithm alpha provides a family @xmath105 of subsets of @xmath106 ( by modifying @xmath95 at loop @xmath102 by extending some of its members or by adding new members ) such that each maximal independent set of @xmath107 is contained in some member of @xmath105 , and finally returns the family @xmath108 . * algorithm alpha * + * input : * a @xmath11-free graph @xmath7 and a vertex ordering @xmath99 of @xmath7 . + * output : * a family @xmath95 of subsets of @xmath17 . + + @xmath109 ; + * for * @xmath110 * to * @xmath46 * do * + * begin * + 1 . [ extension of some members of @xmath95 ] + @xmath112 * do * + @xmath113 is an independent set * then * @xmath114 . [ addition of new members to @xmath95 ] + @xmath10 of @xmath107 containing @xmath115 ( i.e. , for each edge @xmath116 of @xmath107 ) * do * + @xmath117 ; + @xmath118 . + * end . * then the mwis problem can be solved for @xmath11-free graphs by the following algorithm . * algorithm @xmath11-free - mwis * + * input : * a @xmath11-free graph @xmath7 . + * output : * a maximum weight independent set of @xmath7 . * execute algorithm alpha for @xmath7 . let @xmath95 be the resulting family of subsets of @xmath17 . * for each @xmath112 , compute a maximum weight independent set of @xmath119 $ ] ( note that each @xmath112 is an independent set since @xmath7 is @xmath11-free ) . then choose a best solution , i.e. , one of maximum weight . then one obtains the following result . * @xcite * for @xmath11-free graphs , the mwis problem can be solved in time @xmath120 by algorithm @xmath11-free - mwis . @xmath80 first let us introduce a preparatory result . for each @xmath121 , let @xmath122 be the graph drawn in the subsequent figure . note that each @xmath122 contains an induced claw . for each @xmath121 , let @xmath123 denote the set of white vertices of @xmath122 , let @xmath124 denote the set of black vertices of @xmath122 , and let @xmath125 denote the ( white ) vertex at the top of @xmath122 . [ figlk ] for @xmath126,title="fig : " ] [ lemm : claw ] for a graph @xmath7 , assume that @xmath31 is a vertex such that @xmath21 is contained in an induced claw of @xmath7 and @xmath127 $ ] is claw - free . then for each maximal independent set @xmath69 of @xmath7 with @xmath128 , there is a @xmath121 such that @xmath129 for an induced subgraph @xmath122 of @xmath7 with @xmath21 = top@xmath130 . * let @xmath131 be a claw in @xmath7 with , say , @xmath132 . let @xmath69 be a maximal independent set of @xmath7 containing @xmath21 , and let @xmath133 . then for @xmath134 , @xmath135 is a maximal independent set of @xmath136 $ ] . let us distinguish between the following cases . * case 1 * @xmath119 $ ] is connected . by assumption , @xmath21 is contained in an induced claw of @xmath7 . let us distinguish between two subcases . * case 1.1 @xmath21 is the center of @xmath131. since @xmath119 $ ] is claw - free , each of @xmath137 has at most two neighbors in @xmath135 . case 1.1.1 * if a vertex of @xmath137 , say @xmath138 , has two neighbors in @xmath135 , say @xmath139 then @xmath140 with @xmath141 , @xmath142 , and @xmath21 = top@xmath143 . * case 1.1.2 * if none of @xmath137 has a neighbor in @xmath135 then @xmath144 with @xmath145 and @xmath21 = top@xmath146 . * case 1.1.3 * now assume that cases 1.1.1 and 1.1.2 are excluded . this means that one of @xmath137 , say without loss of generality @xmath138 , has exactly one neighbor in @xmath135 and @xmath147 and @xmath148 have at most one neighbor in @xmath135 . let @xmath149 for @xmath150 . note that not both of @xmath147 and @xmath148 are adjacent to @xmath151 since @xmath34 is claw - free , and in general , @xmath152 and @xmath148 do not have any common neighbor in @xmath135 . if @xmath153 then we have @xmath154 with @xmath155 and @xmath21 = top@xmath156 . if @xmath147 has exactly one neighbor in @xmath135 , say @xmath157 , and @xmath158 then if @xmath159 , we have @xmath160 with @xmath161 and @xmath21 = top@xmath162 , and if @xmath163 , we have @xmath164 with @xmath165 and @xmath21 = top@xmath166 , and similarly for the case when @xmath148 has exactly one neighbor in @xmath135 , and @xmath167 . finally , assume that both @xmath147 and @xmath148 have a neighbor in @xmath135 , i.e. , there are @xmath168 with @xmath169 and @xmath170 . if @xmath171 are pairwise distinct then we have @xmath172 with @xmath173 and @xmath21 = top@xmath174 . now assume that @xmath175 ( recall that @xmath176 is impossible ) . without loss of generality , let @xmath177 . then we have @xmath178 with @xmath179 and @xmath21 = top@xmath180 . * case 1.2 @xmath21 is a leaf of @xmath131. without loss of generality , let @xmath147 be the center of @xmath131 . since @xmath119 $ ] is claw - free , @xmath147 has at most two neighbors in @xmath135 , and if @xmath181 ( @xmath182 , respectively ) , the same holds for @xmath138 ( @xmath148 , respectively ) . the following subcases are exhaustive by symmetry . case 1.2.1 * if @xmath183 then @xmath140 with @xmath184 and @xmath21 = top@xmath143 . * case 1.2.2 * if exactly one of @xmath185 is in @xmath135 , say without loss of generality , @xmath186 and @xmath187 ( and more generally , only one of the neighbors of @xmath147 is in @xmath135 - otherwise we have case 1.2.1 ) then @xmath148 has a neighbor in @xmath135 , say @xmath188 , since @xmath135 is a maximal independent set of @xmath136 $ ] . then clearly , @xmath188 is nonadjacent to @xmath138 and @xmath21 and is nonadjacent to @xmath147 ( otherwise @xmath147 would have two neighbors in @xmath135 ) . then @xmath189 with @xmath190 and @xmath21 = top@xmath191 . * case 1.2.3 * now assume that cases 1.2.1 and 1.2.2 are excluded . thus , @xmath192 . then both @xmath138 and @xmath148 must have a neighbor in @xmath135 since @xmath135 is a maximal independent set of @xmath136 $ ] . if no neighbor of @xmath138 or @xmath148 in @xmath135 is adjacent to @xmath147 then both @xmath138 and @xmath148 have exactly one neighbor in @xmath135 , else a claw in @xmath119 $ ] would arise involving @xmath147 . let @xmath193 with @xmath149 , @xmath194 . if @xmath159 then @xmath195 with @xmath196 and @xmath21 = top@xmath197 . if @xmath177 then @xmath198 with @xmath199 and @xmath21 = top@xmath200 . now assume that , without loss of generality , a neighbor @xmath201 of @xmath138 is adjacent to @xmath147 . we claim : \(i ) @xmath188 is adjacent to @xmath148 , since otherwise case 1.2.2 holds with @xmath188 instead of @xmath138 ; \(ii ) @xmath138 and @xmath148 have at most one more neighbor in @xmath135 , and such a neighbor is non - adjacent to @xmath147 , since otherwise case 1.2.1 holds ( i.e. , @xmath147 has two neighbors in @xmath135 ) . if neither @xmath138 nor @xmath148 have another neighbor in @xmath135 then @xmath202 with @xmath203 and @xmath21 = top@xmath204 . if there is @xmath150 with @xmath205 , @xmath149 and the only neighbor of @xmath148 in @xmath135 is @xmath188 then @xmath206 with @xmath207 and @xmath21 = top@xmath208 , and similarly if @xmath148 has two neighbors @xmath209 and @xmath138 has only neighbor @xmath201 . finally , if @xmath138 and @xmath148 have another neighbor in @xmath135 , say @xmath193 , @xmath210 with @xmath149 and @xmath194 then we have : if @xmath159 then @xmath211 with @xmath212 and @xmath21 = top@xmath213 , and if @xmath163 then @xmath214 with @xmath215 and @xmath21 = top@xmath216 . * case 2 * : @xmath119 $ ] is not connected . this case can be treated similarly as case 1 in which @xmath119 $ ] is connected . if @xmath7 is not connected then we can solve mwis separately for each component of @xmath7 . if @xmath7 is connected and @xmath21 is the leaf of a claw then obviously , @xmath119 $ ] is connected . thus , we can assume that @xmath21 is the center of a claw @xmath131 , and we can follow the arguments of case 1 . for brevity let us omit the proof , which can be split into the subcases in which @xmath119 $ ] has two or three components . finally we have @xmath129 , for some @xmath217 , with @xmath21 = top@xmath130 . @xmath80 now we show that for claw+claw - free graphs , mwis can be solved in time @xmath218 . for this , we need the following notion : [ defi : good family ] let @xmath7 be a graph and let @xmath95 be a family of subsets of @xmath17 . then @xmath95 is a _ good claw - free family of _ @xmath7 if the following holds : 1 . each member of @xmath95 induces a claw - free subgraph in @xmath7 . each maximal independent set of @xmath7 is contained in some member of @xmath95 . 3 . @xmath95 contains polynomially many members and can be computed in polynomial time . the basic step is the subsequent algorithm gamma(2 ) ( based on the corresponding algorithm alpha of section [ 2k2fr ] ) which , for any claw+claw - free ( i.e. , 2claw - free ) input graph @xmath7 , computes a good claw - free family @xmath95 of @xmath7 . the approach is based on farber s idea for mwis on @xmath11-free graphs described in algorithm alpha of section [ 2k2fr ] . * algorithm gamma(2 ) * + * input : * a claw+claw - free graph @xmath7 and a vertex - ordering @xmath99 of @xmath7 . + * output : * a good claw - free family @xmath95 of @xmath7 . + @xmath109 ; + * for * @xmath110 * to * @xmath46 * do * + * begin * + 1 . [ extension of some members of @xmath95 ] + @xmath112 * do * + @xmath219 $ ] is claw - free * then * @xmath114 . [ addition of new members to @xmath95 ] + induced @xmath122 of @xmath107 , @xmath121 , with @xmath115 = top@xmath130 * do * + compute a good claw - free family , say @xmath220 , of @xmath221 $ ] . + @xmath222 , set @xmath223 . + * [ prop : claw+claw ] step @xmath224 of algorithm gamma@xmath225 is well defined , i.e. , @xmath221 $ ] is claw - free and has a good claw - free family @xmath226formed by one member , namely , @xmath227 which can be computed in constant time . * * subgraph @xmath221 $ ] is claw - free since @xmath7 is assumed to be claw+claw - free , each @xmath122 contains an induced claw and @xmath228 is defined as the anti - neighborhood of @xmath229 . then the subgraph @xmath221 $ ] has a good claw - free family ( formed by one member , namely , @xmath228 ) which can be computed in constant time . @xmath80 for proving the correctness and the time bound of algorithm gamma(2 ) , we need the following lemmas . [ lemmc:1 ] let @xmath7 be a claw+claw - free graph and let @xmath95 be the result of algorithm gamma@xmath225 . then we have : * each member of @xmath95 induces a claw - free subgraph of @xmath7 . * each maximal independent set of @xmath7 is contained in some member of @xmath95 . * @xmath230 : each member of @xmath95 is created either in the initialization step as the empty set or in step 1 or step 2 of some loop . clearly , each member @xmath231 created in step 1 induces a claw - free subgraph in @xmath7 since each member of @xmath95 is extended in step 1 only if the extension preserves its claw - freeness . according to step 2 and to proposition [ prop : claw+claw ] , each member of @xmath95 created in step 2 is the disjoint union of a vertex subset of a claw - free subgraph , namely @xmath123 , and of a claw - free subgraph representing its anti - neighborhood @xmath228 , namely a member of a good claw - free family . therefore , each member of @xmath95 created in step 2 induces a claw - free graph . this completes the proof of statement @xmath230 . @xmath232 : by @xmath105 , let us denote the family @xmath95 resulting by the @xmath102-th loop of algorithm gamma(2 ) . let us show that for all @xmath233 , each maximal independent set of @xmath107 is contained in a member @xmath34 of @xmath105 . the proof is done by induction . for @xmath234 , the statement is trivial . then let us assume that the statement holds for @xmath235 and prove that it holds for @xmath102 . let @xmath69 be a maximal independent set of @xmath107 . if @xmath236 , then by the induction assumption , @xmath69 is contained in some member of @xmath237 , and thus of @xmath105 , since each member of @xmath237 is a ( not necessarily proper ) subset of a member of @xmath105 . if @xmath238 , then let us consider the following argument . by the induction assumption , let @xmath239 with @xmath240 . note that for all @xmath241 , @xmath242 , each member of @xmath243 induces a claw - free graph , as one can easily verify by an argument similar to the proof of statement @xmath230 . thus , @xmath119 $ ] is claw - free . then let us consider the following two cases which are exhaustive by definition of algorithm gamma(2 ) . * case 1 * : @xmath219 $ ] is claw - free . then @xmath69 is contained in the set @xmath113 , which is a member of @xmath105 since it is generated by step 1 of the algorithm at loop @xmath102 . * case 2 * : @xmath219 $ ] is not claw - free . then by lemma [ lemm : claw ] , since @xmath119 $ ] is claw - free , there is a @xmath121 such that @xmath244 for an induced subgraph @xmath122 of @xmath107 with @xmath115 = top@xmath130 , and @xmath245 is contained in @xmath105 since it is generated by step 2 of algorithm gamma(2 ) at loop @xmath102 . @xmath80 [ lemmc:3 ] the family @xmath95 produced by algorithm gamma@xmath225 contains @xmath246 members and can be computed in @xmath247 time , which is also the time bound of algorithm gamma@xmath225 . * proof . * the members of @xmath95 are created either in the initialization step or in step 2 of all the loops of algorithm gamma(2 ) and then are possibly ( iteratively ) extended in step 1 of algorithm gamma(2 ) . concerning the member created in the initialization step , i.e. , the empty set : this member is created in constant time and is possibly ( iteratively ) extended by step 1 of each loop in @xmath248 time ( and the number of loops is @xmath46 ) . then this member can be computed in @xmath249 time . concerning the members created in step 2 of all the loops : such members are created with respect to all induced @xmath122 , @xmath250 ( the maximum number of vertices in any @xmath122 is 7 ) , of @xmath107 , i.e. , with respect to a family of @xmath251 subsets of @xmath107 ( in fact the algorithm produces the anti - neighborhoods of all @xmath122 for @xmath121 of @xmath107 just once since at loop @xmath102 all such @xmath122 contain @xmath115 ) . then for the respective anti - neighborhood , namely @xmath228 , of each such subset the algorithm computes a good claw - free family . by proposition [ prop : claw+claw ] , @xmath221 $ ] is claw - free and has a good claw - free family ( which contains one member and can be computed in constant time ) . therefore the cardinality of the family of such members is @xmath251 and all such members can be created in @xmath251 time ( since each such member can be created in step 2 in constant time ) . then such members are possibly ( iteratively ) extended in step 1 in @xmath248 time ( and the number of loops is @xmath46 ) . then such members can be computed in @xmath252 time . therefore , @xmath253 contains @xmath246 members and can be computed in @xmath247 time , which is also the time bound of algorithm gamma(2 ) . @xmath80 note that lemmas [ lemmc:1 ] and [ lemmc:3 ] directly imply the following . [ coro : claw+claw ] every claw+claw - free graph has a good claw - free family which can be computed by algorithm gamma@xmath225 . @xmath80 then the mwis problem can be solved for claw+claw - free graphs by the following algorithm . * algorithm mwis(2 ) * + * input : * a claw+claw - free graph @xmath7 . + * output : * a maximum weight independent set of @xmath7 . * execute algorithm gamma(2 ) for @xmath7 . let @xmath95 be the resulting family of subsets of @xmath17 . * for each @xmath112 , compute a maximum weight independent set of @xmath119 $ ] . then choose a best solution , i.e. , one of maximum weight . algorithm mwis@xmath225 is correct and can be done in @xmath218 time . * proof . * _ correctness _ : by lemma [ lemmc:1 ] @xmath232 , algorithm mwis(2 ) is correct . _ time bound _ : by lemma [ lemmc:3 ] , step ( 1 ) can be executed in @xmath247 time . by lemma [ lemmc:3 ] , the family @xmath95 contains @xmath246 members . then , by lemma [ lemmc:1 ] @xmath230 and theorem [ theo : claw ] , step ( 2 ) can be executed in @xmath218 time . thus , algorithm mwis(2 ) can be executed in time @xmath218 . @xmath80 then one obtains the following result . [ theo : wis - claw+claw ] for claw+claw - free graphs , the mwis problem can be solved in time @xmath218 by algorithm mwis@xmath225 . @xmath80 in this section we show that for any fixed @xmath4 , mwis for @xmath6claw - free graphs can be solved in polynomial time . for this , we first describe the subsequent algorithm gamma(@xmath6 ) , which for any @xmath6claw - free input graph @xmath7 computes a good claw - free family @xmath95 of @xmath7 . the approach recursively uses algorithm gamma@xmath254 for algorithm gamma(@xmath6 ) , starting with algorithm gamma(2 ) of subsection [ claw+clawfr ] . * algorithm gamma@xmath255 * + * input : * an @xmath6claw - free graph @xmath7 and a vertex - ordering @xmath99 of @xmath7 . + * output : * a good claw - free family @xmath253 of @xmath7 . + @xmath109 ; + * for * @xmath110 * to * @xmath46 * do * + * begin * + 1 . [ extension of some members of @xmath95 ] + @xmath112 * do * + @xmath219 $ ] is claw - free * then * @xmath114 . [ addition of new members to @xmath95 ] + induced @xmath122 of @xmath107 , @xmath121 , with @xmath115 = top@xmath130 * do * + compute a good claw - free family , say @xmath220 , of @xmath221 $ ] by + algorithm gamma@xmath254 . + @xmath222 , set @xmath223 . + * * assumption 1 . * to prove the subsequent proposition [ prop : lclaw ] , lemmas [ lemmc:1l ] and [ lemmc:3l ] , and corollary [ coro : lclaw ] , we need to consider them as a _ unique _ result , in order to give a proof by induction on @xmath6 . for @xmath6 = 2 , the proof of proposition [ prop : lclaw ] , of lemmas [ lemmc:1l ] and [ lemmc:3l ] , and of corollary [ coro : lclaw ] is respectively that of proposition [ prop : claw+claw ] , of lemmas [ lemmc:1 ] and [ lemmc:3 ] , and of corollary [ coro : claw+claw ] . * subgraph @xmath221 $ ] is @xmath254claw - free since @xmath7 is @xmath6claw - free and since @xmath228 is defined as the anti - neighborhood of @xmath122 containing an induced claw . then by assumption 1 , i.e. , by corollary [ coro : lclaw ] with respect to @xmath256 , subgraph @xmath221 $ ] has a good claw - free family which can be computed by algorithm gamma@xmath254 . @xmath80 * according to assumption 1 , the proof is similar to that of lemma [ lemmc:1 ] , with proposition [ prop : lclaw ] instead of proposition [ prop : claw+claw ] and with algorithm gamma(@xmath6 ) instead of algorithm gamma(2 ) . @xmath80 * proof . * the members of @xmath95 are created either in the initialization step or in step 2 of all the loops of algorithm gamma(@xmath6 ) and then are possibly ( iteratively ) extended in step 1 of algorithm gamma(@xmath6 ) . concerning the member created in the initialization step , i.e. , the empty set , this member is created in constant time and is possibly ( iteratively ) extended by step 1 of each loop in @xmath248 time ( the number of loops is @xmath46 ) . then this member can be computed in @xmath249 time . concerning the members created in steps 2 of all the loops , such members are created with respect to all induced @xmath122 of @xmath7 , i.e. , with respect to a family of @xmath251 subsets of @xmath7 ( in fact , the algorithm produces the anti - neighborhoods of all @xmath122 for @xmath121 of @xmath107 just once since at loop @xmath102 all such @xmath122 contain @xmath115 as their top vertex ) . then for the respective anti - neighborhood , namely @xmath228 , of each such subset the algorithm computes a good claw - free family . by proposition [ prop : lclaw ] , @xmath221 $ ] is @xmath254claw - free and has a good claw - free family . therefore the cardinality of the family of such members is bounded by a polynomial and all such members can be created in polynomial time ( since each such member can be created in step 2 in polynomial time ) . then such members are possibly ( iteratively ) extended in step 1 in @xmath248 time ( the number of loops is @xmath46 ) . thus , such members can be computed in polynomial time . then for @xmath6claw - free graphs , the mwis problem can be solved by the following algorithm . + * algorithm mwis(@xmath6 ) * + * input : * an @xmath6claw - free graph @xmath7 . + * output : * a maximum weight independent set of @xmath7 . * execute algorithm gamma@xmath255 for @xmath7 . let @xmath95 be the resulting family of subsets of @xmath17 . * for each @xmath112 , compute a maximum weight independent set of @xmath119 $ ] . then choose a best solution , i.e. , one of maximum weight . _ time bound _ : by lemma [ lemmc:3l ] , step ( 1 ) can be executed in polynomial time . by lemma [ lemmc:3l ] , the family @xmath95 contains polynomially many members . then by lemma [ lemmc:1l ] @xmath230 and by theorem [ theo : claw ] , step ( 2 ) can be executed in polynomial time . thus , algorithm mwis@xmath255 can be executed in polynomial time . @xmath80 v.e . alekseev , on the local restriction effect on the complexity of finding the graph independence number , _ combinatorial - algebraic methods in applied mathematics _ , gorkiy university press , gorkiy ( 1983 ) 3 - 13 ( in russian ) . alekseev , on the number of maximal independent sets in graphs from hereditary classes , _ combinatorial - algebraic methods in discrete optimization _ , gorkiy university press , gorkiy ( 1991 ) 5 - 8 ( in russian ) . alekseev , a polynomial algorithm for finding largest independent sets in fork - free graphs , _ discrete analysis and operations research _ ser . 1 , 6 ( 1999 ) 3 - 19 ( in russian ) , _ discrete applied mathematics _ 135 ( 2004 ) 3 - 16 . y. faenza , g. oriolo , g. stauffer , an algorithmic decomposition of claw - free graphs leading to an o(@xmath257)-algorithm for the weighted independent set problem , extended abstract in : proceedings soda 2011 , pp . 630 - 646 . d. lokshtanov , m. vatshelle , y. villanger , independent sets in @xmath78-free graphs in polynomial time , http://www.ii.uib.no/@xmath258martinv/papers/isinp5free.pdf ; extended abstract in : proceedings of soda 2014 : 570 - 581 .	the maximum weight independent set ( mwis ) problem is a well - known np - hard problem . a popular way to study mwis is to detect graph classes for which mwis can be solved in polynomial time , with particular reference to hereditary graph classes , i.e. , defined by a hereditary graph property or equivalently by forbidding one or more induced subgraphs . for graphs @xmath0 , @xmath1 denotes the disjoint union of @xmath2 and @xmath3 , and for a constant @xmath4 , @xmath5 denotes the disjoint union of @xmath6 copies of @xmath7 . a _ claw _ has vertices @xmath8 , and edges @xmath9 . mwis can be solved for claw - free graphs in polynomial time ; the first two polynomial time algorithms were introduced in 1980 by @xcite , then revisited by @xcite , and recently improved by @xcite , and by @xcite with the best known time bound in @xcite . furthermore mwis can be solved for the following extensions of claw - free graphs in polynomial time : fork - free graphs @xcite , @xmath10+claw - free graphs @xcite , and apple - free graphs @xcite . this manuscript shows that for any constant @xmath6 , mwis can be solved for @xmath6claw - free graphs in polynomial time . our approach is based on farber s approach showing that every @xmath11-free graph has @xmath12 maximal independent sets @xcite , which directly leads to a polynomial time algorithm for mwis on @xmath11-free graphs by dynamic programming . solving mwis for @xmath6claw - free graphs in polynomial time extends known results for claw - free graphs , for @xmath13-free graphs for any constant @xmath6 @xcite , for @xmath10+claw - free graphs , for @xmath14-free graphs @xcite , and solves the open questions for @xmath15-free graphs and for @xmath16+claw - free graphs being two of the minimal graph classes , defined by forbidding one induced subgraph , for which the complexity of mwis was an open problem . institut fr informatik , universitt rostock , d-18051 rostock , germany . + ` [email protected] ` dipartimento di economia , universit degli studi `` g. dannunzio '' , pescara 65121 , italy . + ` [email protected] ` _ ad laudem domini _
softened or `` low momentum '' interactions are widely used in contemporary nuclear structure calculations because they exhibit faster convergence for methods using basis expansions ( this includes coupled cluster , configuration interaction , in - medium similarity renormalization group , and self - consistent green s function methods ) @xcite . such interactions are derived using unitary transformations starting from chiral effective field theory or phenomenological interactions that exhibit significant coupling of high- and low - momentum physics . when done in small steps , these unitary transformations are a type of renormalization group ( rg ) transformation . the rg decoupling scale can be associated with the resolution of the interaction @xcite . but how do we handle observables involving external probes when using such interactions ? nuclear structure has conventionally been treated as largely separate from nuclear reactions . however , this separation implies a unique factorization of experimental cross sections into the structure and reaction parts . the rg perspective informs us that such a division is itself inevitably resolution dependent @xcite . in some circumstances the dependence is small and one can define the separation with negligible ambiguity . but the significant ( and beneficial ) changes to wave functions from evolving to lower resolution with rg methods imply significant changes to this separation . that is , what is structure at one resolution becomes part of the reaction mechanism at another resolution ( and vice versa ) . this separation is not only scale dependent , it is _ scheme _ dependent as well ; that is , it depends on _ how _ the separation is carried out and on the details of the original hamiltonian . this observation raises questions of consistency and uniqueness in analyzing and interpreting nuclear experimental data . for example , it is clear from previous calculations using the similarity rg ( or srg ) @xcite that the high - momentum tail of the momentum distribution in a nucleus is dramatically resolution dependent for the range of decoupling scales used in present - day nuclear structure calculations . how then can such a distribution be said to be extracted from experiment ? yet it is common in the literature that high - momentum components are treated as measurable , at least implicitly @xcite . in fact , what can be extracted is the momentum distribution at some scale , and with the specification of a scheme . this makes momentum distributions model dependent @xcite . this is relevant for recent work to extract a `` nuclear contact '' @xcite . for some physical systems under certain conditions , such as the high - momentum @xmath0 tail in cold atoms at unitarity @xcite , the scale and scheme dependence is negligible , so it _ can _ be determined essentially uniquely . but for nuclei this dependence may be substantial and one needs to carefully define the short - distance content of the nuclear contact . as illustrated in @xcite , details of the nuclear shell structure such as single - particle energies are not measurable . their extraction from experimental data involves fixing a scale and a scheme . in general , to be consistent between structure and reactions one must calculate cross sections or decay rates within a single framework . that is , one must use the same hamiltonian and consistent operators throughout the calculation ( which means the same scale and scheme ) . such consistent calculations have existed for some time for few - body nuclei ( e.g. , see @xcite ) and are becoming increasingly feasible for heavier nuclei because of advances in reaction technology , such as using complex basis states to handle continuum physics . recent examples in the literature include no core shell model resonating group method ( ncsm / rgm ) @xcite , coupled cluster @xcite , and lattice eft calculations @xcite . but there are many open questions about constructing consistent currents and how to compare results from two such calculations . some work along this direction was done in @xcite where the mean values for eigenstates of renormalized hamiltonians were calculated using evolved operators . we seek to extend this to transition matrix elements and explore the connection to high - momentum physics in a nucleus in a controlled manner . in particular , we will take the first steps in exploring the interplay of structure and reaction as a function of kinematic variables and srg decoupling scale @xmath1 in a controlled calculation of a knock - out process . there are various complications for such processes . with rg evolution , a hamiltonian even with only a two - body potential initially will develop many - body components as the decoupling scale decreases . similarly , a one - body current will develop two- and higher - body components . our strategy is to avoid dealing with all of these complications simultaneously by considering the cleanest knock - out process : deuteron electrodisintegration with only an initial one - body current . with a two - body system , there are no three - body forces or three - body currents to contend with . yet it still includes several key ingredients to investigate : i ) the wave function will evolve with changes in resolution ; ii ) at the same time , the one - body current develops two - body components , which are simply managed ; and iii ) there are final - state interactions ( fsi ) . it is these ingredients that will mix under the rg evolution . we can focus on different effects or isolate parts of the wave function by choice of kinematics . for example , we can examine when the impulse approximation is best and to what extent that is a resolution - dependent assessment . we will vary the interaction resolution using srg transformations , which have proven to be technically feasible for evolving three - body forces @xcite . the srg series of unitary transformations ensures that cross sections are invariant under changes in resolution . as the srg @xmath1 is varied , the hamiltonian @xmath2 and the nuclear wave function change , the current operator changes , and the fsi change as well . the question we address is : how do these combine to achieve the invariance of the observable cross section ? the electron scattering knock - out process is particularly interesting because of the connection to past , present , and planned experiments @xcite . the conditions for clean factorization of structure and reactions in this context is closely related to the impact of 3n forces , two - body currents , and final - state interactions , which have not been cleanly understood as yet @xcite . all of this becomes particularly relevant for high - momentum - transfer electron scattering . this physics is conventionally explained in terms of short - range correlation ( src ) phenomenology @xcite . srcs are two- or higher - body components of the nuclear wave function with high relative momentum and low center - of - mass momentum . these explanations would seem to present a puzzle for descriptions of nuclei with low - momentum hamiltonians , for which srcs are essentially absent from the wave functions . this puzzle is resolved by the unitary transformations that mandate the invariance of cross section . the physics that was described by srcs in the wave functions must shift to a different component , such as a two - body contribution from the current . this may appear to complicate the reaction problem just as we have simplified the structure part , but past work and analogies to other processes suggests that factorization may in fact become cleaner @xcite . one of our goals is to elucidate this issue , although we will only start to do so in the present work . the underlying picture is analogous to that used in deep inelastic scattering ( dis ) @xcite . in the analysis of dis , one introduces both a renormalization and a factorization scale . in the usual dimensional regularization minimal subtraction scheme @xmath3 , the renormalization scale @xmath4 sets the division between long- and short - distance physics in the hamiltonian this is manifested in the running coupling @xmath5 where the choice of @xmath6 is to optimize the efficacy of the qcd perturbation expansion . the factorization scale @xmath7 dictates the division between what goes into the reaction part , namely the purely hard process described in pqcd , and the structure part , which is subsumed into the soft parton distribution functions . changing @xmath7 changes the balance . in many cases these two scales are chosen to be the same and equal to the magnitude of the four - momentum transfer @xmath8 , in order to minimize the contribution of logarithms that can disturb the perturbative expansion . the nuclear analogs to these two scales are tied together in the srg evolution . in particular , the decoupling scale , roughly given by the value of the srg flow parameter @xmath1 , sets both these scales . for the hamiltonian , this scale clearly sets the division between low and high momentum . for operators acting on wave functions , the decoupling dictates the division between the two ; e.g. , the scale at which a one - body current is largely replaced by a two - body current . in the present paper , we will illustrate the combined interplay with quantitative calculations . we build upon previous work by anderson _ et al._on srg operator evolution for the deuteron @xcite , and the work by yang and phillips @xcite in applying chiral eft to deuteron electrodisintegration . this paper is organized as follows . in sec . [ sec : formalism ] , we briefly review the electrodisintegration formalism and develop the machinery we need for the srg evolution , which we accomplish in practice by appropriate insertions of unitary transformation matrices . in sec . [ sec : results ] , we present proof - of - principle tests and illustrate the interplay of the different components that enter in the calculation of the disintegration process . we give representative results for selected kinematics . in sec . [ sec : conclusion ] , we summarize our observations and plans to extend the calculations to other kinematics , as well as beyond the deuteron system and one - body initial currents . deuteron electrodisintegration is the simplest nucleon - knockout process and has been considered as a test ground for various @xmath9 models for a long time ( see , for example , refs . it has also been well studied experimentally @xcite . as outlined in the introduction , the absence of three - body currents and forces makes it an ideal starting point for studying the interplay with srg evolution of the deuteron wave function , current , and final - state interactions . we follow the approach of ref . @xcite , which we briefly review . the kinematics for the process in the laboratory frame is shown in fig . [ fig : deut_dis_kinematics ] . the virtual photon from electron scattering transfers enough energy and momentum to break up the deuteron into a proton and neutron . the differential cross section for deuteron electrodisintegration for unpolarized scattering in the lab frame is given by @xcite @xmath10 \,.\end{gathered}\ ] ] here @xmath11 and @xmath12 are the solid angles of the electron and the proton , @xmath13 and @xmath14 are the magnitude of incoming and outgoing electron 3-momenta , @xmath15 is the 4-momentum - squared of the virtual photon , and @xmath16 is the fine structure constant . @xmath17 is the angle between the scattering plane containing the electrons and the plane spanned by outgoing nucleons . @xmath18 are electron kinematic factors , and @xmath19 are the deuteron structure functions . these structure functions contain all the dynamic information about the process . the four structure functions are independent and can be separated by combining cross - section measurements carried out with appropriate kinematic settings @xcite . structure functions are thus cross sections up to kinematic factors and are independent of the srg scale @xmath1 . in this work we focus on the longitudinal structure function @xmath20 , following the approach of ref . @xcite . as in ref . @xcite , we carry out the calculations in the center - of - mass frame of the outgoing proton - neutron pair . in this frame the photon four - momentum is @xmath21 , which can be obtained from the initial electron energy and @xmath22 , the electron scattering angle . we denote the momentum of the outgoing proton by @xmath23 and take @xmath24 to be along @xmath25-axis . the angles of @xmath26 are denoted by @xmath27 . the longitudinal structure function can be written as @xmath28 where @xmath29 and @xmath30 are the spin quantum numbers of the final neutron - proton state , @xmath31 is the polarization index of the virtual photon , and @xmath32 is the angular momentum of the initial deuteron state . the amplitude @xmath33 is given by @xcite @xmath34 where @xmath35 is the final - state wavefunction of the outgoing neutron - proton pair , @xmath36 is the initial deuteron state , and @xmath37 is the current operator that describes the momentum transferred by the photon . the variables in eq . are : * fine - structure constant @xmath16 ; * outgoing proton ( neutron ) 3-momentum @xmath38 ; * proton energy @xmath39 , where @xmath40 is the average of proton and neutron mass * deuteron energy @xmath41 , where @xmath42 is the mass of the deuteron . as mentioned before , all of these quantities are in the center - of - mass frame of the outgoing nucleons . for @xmath43 , @xmath44 and therefore only @xmath45 contributes . the one - body current matrix element is given by @xmath46 where @xmath47 and @xmath48 are the electric form factors of the proton and the neutron , and the deuteron state has isospin @xmath49 . the final - state wave function of the outgoing proton - neutron pair can be written as @xmath50 where @xmath51 denotes a relative plane wave , @xmath52 and @xmath53 are the green s function and the @xmath53-matrix respectively , and @xmath54 is the energy of the outgoing nucleons . the second term in eq . describes the interaction between the outgoing nucleons . in the impulse approximation ( ia ) as defined here , the interaction between the outgoing nucleons is ignored and @xmath55 . the plane wave @xmath56 will have both isospin @xmath57 and @xmath58 components . the current @xmath45 , @xmath52 , and the @xmath53-matrix are diagonal in spin space . the deuteron has spin @xmath59 and therefore the final state will also have @xmath59 . hence , we have & = _ t = 0,1 ( \|p^ s=1 m_s_f + & + ( -1)^t \|-p^ s=1 m_s_f ) \|t . [ eq : phi_def ] using eqs . and , the overlap matrix element in ia becomes @xmath60 \ , , \label{eq : overlap_ia}\end{gathered}\ ] ] where @xmath61 is the solid angle between the unit vector @xmath62 and @xmath63 . @xmath64 is the deuteron wave function in momentum space defined as @xmath65 the @xmath66-wave @xmath67 and @xmath68-wave @xmath69 components of the deuteron wave function satisfy the normalization condition @xmath70 in deriving eq . we have used the property of the spherical harmonics that @xmath71 in our work we follow the conventions of ref . @xcite . since @xmath72 and @xmath73 are the angles of @xmath26 , @xmath74 and @xmath75 , where @xmath76 and @xmath77 the overlap matrix element including the final - state interactions ( fsi ) is given by @xmath78 the first term on the right side of eq . has already been evaluated in eq . . therefore , the term we still need to evaluate is @xmath79 . the @xmath53-matrix is most conveniently calculated in a partial - wave basis . hence , the fsi term is evaluated by inserting complete sets of states in the form @xmath80 the outgoing plane - wave state in the partial - wave basis is given by @xmath81 the green s function is diagonal in @xmath82 , @xmath83 , @xmath84 , @xmath66 , and @xmath85 , so we have @xmath86 we also need to express the current in eq . in the partial - wave basis . to begin with , let us just work with first term in eq . , which we denote by @xmath87 . in the partial - wave basis , it is written as @xmath88 here @xmath32 is the deuteron quantum number , which is preserved throughout . we have used the deuteron quantum numbers in the ket in anticipation that we will always evaluate the matrix element of @xmath45 with the deuteron wave function on the right . @xmath89 is as defined in eq . . in deriving eq . we have also made use of the relation @xcite @xmath90 equations , , and can be combined to obtain @xmath91 we denote the second term in the one - body current eq . by @xmath92 . the expression for @xmath93 is analogous to eq . , the only differences being that the form - factor coefficient is @xmath94 and the input arguments for the second associated legendre polynomial and the deuteron wave function are different . the two factors respectively become @xmath95 and @xmath96 , where @xmath97 is defined in eq . . it can be shown that @xmath98 . thus , @xmath99 using this we can evaluate the overlap matrix element in eq . . as outlined in eqs . and , this matrix element is related to the longitudinal structure function @xmath20 . recall that the deuteron spin is conserved throughout and therefore @xmath100 in eq . . in sec [ subsec : illustrative_examples ] we present results for @xmath20 both in the ia and including the fsi . these results match those of ref . @xcite , verifying the accuracy of the calculations presented above . as outlined in the introduction , we want to investigate the effect of unitary transformations on calculations of @xmath20 . let us start by looking at the ia matrix element : & = _ a + _ b , [ eq : a_b_split_up ] where we decompose the unitary matrix @xmath101 into the identity and a residual @xmath102 , @xmath103 the matrix @xmath102 is smooth and therefore amenable to interpolation . the @xmath101 matrix is calculated following the approach in @xcite . the terms in eq . can be further split into @xmath104 and @xmath105 = \underbrace{{\langle}\phi \| \widetilde{u}^\dag \ , \widetilde{u}\ , j_0 \ , \widetilde{u}^\dag \| \psi_i^\lambda { \rangle}}_{a_1 } + \underbrace{{\langle}\phi \| \widetilde{u}^\dag \ , \widetilde{u}\ , j_0 \| \psi_i^\lambda { \rangle}}_{a_2 } \\ \null + \underbrace{{\langle}\phi \| \widetilde{u}^\dag j_0 \ , \widetilde{u}^\dag \| \psi_i^\lambda { \rangle}}_{a_3 } + \underbrace{{\langle}\phi \| \widetilde{u}^\dag j_0 \| \psi_i^\lambda{\rangle}}_{a_4 } \ , . \label{eq : a_split_up}\end{gathered}\ ] ] the @xmath106 term is the same as in eq . , but with the deuteron wave function replaced by the evolved version @xmath107 . inserting complete sets of partial - wave basis states as in eq . and using eqs . and , we can obtain the expressions for @xmath108 , @xmath109 , @xmath110 and @xmath111 . these expressions are given in appendix [ appendix : evol_eqns ] . using the expressions for @xmath112 and @xmath113 , we can obtain results for @xmath20 in the ia with one or more components of the overlap matrix element @xmath114 evolved . when calculated in ia , @xmath20 with all components evolved matches its unevolved counterpart , as shown later in sec . [ subsec : illustrative_examples ] . the robust agreement between the evolved and unevolved answers indicates that the expressions derived for @xmath115 are correct and that there is no error in generating the @xmath101-matrices . in sec . [ subsec : numerical_implementation ] we provide some details about the numerical implementation of the equations presented here . let us now take into account the fsi and study the effects of evolution . the overlap matrix element should again be unchanged under evolution , @xmath116 where @xmath117 is given by eq . . furthermore , @xmath118 where @xmath119 is the evolved @xmath53-matrix , _ i.e. _ , the @xmath53-matrix obtained by solving the lippmann schwinger equation using the evolved potential , as discussed in appendix [ sec : finalstate ] . thus @xmath120 the term @xmath121 is the same that we already encountered in eq . . the term @xmath122 can also be split up into four terms : @xmath123 = \underbrace{{\langle \phi\|t_\lambda ^\dag \ , g_0^\dag \ , \widetilde{u } \ , j_0 \ , \widetilde{u}^\dag\|\psi_i^\lambda\rangle}}_{f_1 } + \underbrace{{\langle \phi\|t_\lambda ^\dag \ , , \widetilde{u } \ , j_0 \|\psi_i^\lambda\rangle}}_{f_2 } \\ \null + \underbrace{{\langle \phi\|t_\lambda ^\dag \ , g_0^\dag \ , j_0 \ , \widetilde{u}^\dag\|\psi_i^\lambda\rangle}}_{f_3 } + \underbrace{{\langle \phi\|t_\lambda ^\dag \ , g_0^\dag \ , j_0\|\psi_i^\lambda\rangle}}_{f_4 } \ , . \label{eq : f_split_up}\end{gathered}\ ] ] the expression for @xmath124 can easily be obtained from eqs . and by replacing the deuteron wave function and the @xmath53-matrix by their evolved counterparts . as before , we insert complete sets of partial - wave basis states using eq . and evaluate @xmath125 , @xmath126 , and @xmath127 ; see eqs . , , and . figures in sec . [ subsec : illustrative_examples ] compare @xmath20 calculated from the matrix element with all components evolved to the unevolved @xmath20 . we find an excellent agreement , validating the expressions for @xmath128 . there are various practical issues in the calculation of evolved matrix elements that are worth detailing . we use c++11 for our numerical implementation of the expressions discussed in the previous section . matrix elements with a significant number of components evolved are computationally quite expensive due to a large number of nested sums and integrals ( see in particular appendix [ appendix : evol_eqns ] ) . the deuteron wave function and @xmath9 @xmath53-matrix are obtained by discretizing the schrdinger and lippmann schwinger equations , respectively ; these equations are also used to interpolate the @xmath53-matrix and wave function to points not on the discretized mesh . for example , if we write the momentum - space schrdinger equation neglecting channel coupling here for simplicity as @xmath129 it can be solved numerically as a simple matrix equation by setting @xmath130 . for any @xmath131 not on this mesh , the sum in eq . can then be evaluated to get @xmath132 . this technique is based on what has been introduced in connection with contour - deformation methods in break - up scattering calculations @xcite . to interpolate the potential , which is stored on a momentum - space grid , we use the two - dimensional cubic spline algorithm from alglib @xcite . in order to avoid unnecessary recalculation of expensive quantities in particular of the off - shell @xmath53-matrix while still maintaining an implementation very close to the expressions given in this paper , we make use of transparent caching techniques . for most integrations , in particular those involving a principal value , we use straightforward nested gaussian quadrature rules ; only in a few cases did we find it more efficient to use adaptive routines for multi - dimensional integrals . with these optimizations , the calculations can in principle still be run on a typical laptop computer . in practice , we find it more convenient to use a small cluster , with parallelization implemented using the tbb library @xcite . on a node with 48 cores , generating data for a meaningful plot ( like those shown in sec . [ sec : results ] ) can then be done in less than an hour . for higher resolution and accuracy , we used longer runs with a larger number of data and integration mesh points . recall that from eqs . and we have @xmath133 when all three components the final state , the current , and the initial state are evolved consistently , then @xmath20 is unchanged . however , if we miss evolving a component , then we obtain a different result . it is instructive to illustrate this through a first - order analytical calculation . let us look at the effects due to the evolution of individual components for a general matrix element @xmath134 . the evolved initial state is given by @xmath135 where @xmath102 is the smooth part of the @xmath101-matrix defined in eq . . similarly , we can write down the expressions for the evolved final state and the evolved operator as @xmath136 and @xmath137 we assume here that @xmath102 is small compared to @xmath138 ( which can always be ensured by choosing the srg @xmath1 large enough ) and therefore keep terms only up to linear order in @xmath102 . using eqs . , , and , we get an expression for the evolved matrix element in terms of the unevolved one and changes to individual components due to evolution : @xmath139 @xmath140 we see that the change due to evolution in the operator is equal and opposite to the sum of changes due to the evolution of the initial and final states . we also find that changes in each of the components are of the same order , and that they mix ; this feature persists to higher order . therefore , if one misses evolving an individual component , one will not reproduce the unevolved answer . for our analysis , we studied the effect of evolution of individual components on @xmath20 for selected kinematics in the ranges @xmath141@xmath142 and @xmath143@xmath144 , where @xmath145 is the energy of outgoing nucleons and @xmath146 is the three - momentum transferred by the virtual photon ; both are taken in the center - of - mass frame of the outgoing nucleons . this range was chosen to cover a variety of kinematics and motivated by the set covered in ref . we use the argonne @xmath147 potential ( av18 ) @xcite for our calculations . it is one of the widely used potentials for nuclear few - body reaction calculations , particularly those involving large momentum transfers @xcite . how strong the evolution of individual components ( or a subset thereof ) affects the result for @xmath20 depends on the kinematics . one kinematic configuration of particular interest is the so - called quasi - free ridge . as discussed in sec . [ subsec : disintegration_primer ] , the four - momentum transferred by the virtual photon in the center - of - mass frame is @xmath148 . the criterion for a configuration to lie on the quasi - free ridge is @xmath149 . physically , this means that the nucleons in the deuteron are on their mass shell . as shown in ref . @xcite , at the quasi - free ridge the energy of the outgoing nucleons ( @xmath145 ) and the photon momentum transfer are related by @xmath150 which reduces to @xmath151 the quasi - free condition in the center - of - mass frame is the same as the quasi - elastic condition in the lab frame . there , the quasi - elastic ridge is defined by @xmath152 , where @xmath153 is the invariant mass . on the quasi - elastic ridge , the so - called missing momentum . ] vanishes , @xmath154 . in fig . [ fig : evolution_at_qfr ] we plot @xmath20 along the quasi - free ridge both in the impulse approximation ( ia ) and with the final - state interactions ( fsi ) included as a function of energy of the outgoing nucleons for a fixed angle , @xmath155 of the outgoing proton . @xmath156 and @xmath146 in fig . [ fig : evolution_at_qfr ] are related by eq . . comparing the solid curve labeled @xmath157 in the legend to the dashed curve ( labeled @xmath158 ) we find that fsi effects are minimal for configurations on the quasi - free ridge especially at large energies . calculated at various points on the quasi - free ridge for @xmath159 for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 is evolved . there are no appreciable evolution effects all along the quasi - free ridge . the effect due to evolution of the final state is small as well and is not shown here to avoid clutter . @xmath20 calculated in the impulse approximation is also shown for comparison . ] in an intuitive picture , this is because after the initial photon is absorbed , both the nucleons in the deuteron are on their mass shell at the quasi - free ridge , and therefore no fsi are needed to make the final - state particles real . as we move away from the ridge , fsi become more important , as additional energy - momentum transfer is required to put the neutron and the proton on shell in the final state . the difference between full @xmath20 and @xmath20 in ia at small energies is also seen to hold for few - body nuclei @xcite . figure [ fig : evolution_at_qfr ] also shows @xmath20 calculated from evolving only one of the components of the matrix element in eq . . we note that the effects of srg evolution of the individual components are minimal at the quasi - free ridge as well . the kinematics at the quasi - free ridge are such that only the long - range ( low - momentum ) part of the deuteron wave function is probed , the fsi remains small under evolution , and then unitarity implies minimal evolution of the current . as one moves away from the quasi - free ridge , the effects of evolution of individual components become prominent . note that @xmath160 and therefore the unevolved vs. all - evolved @xmath20 overlap in fig . [ fig : evolution_at_qfr ] . . the effects of evolution get progressively prominent as one moves further away from the quasi - free ridge . the kinematics of the labeled points are considered in sec . [ subsec : illustrative_examples ] . ] figure [ fig : more-1p5 ] shows the ` phase space ' of kinematics for srg @xmath161 . the quasi - free ridge is along the solid line in fig . [ fig : more-1p5 ] . in the shaded region the effects generated by the evolution of individual components are weak ( only a few percent relative difference ) . as one moves away from the quasi - free ridge , these differences get progressively more prominent . the terms ` small ' and ` weak ' in fig . [ fig : more-1p5 ] are used in a qualitative sense . in the shaded region denoted by ` weak effects ' , the effects of evolution are not easily discernible on a typical @xmath20 versus @xmath72 plot , as seen in fig . [ fig:100_10_quasi_free_fsi ] , whereas in the region labeled by ` strong effects ' , the differences due to evolution are evident on a plot ( e.g. , see fig . [ fig:30_16_fsi ] ) . the size of the shaded region in fig . [ fig : more-1p5 ] depends on the srg @xmath1 . it is large for high @xmath1 s and gets smaller as the @xmath1 is decreased ( note that smaller srg @xmath1 means greater evolution ) . in the next subsection we look in detail at a few representative kinematics , indicated by points in fig . [ fig : more-1p5 ] . as a representative of quasi - free kinematics , we choose @xmath162 and @xmath163 and plot @xmath20 as a function of angle in fig . [ fig:100_10_quasi_free_fsi ] . the effect of including fsi is small for this configuration for all angles . also , the effects due to evolution of the individual components are too small to be discernible . all this is consistent with the discussion in the previous section . calculated for @xmath162 and @xmath163 ( point `` 1 '' in fig . [ fig : more-1p5 ] ) for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 is evolved . @xmath72 is the angle of the outgoing proton in the center - of - mass frame . there are no discernible evolution effects for all angles . the effect due to evolution of the final state is small as well and is not shown here to avoid clutter . @xmath20 calculated in the ia , @xmath158 , is also shown for comparison . ] next we look at the kinematics @xmath164 and @xmath165 , which is near the quasi - free ridge . this is the point `` 2 '' in fig . [ fig : more-1p5 ] . as seen in fig . [ fig:10_4_fsi ] , the different curves for @xmath20 obtained from evolving different components start to diverge . figure [ fig:10_4_fsi ] also shows @xmath20 calculated in ia . comparing this to the full @xmath20 including fsi , we see that the effects due to evolution are small compared to the fsi contributions . this smallness prevents us from making any systematic observations about the effects due to evolution at this kinematics . we thus move on to kinematics which show more prominent effects . calculated for @xmath164 and @xmath166 ( point `` 2 '' in fig . [ fig : more-1p5 ] ) for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 is evolved . @xmath20 calculated in the ia , @xmath158 , is also shown for comparison . the effects due to evolution of individual components on @xmath20 are discernible , but still small ( compared to the fsi contribution ) . the effect due to evolution of the final state is small as well and is not shown here to avoid clutter . ] we next look in the region where @xmath167 , _ i.e. _ , below the quasi - free ridge in fig . [ fig : more-1p5 ] . we look at two momentum transfers @xmath168 and @xmath169 for @xmath170 , which are points `` 3 '' and `` @xmath171 '' in fig . [ fig : more-1p5 ] . figures [ fig:30_16_fsi ] and [ fig:30_25_fsi ] indicate the effects on @xmath20 from evolving individual components of the matrix elements . it is noteworthy that in both cases evolution of the current gives a prominent enhancement , whereas evolution of the initial and final state gives a suppression . when all the components are evolved consistently , these changes combine and we recover the unevolved answer for @xmath20 . this verifies the accurate implementation of the equations derived in sec . [ sec : evolution_set_up ] . calculated for @xmath172 and @xmath168 ( point `` 3 '' in fig . [ fig : more-1p5 ] ) for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 is evolved . prominent enhancement with evolution of the current only and suppression with evolution of the initial state and the final state only , respectively . ] calculated for @xmath172 and @xmath169 ( point `` 3@xmath173 '' in fig . [ fig : more-1p5 ] ) for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 is evolved . prominent enhancement with evolution of the current only and suppression with evolution of the initial state and the final state only , respectively . ] it is possible to qualitatively explain the behavior seen in figs . [ fig:30_16_fsi ] and [ fig:30_25_fsi ] . as noted in eq . , the overlap matrix element is given by the sum of the ia part and the fsi part . below the quasi - free ridge these two terms add constructively . in this region , @xmath20 calculated in impulse approximation is smaller than @xmath20 calculated by including the final - state interactions . [ [ evolving - the - initial - state ] ] evolving the initial state + + + + + + + + + + + + + + + + + + + + + + + + + + let us first consider the effect of evolving the initial state only . we have @xmath174 as seen in eq . , in the term @xmath175 the deuteron wave function is probed between @xmath176 and @xmath177 . these numbers are @xmath178 and @xmath179 for @xmath170 , @xmath168 and @xmath170 , @xmath169 , respectively . the evolved deuteron wave function is significantly suppressed at these high momenta . this behavior is reflected in the deuteron momentum distribution plotted in fig . [ fig : deut_md ] . the deuteron momentum distribution @xmath180 is proportional to the sum of the squares of @xmath66- and @xmath68- state deuteron wave functions . thus , the first ( ia ) term in eq . is much smaller than its unevolved counterpart in eq . , for all angles . we note that even though we only use the av18 potential to study changes due to evolution , these changes will be significant for other potentials as well . lo chiral eft @xcite potentials , and for the av18 potential evolved to two srg @xmath1 s . ] evaluation of the second ( fsi ) term in eq . involves an integral over all momenta , as indicated in eq . . we find that latexmath:[$\|{\langle \phi\|t^\dag \ , { g_0}^\dag \ , j_0\|\psi_i^\lambda\rangle}\| < mentioned before , because the terms @xmath158 and @xmath182 add constructively below the quasi - free ridge and because the magnitude of both these terms decreases upon evolving the wave function , we have @xmath183 the above relation holds for most combinations of @xmath32 and @xmath184 . for those @xmath32 and @xmath184 for which eq . does not hold , the absolute value of the matrix element is much smaller than for those for which the eq . _ does _ hold , and therefore we have @xmath20 calculated from @xmath185 smaller than the @xmath20 calculated from @xmath157 , as seen in figs . [ fig:30_16_fsi ] and [ fig:30_25_fsi ] . [ [ evolving - the - final - state ] ] evolving the final state + + + + + + + + + + + + + + + + + + + + + + + + as indicated in eq . , evolving the final state entails the evolution of the @xmath53-matrix . the overlap matrix element therefore is @xmath186 the ia term is the same as in the unevolved case . the srg evolution leaves the on - shell part of the @xmath53-matrix which is directly related to observables invariant . the magnitude of the relevant off - shell @xmath53-matrix elements decreases on evolution , though . as a result we have @xmath187 this is reflected in @xmath20 as calculated from the evolved final state , and seen in figs . [ fig:30_16_fsi ] and [ fig:30_25_fsi ] . the effect of evolution of the initial state and the final state is to suppress @xmath20 . when all the three components are evolved , we reproduce the unevolved answer as indicated in fig . [ fig:30_16_fsi ] and [ fig:30_25_fsi ] . it is therefore required that we find a huge enhancement when just the current is evolved . the kinematics @xmath170 , @xmath169 is further away from the quasi - free ridge than @xmath170 , @xmath168 . the evolution effects discussed above get progressively more prominent the further away one is from the quasifree ridge . this can be verified by comparing the effects due to evolution of individual components in figs . [ fig:30_16_fsi ] and [ fig:30_25_fsi ] . as remarked earlier , away from the quasi - free ridge the fsi become important . nonetheless , it is still instructive to look at @xmath20 calculated in the ia at these kinematics . in ia @xmath188 calculated for @xmath172 and @xmath168 for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 are evolved . ] in ia @xmath188 calculated for @xmath172 and @xmath169 for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 are evolved . ] note that the ( unevolved ) @xmath20 calculated in the ia , shown in figs . [ fig:30_16_ia ] and [ fig:30_25_ia ] , is smaller than the full @xmath20 that takes into account the final state interactions ( _ cf . _ the corresponding curves in figs . [ fig:30_16_fsi ] and [ fig:30_25_fsi ] ) . this is consistent with the claim made earlier that below the quasi - free ridge the two terms in eq . add constructively . the results in figs . [ fig:30_16_ia ] and [ fig:30_25_ia ] can again be qualitatively explained based on our discussion above . the evolution of the deuteron wave function leads to suppression as the evolved wave function does not have strength at high momentum . the evolved current thus leads to enhancement . evolution of both the current and the initial state decreases @xmath20 from just the evolved current value , but it is not until we evolve all three components final state , current , and the initial state that we recover the unevolved answer . in ia calculated at @xmath159 for @xmath172 and @xmath169 for the av18 potential when the current operator in eq . used to calculate @xmath20 is evolved to various srg @xmath1 s . the horizontal dotted line is the unevolved answer . ] as expected , the effect due to evolution increases with further evolution . this is illustrated in fig . [ fig : j0_evolution_vs_lambda ] , where we investigate the effects of the current - operator evolution on @xmath20 as a function of the srg @xmath1 . to isolate the effect of operator evolution , we only look at @xmath20 calculated in ia at a specific angle in fig . [ fig : j0_evolution_vs_lambda ] . finally , we look at an example from above the quasi - free ridge . figure [ fig:100_0p5_fsi ] shows the effect of evolution of individual components on @xmath20 for @xmath189 and @xmath190 , which is point `` 4 '' in fig . [ fig : more-1p5 ] . the effects of evolution in this case are qualitatively different from those found below the quasi - free ridge . for instance , we see a peculiar suppression in @xmath20 calculated from the evolved deuteron wave function at small angles , but an enhancement at large angles . an opposite behavior is observed for the final state . it is again possible to qualitatively explain these findings . calculated for @xmath162 and @xmath190 ( point `` 4 '' in fig . [ fig : more-1p5 ] ) for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 is evolved . opposite effects from the evolution of the initial state and the final state . ] [ [ evolving - the - initial - state-1 ] ] evolving the initial state + + + + + + + + + + + + + + + + + + + + + + + + + + above the quasi - free ridge , the ia and fsi terms in eq . add destructively . this can be seen by comparing the unevolved @xmath20 curves in figs . [ fig:100_0p5_fsi ] and [ fig:100_0p5_ia ] . including the fsi brings down the value of @xmath20 when one is above the quasi - free ridge . at small angles , the magnitude of the ia term in eq . is larger than that of the fsi term . the deuteron wave function for this kinematics is probed between @xmath191 and @xmath192 . with the wave - function evolution , the magnitude of the ia term in eq . decreases , whereas the magnitude of the fsi term in that equation slightly increases compared to its unevolved counterpart . still , at small angles , we have @xmath193 , which leads to @xmath194 and thus to the suppression of @xmath20 at small angles observed in fig . [ fig:100_0p5_fsi ] . at large angles , the magnitude of the ia term in eq . is smaller than that of the fsi term . with the wave - function evolution , the magnitude of ia term decreases substantially ( large momenta in the deuteron wave function are probed at large angles , _ cf . _ eq . ) , whereas the fsi term in eq . remains almost the same . this results in increasing the difference between the two terms in eq . as the srg @xmath1 is decreased . as mentioned before , above the quasi - free ridge , the ia and fsi terms in eq . add destructively and we therefore end up with @xmath195 ) . [ [ evolving - the - final - state-1 ] ] evolving the final state + + + + + + + + + + + + + + + + + + + + + + + + the expression to consider is eq . . with the evolution of the @xmath53-matrix , the magnitude of the term @xmath196 decreases , and because of the opposite relative signs of the two terms in eq . and because at small angles the magnitude of the ia term is larger than the fsi term the net effect is latexmath:[$\|{\langle \psi_f^\lambda\|j_0\|\psi_i\rangle}\| > @xmath20 with evolved final state at small angles , as seen in fig . [ fig:100_0p5_fsi ] . at large angles the magnitude of the ia term in eq . is smaller than that of the fsi term . with the evolution of the @xmath53-matrix , the magnitude of the fsi term decreases and the difference between the ia and the fsi terms decreases as well . this leads to the observed overall suppression in @xmath20 at large angles due to the evolution of the final state seen in fig . [ fig:100_0p5_fsi ] . for those few ( @xmath184 , @xmath32 ) combinations for which the above general observations do not hold , the value of individual components is too small to make any qualitative difference . in ia @xmath188 calculated for @xmath162 and @xmath190 for the av18 potential . legends indicate which component of the matrix element in eq . used to calculate @xmath20 are evolved . ] figure [ fig:100_0p5_ia ] shows the effect of evolution of individual components on @xmath20 calculated in the ia for the kinematics under consideration . again the evolved deuteron wave function does not have strength at high momenta and therefore @xmath20 calculated from @xmath198 has a lower value than its unevolved counterpart . unitary evolution means that the effect of the evolved current is always such that it compensates the effect due to the evolution of the initial and final states . in future work we will examine more directly the behavior of the current as it evolves to better understand how to carry over the results observed here to other reactions . nuclear properties such as momentum distributions are extracted from experiment by invoking the factorization of structure , which includes descriptions of initial and final states , and reaction , which includes the description of the probe components . the factorization between reaction and structure depends on the scale and scheme chosen for doing calculations . unlike in high - energy qcd , this scale and scheme dependence of factorization is often not taken into account in low - energy nuclear physics calculations , but is potentially critical for interpreting experiment . in our work we investigated this issue by looking at the simplest knockout reaction : deuteron electrodisintegration . we used srg transformations to test the sensitivity of the longitudinal structure function @xmath20 to evolution of its individual components : initial state , final state , and the current . we find that the effects of evolution depend on kinematics , but in a _ systematic _ way . evolution effects are negligible at the quasi - free ridge , indicating that the scale dependence of individual components is minimal there . this is consistent with the quasi - free ridge mainly probing the long - range part of the wave function , which is largely invariant under srg evolution . this is also the region where contributions from fsi to @xmath20 are minimal . the effects get progressively more pronounced the further one moves away from the quasifree ridge . the nature of these changes depends on whether one is above or below the quasifree ridge in the `` phase - space '' plot ( fig . 3 ) . as indicated in sec . iiic , these changes can also be explained qualitatively by looking at the overlap matrix elements . this allows us to predict the effects due to evolution depending on kinematics . our results demonstrate that scale dependence needs to be taken into account for low - energy nuclear calculations . while we showed this explicitly only for the case of the longitudinal structure function in deuteron disintegration , we expect the results should qualitatively carry over for other knock - out reactions as well . an area of active investigation is the extension of the formalism presented here to hard scattering processes . srg transformations are routinely used in nuclear structure calculations because they lead to accelerated convergence for observables like binding energies . we demonstrated that srg transformations can be used for nuclear knock - out reactions as well as long as the operator involved is also consistently evolved . the evolved operator appears to be complicated compared to the unevolved one , but the factorization might be cleaner with the evolved operator if one can exploit an operator product expansion @xcite . we plan to explore this in our future work . we also plan to use pionless eft as a framework to quantitatively study the effects of operator evolution . it should be a good starting point to understand in detail how a one - body operator develops strength in two- and higher - body sectors upon evolution . this can give insight on the issue of power counting of operator evolution . pionless eft has been employed previously to study deuteron electrodisintegration in ref . @xcite , where it was used to resolve a discrepancy between theory and experiment . extending our work to many - body nuclei requires inclusion of 3n forces and 3n currents . consistent evolution in that case would entail evolution in both two and three - body sectors . however , srg transformations have proven to be technically feasible for evolving three - body forces @xcite . thus , extending our calculations to many - body nuclei would be computationally intensive , but is feasible in the existing framework . including the effects of fsi is challenging for many - body systems and has been possible only recently for light nuclei @xcite . it would be interesting to investigate if the scale and scheme dependence of factorization allows us to choose a scale where the fsi effects are minimal . we are grateful to s. jeschonnek , a. schwenk , and h. hergert for useful discussion . we would also like to thank h. arenhvel and c .- j . yang for patiently answering our questions about the deuteron electrodisintegration formalism , as well as for sharing their data with us to compare results . this work was supported in part by the national science foundation under grant no . phy1306250 , the nuclei scidac collaboration under doe grant de - sc0008533 and by the erc grant no . 307986 strongint . here we document the expressions used in sec . [ sec : evolution_set_up ] . as seen in eq . , in order to evaluate the term @xmath199 we split it into four terms : @xmath108 , @xmath109 , @xmath110 , and @xmath106 . @xmath106 is obtained from eq . by using the evolved deuteron wave function instead of the unevolved one . the expressions for the terms @xmath110 , @xmath109 , and @xmath108 are as follows : @xmath200 @xmath201 @xmath202 in deriving the equations for @xmath108 , @xmath109 , and @xmath110 we have made use of the fact that the matrix elements with @xmath45 are twice the matrix elements with @xmath203 , _ i.e. _ , @xmath204 , and similarly for @xmath109 and @xmath108 ( _ cf . _ sec . [ sec : calculating_fl ] ) . evaluating eq . involves calculating the individual terms @xmath205 , @xmath206 , @xmath207 , and @xmath208 . the expressions for @xmath208 and @xmath207 can be obtained from expressions for @xmath109 and @xmath108 , respectively , by replacing @xmath102 with @xmath209 . the @xmath101-matrices are real . therefore , @xmath209 is obtained from @xmath102 by interchanging momentum and angular momentum indices . the expressions for @xmath206 and @xmath205 are @xmath210 and @xmath211 evaluating the evolved current while including the final - state interactions involves computing the terms @xmath127 , @xmath126 , @xmath125 , and @xmath124 , as indicated in eq . . @xmath124 is obtained from eqs . and by replacing the deuteron wave function and the @xmath53-matrix by their evolved counterparts . the expressions for the terms @xmath125 , @xmath126 , and @xmath127 are then as follows : @xmath212 @xmath213 @xmath214 the interacting final neutron - proton state @xmath215 as defined in eq . is the formal solution of the lippmann schwinger ( ls ) equation for the scattering wave function , the @xmath53-matrix , in turn , is defined by the ls equation @xmath216 the subsitution @xmath217 and the limit @xmath218 are implied to select outgoing boundary conditions . we want to show now that the srg - evolved final state can be obtained directly by using the solution @xmath219 of eq . with @xmath220 in the second line of eq . , which is the same as eq . in sec [ sec : formalism ] , _ i.e. _ , @xmath221 where @xmath222 in this section , we suppress all spin and isospin degrees of freedom , and only denote the ( arbitrary ) energy parameter as @xmath145 for consistency with sec . [ sec : formalism ] . first , it is important to recall that by definition the free hamiltonian @xmath223 does not evolve , so that for @xmath224 we have @xmath225 in other words , the evolved potential @xmath226 is defined such that it absorbs the evolution of the initial free hamiltonian ( kinetic energy ) as well . here , @xmath229 is the free green s function ( which does not change under the srg evolution because @xmath223 does not ) , and @xmath25 is an arbitrary complex energy parameter that is set to @xmath230 to recover the physically relevant case . the green s functions can be expressed in terms of the @xmath53-matrix as @xmath231 and analogously for the evolved version . furthermore , the green s functions can be written in their spectral representations 54ifxundefined [ 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 10.1088/0034 - 4885/76/12/126301 [ * * , ( ) ] , link:\doibase 10.1016/j.ppnp.2010.03.001 [ * * , ( ) ] , link:\doibase 10.1088/0954 - 3899/37/6/064005 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.82.054001 [ * * , ( ) ] , in http://inspirehep.net/record/1255127/files/arxiv:1309.5771.pdf[__ ] ( ) p. , link:\doibase 10.1142/s0217751x08041207 [ * * , ( ) ] , link:\doibase 10.1016/j.ppnp.2012.04.002 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.89.044303 [ * * , ( ) ] , link:\doibase 10.1142/s0218301315300039 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.90.064006 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.92.044003 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/physrevlett.114.012501 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.92.054311 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.110.055305 [ * * , ( ) ] link:\doibase 10.1103/physrevc.92.034313 [ * * , ( ) ] , link:\doibase 10.1103/revmodphys.81.1773 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1103/revmodphys.87.1067 [ ( ) , 10.1103/revmodphys.87.1067 ] , @noop ( ) , ( ) link:\doibase 10.1103/physrevlett.111.122502 [ * * , ( ) ] , link:\doibase 10.1140/epja / i2013 - 13151 - 3 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.90.011301 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.92.024003 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.103.082501 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.83.034301 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.85.021002 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.87.061001 [ * * , ( ) ] , @noop _ _ ( , , ) @noop `` , '' ( ) link:\doibase 10.1016/j.physrep.2015.06.002 [ * * , ( ) ] link:\doibase 10.1103/physrevc.86.064304 [ * * , ( ) ] , link:\doibase 10.1140/epja / i2013 - 13122 - 8 [ * * , ( ) ] , link:\doibase 10.1140/epja / i2004 - 10061 - 5 [ * * , ( ) ] , link:\doibase 10.1016/s0375 - 9474(98)00028 - 1 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.262502 [ * * , ( ) ] , @noop * * , ( ) _ _ , @noop ph.d . thesis , ( ) link:\doibase 10.1103/physrevc.46.455 [ * * , ( ) ] @noop _ _ ( , , ) _ _ , @noop , ( ) link:\doibase 10.1103/physrev.137.b935 [ * * , ( ) ] @noop _ _ ( , , ) @noop `` , '' @noop `` , '' ( ) @noop * * , ( ) , link:\doibase 10.1103/physrevc.51.38 [ * * , ( ) ] , link:\doibase 10.1103/revmodphys.70.743 [ * * , ( ) ] link:\doibase 10.1088/0954 - 3899/41/12/123002 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.63.024001 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1103/physrevc.77.064001 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.91.062501 [ * * , ( ) ] ,	renormalization group ( rg ) methods used to soften hamiltonians for nuclear many - body calculations change the effective resolution of the interaction . for nucleon knock - out processes , these rg transformations leave cross sections invariant , but initial - state wave functions , interaction currents , and final - state interactions are individually altered . this has implications for the factorization of nuclear structure and reactions . we use deuteron electrodisintegration as a controlled laboratory for studying how structure and reaction components are modified under rg evolution , without the complication of three - body forces or currents . the dependence of these changes on kinematics is explored .
[ [ section ] ] the maximal number of rational points that a ( smooth , geometrically irreducible ) curve of genus @xmath4 over a finite field @xmath5 can have , is denoted by @xmath6 . one has the estimate ( see @xcite ) @xmath7\ ] ] in which the notation @xmath8 $ ] for @xmath9 means the largest integer @xmath10 . the upper bound here is called the hasse - weil - serre bound . for @xmath11 , it is a classical result of deuring @xcite , @xcite that @xmath12 $ ] , except when @xmath13 with @xmath14 prime and @xmath15 odd and @xmath14 divides @xmath16 $ ] , in which case @xmath17 $ ] . for @xmath18 an explicit formula is due to j - p . he stated and proved the result during a course @xcite he gave at harvard university in 1985 ; a nice survey including some modifications of the original proof can be found in chapter 5 of the thesis @xcite . the final result is that if @xmath19 is a square and @xmath20 then @xmath21 $ ] . moreover @xmath22 - 2 $ ] and @xmath23 - 3 $ ] . in case @xmath19 is not a square , then also @xmath21 $ ] except when either @xmath24)>1 $ ] or @xmath19 can be written in one of the forms @xmath25 , @xmath26 or @xmath27 . in these remaining cases , one has that if @xmath28\geq\frac{\sqrt{5}-1}{2}$ ] then @xmath29 $ ] and if @xmath28<\frac{\sqrt{5}-1}{2}$ ] then @xmath29 - 1 $ ] . for @xmath30 no such result is known . the best known lower bounds in case @xmath31 and @xmath19 a power of @xmath32 or @xmath3 which is @xmath33 can be found in @xcite . in @xcite j - p . serre gives values of @xmath34 for @xmath35 and for @xmath36 . moreover he shows in @xcite that @xmath37 . hence we have the following table . @xmath38 the entries for @xmath39 are obtained using the technique from the current paper ; its main goal is to give lower bounds for @xmath34 by restricting ourselves to one specific family of curves of genus @xmath3 . [ [ section-1 ] ] let @xmath40 be a field of characteristic different from @xmath32 . the plane quartic @xmath41 given by @xmath42 is for @xmath43 with @xmath44 a geometrically irreducible , smooth curve of genus @xmath3 . the degree @xmath45 polynomials given here are fixed by the subgroup @xmath46 generated by @xmath47 and @xmath48 . the group @xmath49 is isomorphic to @xmath50 , the symmetric group on @xmath45 elements . hence @xmath49 is contained in the group of automorphisms of @xmath51 . for general @xmath52 the automorphism group of @xmath51 in fact equals @xmath49 . these curves occur in the classification of non - hyperelliptic genus @xmath3 curves with nontrivial automorphism group , as given in @xcite and in ( * ? ? ? * table 5.6 , pp . 63 - 64 ) . suppose @xmath53 . by @xmath54 we denote the elliptic curve given by the equation @xmath55 . if moreover @xmath56 then we write @xmath57 for the elliptic curve with equation @xmath58 . the relation with the curves @xmath51 is as follows . [ jacobian ] suppose @xmath40 is a field of characteristic different from @xmath32 and @xmath59 . then the jacobian of the curve @xmath60 given by @xmath42 is over @xmath40 isogenous to the product @xmath61 , where @xmath57 denotes the elliptic curve with equation @xmath58 . _ most of this is shown in @xcite ; one takes the quotient of @xmath41 by the involution @xmath62 . the resulting curve has genus @xmath63 and it admits an involution without any fixed points . taking the quotient again results in an elliptic curve , given by @xmath64 . the @xmath32-isogeny with kernel generated by @xmath65 maps this curve onto @xmath66 ( compare the formulas for @xmath32-isogenies as given in ( * ? ? ? 4 ) ) . write @xmath67 for the composition of all maps described here . then @xmath68 where @xmath47 is one of the automorphisms of @xmath51 . the fact that @xmath69 induces an isomorphism between the spaces of regular @xmath63-forms implies that @xmath69 induces an isogeny between @xmath70 and the triple product of @xmath57 . [ comparetraces ] with notations as above , one finds for @xmath71 with @xmath19 odd and @xmath72 that @xmath73 _ proof . _ it is a well known fact that @xmath74 equals @xmath75 , where @xmath76 is the trace of frobenius acting on a tate module of @xmath70 . lemma [ jacobian ] implies that this tate module is isomorphic to a direct sum of three copies of the tate module of @xmath66 . hence @xmath77 where @xmath78 is the trace of frobenius on the tate module of @xmath66 . since this trace equals @xmath79 , the result follows . [ [ strategy ] ] our strategy for finding a curve of genus @xmath3 over a finite field @xmath5 with odd characteristic should now be clear : find @xmath52 such that @xmath80 is as large as possible and use corollary [ comparetraces ] . this works quite well for small @xmath19 , using a direct search . in fact , as will be explained in section [ charp ] below , it is not even necessary here to calculate @xmath80 for many values @xmath71 . we obtain a general result when the characteristic of @xmath5 equals @xmath3 , because in that case we deal with a curve @xmath81 which is isomorphic to the curve @xmath82 with @xmath83 . since it is precisely known which values @xmath84 attains ( see @xcite and also section [ charlegendre ] below ) , one obtains a nice explicit lower bound for @xmath85 . in fact , the result implies that the difference between @xmath85 and the hasse - weil - serre bound is bounded independently of @xmath86 : [ resultchar3 ] for every @xmath87 the inequality @xmath88-n_{3^n}(3)\leq\left\ { \begin{array}{cl } 0 & \mbox { if } \ ; n\equiv 2\bmod 4;\\ 12 & \mbox { if } \ ; n\equiv 0\bmod 4;\\ 21 & \mbox { if } \ ; n\equiv 1\bmod 2\end{array}\right.\ ] ] holds . for the proof we refer to section [ char3 ] . note that this proves a special case of a conjecture of j - p . serre @xcite , which says that for _ all _ @xmath19 the difference @xmath2-n_q(3)$ ] should be bounded independently of @xmath19 . [ [ section-2 ] ] in characteristic at least @xmath89 we have not been able to obtain a general result such as given in proposition [ resultchar3 ] . however , the fact that a curve @xmath66 is either isomorphic to @xmath90 or it is a quadratic twist of @xmath90 , implies ( again using @xcite ) that for every finite field @xmath5 of odd characteristic , a curve @xmath51 as above exists for which @xmath91 is at most @xmath92 off from either the hasse - weil - serre upper bound @xmath2 $ ] , or from the analogous lower bound @xmath93 $ ] . this is proven in section [ charp ] . we note that a sharper result of the same kind ( with @xmath92 replaced by @xmath3 ) was obtained by kristin lauter @xcite , @xcite using an entirely different method . as everett howe pointed out to us , it is in fact possible to improve our result slightly by replacing the product @xmath94 we use , by a product @xmath95 in which @xmath96 is an elliptic curve with a rational point of order @xmath32 and @xmath97 maximal under that condition , and @xmath98 is a legendre elliptic curve over @xmath5 with as many rational points as possible . the result of everett howe , franck leprvost and bjorn poonen ( * ? ? ? * prop . 15 ) in this case implies that either this product or its standard quadratic twist is isogenous over @xmath5 to the jacobian of a smooth genus @xmath3 curve over @xmath5 . it may be noted that the estimate obtained in this way is in general still weaker than lauter s result ( it replaces our @xmath92 by @xmath99 instead of by @xmath3 ) . suppose @xmath100 is a field of characteristic @xmath101 , and @xmath102 is an elliptic curve . we will say that @xmath102 is a legendre elliptic curve over @xmath100 if there is a @xmath53 in @xmath100 such that @xmath103 is over @xmath100 isomorphic to @xmath90 given by @xmath55 . a necessary but in general not sufficient condition for an elliptic curve @xmath102 to be a legendre elliptic curve over @xmath100 is that all points of order @xmath32 on @xmath103 are @xmath100-rational . an intrinsic description of legendre elliptic curves is given as follows . take a separable closure @xmath104 of @xmath100 and write @xmath105 for its galois group . [ legendrecriterion ] the statements 1 . @xmath103 is a legendre elliptic curve over @xmath100 ; 2 . @xmath103 can be given by an equation @xmath106 in which at least one of @xmath107 is a square in @xmath108 ; 3 . @xmath103 has all its points of order @xmath32 rational over @xmath100 , and there exists a point @xmath109 $ ] such that @xmath110 is not in the @xmath111-orbit of @xmath112 . are equivalent . _ proof_. the equivalence of ( 1 ) and ( 2 ) is easy . to verify that ( 2 ) and ( 3 ) are equivalent , suppose ( after possibly permuting @xmath113 ) that @xmath114 is a square and that @xmath103 is given by @xmath106 . the point @xmath115 in @xmath116 has order @xmath32 , and the quotient @xmath117 admits an isogeny of degree @xmath32 : @xmath118 defined over @xmath100 ( the dual isogeny of the quotient map ) . a very well known property ( compare ( * ? ? ? * iii 5 ) ) of @xmath119 is that the image @xmath120 equals the kernel of the homomorphism @xmath121 defined by @xmath122 and @xmath123 for all @xmath124 with @xmath125 . hence the condition that @xmath114 be a square is equivalent with the property that the point @xmath126 is in the image of @xmath127 . this means precisely that a pair of points @xmath128 exists which is @xmath111-stable , and @xmath129 . hence @xmath112 is a point of order @xmath45 on @xmath103 , and for all @xmath130 we have @xmath131 . in particular @xmath132 , which means @xmath133 for all @xmath130 . vice versa , suppose given a point @xmath112 of order @xmath45 with the property @xmath133 for all @xmath130 . since all @xmath32-torsion of @xmath103 is @xmath100-rational , we have that @xmath134 $ ] and moreover the condition @xmath133 implies that @xmath135 is in a cyclic subgroup of @xmath136 $ ] which is independent of @xmath137 . hence we have points @xmath138 and @xmath139 of order @xmath32 , where @xmath140 is @xmath111-stable . as we have seen , this implies the statements ( 1 ) and ( 2 ) . [ legfrob ] suppose @xmath19 is a power of an odd prime and @xmath141 is an elliptic curve . let @xmath142 be the frobenius endomorphism ( raising coordinates to the power @xmath19 ) . then @xmath103 is a legendre elliptic curve over @xmath1 if and only if @xmath143 but @xmath144 . _ proof_. the galois group @xmath145 is topologically generated by the @xmath19-th power map , and this generator acts on @xmath103 via the endomorphism @xmath146 . the condition @xmath143 is equivalent with the statement that @xmath103 has all its points of order @xmath32 rational over @xmath5 . in the same manner , the condition @xmath144 precisely means that a point @xmath112 of order @xmath45 exists , with the property @xmath147 . since the galois group @xmath145 acts on @xmath148 $ ] via a ( cyclic ) subgroup of the kernel of @xmath149 , it follows that @xmath133 for all @xmath150 . hence lemma [ legendrecriterion ] implies that @xmath103 is a legendre elliptic curve over @xmath5 . vice versa , if @xmath103 is a legendre elliptic curve over @xmath5 , then by lemma [ legendrecriterion ] we know that @xmath151 $ ] exists with @xmath147 , which implies that @xmath152 is not divisible by @xmath45 in @xmath153 . we have that @xmath143 since @xmath146 acts trivially on all points of order @xmath32 . this proves the corollary . [ atresult ] an elliptic curve @xmath96 ( with @xmath19 odd ) for which @xmath154 is isogenous to a legendre elliptic curve over @xmath5 , except in the following case : @xmath155 with @xmath156 , and @xmath157 . _ proof_. ( this result was first presented in @xcite , however , with a somewhat different proof . the present proof is more conceptual , but it gives less information concerning the possible values of legendre parameters @xmath52 in the supersingular case . ) let @xmath142 be the ( @xmath19-th power ) frobenius . the proof considers two cases . first , suppose @xmath158 . then @xmath159 and @xmath160 . any curve @xmath161 isogenous to @xmath103 then also satisfies @xmath162 and frobenius in @xmath163 is equal to @xmath164 . by corollary [ legfrob ] , one ( and equivalently , all of them ) such curve @xmath161 is a legendre elliptic curve over @xmath5 precisely when @xmath165 is even , but not divisible by @xmath45 . the latter condition is equivalent with @xmath166 . this proves the statement in the case @xmath167 . if @xmath168 then @xmath169\subset{\operatorname{end}}(e)$ ] is an order in the ring of integers of an imaginary quadratic field @xmath100 . we have that @xmath170 where the bar denotes complex conjugation in @xmath100 . the condition @xmath171 implies that @xmath172 is integral . now consider the order @xmath173 $ ] . by construction , @xmath174 satisfies @xmath175 and @xmath176 . it is a result of waterhouse ( * ? ? ? 4.5 ) ( compare @xcite where a mistake in the original result is corrected ) , that a curve @xmath177 exists with an isomorphism @xmath178 such that under this isomorphism frobenius on @xmath161 corresponds to @xmath174 . this implies in particular that @xmath179 and hence @xmath161 and @xmath103 are isogenous . moreover , using corollary [ legfrob ] we know that @xmath161 is a legendre elliptic curve over @xmath5 . this proves the proposition . we will now prove proposition [ resultchar3 ] . take @xmath87 and write @xmath180 , @xmath181 $ ] and @xmath182 with @xmath183 and @xmath184 . as explained in ( [ strategy ] ) , we will examine how close to the upper bound @xmath185 the number of @xmath5-points on a legendre elliptic curve @xmath186 can be , for @xmath71 . if @xmath86 is odd and moreover @xmath187 ( the smallest @xmath86 where this is the case , is @xmath188 which gives @xmath189 and @xmath190 ) , then we replace @xmath191 by @xmath192 . the resulting number @xmath191 satisfies @xmath193 , and moreover we know from @xcite that @xmath96 exists with @xmath194 . if @xmath86 is odd , then proposition [ atresult ] implies the existence of @xmath71 with @xmath195 . hence corollary [ comparetraces ] yields a genus @xmath3 curve @xmath41 with @xmath196 . in particular , this shows that @xmath197 for odd @xmath86 ( in fact , even @xmath198 unless @xmath199 is divisible by @xmath3 ) . if @xmath86 is even , then @xmath200 and ( again using deuring s results @xcite ) an elliptic curve @xmath96 exists with @xmath201 . by proposition [ atresult ] , this number of points occurs for a legendre elliptic curve only in case @xmath202 , i.e. , when @xmath203 . hence under this condition we obtain a curve @xmath41 whose number of points attains the hasse - weil - serre bound . in the remaining case we have @xmath204 . here the number @xmath185 does not occur as @xmath205 , for any @xmath71 . hence we take the largest smaller possibility , which is @xmath206 . proposition [ atresult ] implies that a legendre elliptic curve with this number of points over @xmath5 indeed occurs . it follows that a genus @xmath3 curve @xmath207 exists with @xmath208 . this implies the inequality given in proposition [ resultchar3 ] . the problem which arises when one attempts to adapt the argument presented in section [ char3 ] to finite fields of characteristic @xmath209 , can already be seen in the following result . as we mentioned in the introduction , a somewhat stronger result has been obtained by kristin lauter @xcite , @xcite using quite different techniques . moreover a variant of our proof may be obtained by using a result of everett howe , franck leprvost and bjorn poonen ( * ? ? ? _ write @xmath182 with @xmath183 and @xmath213 . then one of @xmath214 occurs as the number of points on a legendre elliptic curve @xmath98 , except possibly when @xmath215 and @xmath14 divides @xmath216 . in that case , we replace @xmath191 by @xmath217 and we obtain a number of points which does occur . this gives us an elliptic curve @xmath90 . the corresponding curve @xmath218 has either @xmath191 or @xmath219 points , or in case @xmath220 is not a square in @xmath5 this number is @xmath221 or @xmath222 . since this number is at distance at most @xmath223 from one of @xmath224 , corollary [ comparetraces ] implies the result . _ take @xmath228 . since @xmath225 , the elliptic curve @xmath229 is supersingular . this implies @xmath230 ( in case @xmath231 , this follows from the fact that the number of points is a multiple of @xmath45 , and also of course from a direct calculation ) . one concludes that @xmath232 . since @xmath233 as an element of @xmath234 is a square in @xmath5 , the two curves @xmath90 and @xmath235 are isomorphic over @xmath5 . corollary [ comparetraces ] therefore yields that the genus @xmath3 curve @xmath41 attains the hasse - weil - serre bound over @xmath5 . in practice , a fairly efficient method to find @xmath71 for which @xmath236 equals a given number @xmath237 can be given in case @xmath238 a prime or @xmath239 the square of a prime . this works as follows . write @xmath240 . we first treat the case @xmath239 and @xmath241 . exactly one of the two numbers @xmath242 occurs as a number of points of a legendre elliptic curve over @xmath243 , and this number is attained in our family precisely for the supersingular @xmath56 such that @xmath220 is a square in @xmath243 ; the number with the opposite choice of sign occurs for the ones such that @xmath220 is a nonsquare . in the remaining cases , the hasse inequality tells us @xmath244 . hence we find @xmath76 exactly if we know @xmath245 . now @xmath246 is already known , hence it suffices to find a @xmath52 such that @xmath247 . if we write @xmath248 for the nontrivial character with kernel @xmath249 , this means we look for @xmath56 such that @xmath250 . it is well known ( * ? ? ? 4 ) that @xmath251 , with @xmath252 if @xmath238 and @xmath253 if @xmath239 . here @xmath254 is the so - called hasse polynomial , whose coefficients can be computed using an easy recursion . hence in case @xmath238 we have to solve @xmath255 for @xmath256 , and then check whether @xmath257 has the correct value . similarly , when @xmath239 we look for solutions in @xmath5 of @xmath258 . this works reasonably efficient for -say- @xmath259 . using the package kant , we tested which values @xmath171 occur for the curves @xmath260 , for all odd @xmath261 . it turns out that for most @xmath19 , all values are attained . in the table below , we list all @xmath261 where this is _ not _ the case , and for each of them the missing value(s ) @xmath97 . as can be seen from the data , usually there is only one such missing value , which moreover is always one of the minimal or the maximal possible number @xmath262 in the interval @xmath263,q+1+[2\sqrt{q}]\right]$ ] . we list a sign @xmath264 indicating which of these possibilities occurs for a missing value . there are exactly two exceptional cases for @xmath261 . the first one is @xmath265 : here two values do nt occur , namely the maximum @xmath266 and also @xmath267 . the other one is @xmath268 . the two values missing here are the minimal one @xmath269 and also @xmath270 . the following table gives all other @xmath261 with a missing value . @xmath271 the table shows that for all but @xmath272 values @xmath261 , the maximal value of @xmath273 equals @xmath274 , where @xmath275 is the maximal number of points of an elliptic curve over @xmath5 with all its points of order @xmath32 rational . whenever the hasse - weil - serre bound is divisible by @xmath45 , we may be in the lucky circumstance that it is reached using our family of curves . this happens quite frequently , for instance when @xmath19 equals any of the primes @xmath276 , @xmath277 , @xmath278 . in the case @xmath279 the bound @xmath280 is a multiple of @xmath45 , but as the table above shows , our curves do not attain it . the data seems to indicate that for much more than @xmath281 of the prime powers @xmath19 , all possible values @xmath282 are attained by the family @xmath283 . moreover , the only occurrences of a @xmath19 for which more than one value is missing , happened at ` high ' even powers of a prime number . we have no theoretical explanation for this . a numerical test over all @xmath284 revealed exactly one more case where two values are missing , namely at @xmath285 . we have also not been able to explain why in all cases where we found that exactly one value is missing , this missing value is one of the maximal or minimal number of points . xxxx r. auer and j. top , _ legendre elliptic curves over finite fields , _ accepted for publ . in j. number theory , 2001 . m. deuring , _ die typen der multiplicatorenringe elliptischer funktionenkrper _ hamburg * 14 * ( 1941 ) , 197272 . g. van der geer and m. van der vlugt , _ tables for the function @xmath6 , _ available from http://www.science.uva.nl/~geer/ p. henn , _ die automorphismengruppen der algebraischen funktionenkrper vom geschlecht @xmath3 . _ inaugural dissertation , heidelberg , 1976 . e. w. howe , f. leprvost and b. poonen , _ large torsion subgroups of split jacobians of curves of genus two or three , _ forum math . * 12 * ( 2000 ) , 315364 . k. lauter , _ genus three curves over finite fields _ , notes of a lecture , available from http://msri.org/publications/ln/msri/2000/crypto/lauter/1/ k. lauter , _ the maximum or minimum number of rational points on curves of genus three over finite fields , _ ( with an appendix by j - p . serre ) , available from http://arxiv.org/abs/math.ag/0104086 , accepted for publication in compos . r. schoof , _ nonsingular plane cubic curves over finite fields , _ j. combin . theory ser . a * 46 * ( 1987 ) , 183211 . serre , _ sur le nombre des points rationnels dune courbe algbrique sur un corps fini _ , c. r. acad . paris sr . i * 296 * ( 1983 ) , 397402 . serre , _ rsum des cours de 19831984 _ , annuaire du collge de france ( 1984 ) , 7983 . reprinted in vol . 3 of jean - pierre serre , vres , collected papers . new york , etc . : springer - verlag , 1985 . serre , rational points on curves over finite fields , lectures given at harvard university , 1985 . notes by f. q. gouva . v. shabat , _ curves with many points . thesis , amsterdam , 2001 . j. h. silverman , the arithmetic of elliptic curves . new york , etc . : springer - verlag , gtm 106 , 1986 . j. h. silverman and j. tate , rational points on elliptic curves . new york , etc . : springer - verlag , 1992 . j. top , _ hecke l - series related with algebraic cycles or with siegel modular forms . thesis , utrecht , 1989 . a. m. vermeulen , _ weierstrass points of weight two on curves of genus three . thesis , amsterdam , 1983 . w. c. waterhouse , _ abelian varieties over finite fields _ , ann . sci . cole norm . * 2 * ( 1969 ) , 521560 .	we explain a naive approach towards the problem of finding genus 3 curves @xmath0 over any given finite field @xmath1 of odd characteristic , with a number of rational points close to the hasse - weil - serre upper bound @xmath2 $ ] . the method turns out to be successful at least in characteristic @xmath3 .

Subsets and Splits

No saved queries yet

Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.